Information processing method and information processing system

Abstract

An information processing method according to an embodiment of the present technology includes predicting, on the basis of a feature amount of a baseball game up to a first timing, pitch-sequencing information relating to pitch sequencing at a second timing subsequent to the first timing. On the basis of the predicted pitch-sequencing information, explanation information indicating the validity of the prediction of the pitch-sequencing information is generated. Thus, it is possible to provide high-quality information relating to pitch-sequencing predictions.

Description

Information Processing Method and Information Processing System

[0001] The present technology relates to an information processing method and an information processing system applicable to the field of baseball.

[0002] Patent Document 1 discloses a pitch prediction display system that accepts pitch predictions from viewers of a baseball broadcast. In this pitch prediction display system, the 3-row 3-column operation buttons of a television remote control or a mobile phone are made to correspond to a strike zone divided into 3 rows and 3 columns. This allows a user to intuitively grasp the relationship between the operation buttons and positions on the strike zone, facilitating the transmission of pitch predictions.

[0003] Patent Document 2 discloses a baseball pitch analysis system. This pitch analysis system stores the types of pitches a pitcher has thrown in the past, and calculates the pitching probability of each pitch type with respect to the total number of pitches. This makes it possible to obtain the pitching probability of each pitch type with high accuracy.

[0004] Japanese Patent Application Laid-Open No. 2010-115267 Japanese Patent Application Laid-Open No. 2016-52460

[0005] As described above, there is a demand for technology that enables provision of high-quality information related to pitch prediction.

[0006] In view of the above circumstances, an object of the present technology is to provide an information processing method and an information processing system that enable provision of high-quality information related to pitch prediction.

[0007] In order to achieve the above object, an information processing method according to one aspect of the present technology includes predicting pitch information related to a pitch at a second timing after a first timing, based on feature quantities of the game up to the first timing in baseball. Explanation information indicating the validity of the prediction of the pitch information is generated based on the predicted pitch information.

[0008] This information processing method predicts pitching information for a second timing point in a baseball game based on the characteristics of the game up to the first timing point. Furthermore, explanatory information demonstrating the validity of the prediction is generated based on the predicted pitching information. This makes it possible to provide high-quality information regarding pitching predictions.

[0009] The first timing may be the present moment. In this case, the second timing is the timing at which the next pitching sequence will occur, relative to the present moment.

[0010] The aforementioned pitching information may include a prediction of the actual pitching sequence that will be performed.

[0011] As the pitching information, a pitching sequence that is advantageous or disadvantageous to the first team performing the pitching sequence may be predicted.

[0012] The pitching information may include at least one of the pitching course, pitch type, ball speed, spin rate, or spin axis.

[0013] The aforementioned feature may include at least one of the following: the score of the first team that executes the pitching; the score of the second team, which is the opponent of the first team; the pitcher of the first team or his earned run average; the batter of the second team or his batting average; the date; the stadium; the strike count; the ball count; the out count; the wind direction; or the defensive shift.

[0014] If the pitches actually performed up to the first timing are designated as historical pitches, the explanatory information may be generated by designating historical pitches similar to the predicted pitch information as explanatory information.

[0015] At the second timing described above, it may be predicted whether or not a predetermined tactic should be executed.

[0016] The aforementioned predetermined tactics may include pitching changes or pinch-hitting.

[0017] The suitability of each player for executing the aforementioned pitching change or pinch-hitting may also be predicted.

[0018] At a third timing after the second timing, the system may accept the user's specification of the pitching sequence. In this case, the system may predict the result if the pitching sequence specified at the second timing had been performed.

[0019] As the pitching information, a pitching sequence that is advantageous to the first team performing the pitching sequence may be predicted. In this case, among the second timings, timings in which the first team concedes a run and the predicted advantageous pitching sequence is dissimilar to the actual pitching sequence may be extracted.

[0020] The display of the predicted pitching information and the generated explanatory information to the user may be controlled.

[0021] The aforementioned prediction of pitch selection information may be performed using a machine learning model.

[0022] An information processing system according to one embodiment of this technology comprises a prediction unit and a generation unit. The prediction unit predicts pitching information regarding pitching at a second timing after the first timing, based on the characteristic quantities of the game up to a first timing in baseball. The generation unit generates explanatory information indicating the validity of the prediction of pitching information based on the pitching information predicted by the prediction unit.

[0023] The prediction unit may predict the actual pitching sequence as the pitching sequence information.

[0024] The prediction unit may predict, as pitching information, a pitching sequence that is advantageous or disadvantageous to the first team performing the pitching sequence.

[0025] The prediction unit may predict whether or not a predetermined tactic should be executed at the second timing.

[0026] The prediction unit may accept a user's specification of the pitching sequence at a third timing after the second timing. In this case, it may predict the result if the pitching sequence specified at the second timing had been performed.

[0027] The information processing system may further include a display control unit that controls the display to the user of the ball distribution information predicted by the prediction unit and the explanatory information generated by the generation unit.

[0028] This is a schematic diagram showing an example of a UI (User Interface) displayed on a display device. This is a schematic diagram showing an example of the configuration of an information processing system. This is a schematic diagram showing an example of information stored in the memory unit. This is a schematic diagram showing the contents of the pitching sequence log and the pitching sequence log. This is a flowchart showing an example of the learning process. This is a schematic diagram showing the learning content of the predicted pitching sequence model. This is a schematic diagram showing the learning content of the recommended pitching sequence model. This is a flowchart showing an example of the prediction process and the explanatory information generation process. This is a schematic diagram showing the prediction content of the predicted pitching sequence. This is a schematic diagram showing the generation content of the explanatory information. This is a schematic diagram showing an example of a UI related to if-pitcher changes. This is a schematic diagram showing an example of a UI related to if-pinch hitters. This is a schematic diagram showing an example of the configuration of a functional block according to this embodiment. This is a schematic diagram showing an example of information stored in the memory unit. This is a flowchart showing an example of the learning process. This is a schematic diagram showing the contents of the pitching sequence base log and the pitching sequence base quick report log. This is a schematic diagram showing the contents of the pitching sequence detail log and the pitching sequence detail quick report log. This is a schematic diagram showing the contents of the pinch hitter base log and the pinch hitter base quick report log. This is a schematic diagram showing the contents of the pinch hitter detail log and the pinch hitter detail quick report log. This is a flowchart showing an example of prediction processing related to pitching changes. This is a flowchart showing an example of prediction processing related to pinch hitting. This is a schematic diagram showing an example of UI related to simulation. This is a schematic diagram showing an example of the configuration of a functional block according to this embodiment. This is a schematic diagram showing an example of information stored in the memory unit. This is a flowchart showing an example of learning processing. This is a flowchart showing an example of simulation processing. This is a block diagram showing an example of the hardware configuration of a computer capable of realizing an information processing device.

[0029] <First Embodiment> Hereinafter, embodiments relating to the present technology will be described with reference to the drawings.

[0030] [User Interface] Figure 1 is a schematic diagram showing an example of a UI (User Interface) 2 displayed on display device 1. In this technology, UI 2 as shown in Figure 1 is displayed on display device 1, and the user views the displayed UI 2. This UI 2 shows information about a baseball game currently in progress. Here, baseball games include not only professional baseball official games but also amateur baseball games, etc.

[0031] Display device 1 is, for example, a smartphone screen, but it may also be a television, a PC (Personal Computer), a tablet, or the screen of a terminal dedicated to this technology. Furthermore, the television broadcast or internet broadcast of the currently ongoing match may be displayed on display device 1, and the UI 2 related to this technology may be displayed simultaneously on the same screen.

[0032] UI2 includes a score display unit 3, a history ball distribution display unit 4, a ball distribution information display unit 5, an explanatory information display unit 6, and a comment display unit 7. In this example, these are arranged in this order from top to bottom, but the specific position of each display unit is not limited. Also, the information displayed in each display unit is not limited to that in this example, and display units that display other content may be included.

[0033] The score display unit 3 shows the current date, the name of the stadium where the game is being played, the start time of the game, the score of both teams (Team A and Team B) for each inning, the current score of both teams, the number of hits, and the number of errors.

[0034] The history pitch sequence display unit 4 shows the history of pitch sequences in the current at-bat. In this example, four pitch sequences have already been performed in the at-bat, and the speed, type of pitch, and result of each sequence are displayed. In other words, the fifth sequence is about to be performed. The history pitch sequence display is updated in real time each time a new sequence is performed. Note that a sequence of pitches is sometimes referred to as a pitch.

[0035] Pitch types include fastballs, as well as breaking balls such as sliders, curves, forks, changeups, sinkers, and shuuto, and other types of breaking balls may also be included. Results include swings and misses, called strikes, called balls, fouls, walks, hit by pitch, ground outs such as shortstop and second base, fly outs such as outfield flies, infield flies, double plays, triple plays, singles, doubles, triples, home runs, etc.

[0036] The pitching information display unit 5 displays a heatmap 8, pitching icons 9, and a scoreboard 10. The heatmap 8 shows the current batter's batting average for each pitch location. In this example, the locations are divided into 5x5 blocks, and the gray color of the block increases with the batting average. The heatmap 8 is not limited to 5x5; it may be divided into any number of blocks. Alternatively, the batting average may be displayed continuously without being divided into blocks. The heatmap 8 includes a frame indicating the strike zone.

[0037] In this example, batters recorded high batting averages when the pitch was placed near the center of the course, for example, in the 3rd row, 2nd to 4th column. On the other hand, batting averages were low when the pitch was placed outside the strike zone, for example, in the 2nd row, 1st column.

[0038] In this example, the batter's batting average is displayed as heatmap 8, but depending on the operation of the toggle button located below it, the batting average against by pitcher's location, the batter's slugging average by location, etc., will be displayed as heatmap 8.

[0039] The pitch selection icon 9 includes past pitch selection icons 9a, predicted pitch selection icons 9b, recommended pitch selection icons 9c, and bad pitch selection icons 9d. In this example, one of the past pitch selection icons 9a is represented by a symbol. Each pitch selection icon 9 is displayed on the heatmap 8. The position of the pitch selection icon 9 indicates the course of the pitch, the number indicates which pitch selection it is in the at-bat, and the shape and pattern indicate the type of pitch. In addition, the pitch selection icon 9 may include information such as ball speed, spin rate, spin axis, and result.

[0040] The past pitch icons 9a are icons indicating pitches thrown up to the present time. In this example, by referring to the No. 1 past pitch icon 9a, it can be read that the first pitch was thrown at the position of the 4th row and 4th column, and the pitch type was a fastball. Furthermore, by referring to the No. 2 to No. 4 past pitch icons 9a, it is possible to read the course and pitch type for each of the thrown pitches.

[0041] The predicted pitch icon 9b, recommended pitch icon 9c, and bad pitch icon 9d are icons each indicating a pitch predicted by the system. These are displayed distinctively from each other by using different solid lines and broken lines for their edge portions.

[0042] In this example, pitch information regarding the pitch to be thrown at the next timing of pitching with reference to the present (hereinafter may be referred to as the next pitch) is predicted, and displayed as the predicted pitch icon 9b, recommended pitch icon 9c, and bad pitch icon 9d. Here, since the present time is immediately after the 4th pitch has been thrown, the next pitch means the 5th pitch. In fact, the number 5 is displayed on these icons.

[0043] The predicted pitch icon 9b, recommended pitch icon 9c, and bad pitch icon 9d correspond to an embodiment of pitch information relating to the next pitch. The past pitch icons 9a correspond to an embodiment of pitch information relating to pitches thrown up to the present time. Therefore, the pitch icons 9 as a whole correspond to an embodiment of pitch information up to the next pitch. The present time corresponds to an embodiment of the first timing according to the present technology. The next pitch corresponds to an embodiment of the second timing according to the present technology.

[0044] The predicted pitch icon 9b is an icon indicating a prediction result of a pitch that will actually be thrown. In this example, it is predicted that a fastball will be thrown to the course of the 3rd row and 2nd column in the next 5th pitch.

[0045] The recommended pitching icon 9c is an icon that indicates a prediction result of pitching advantageous to the team that is to perform the pitching. Since the current defending team is Team B, the team that is to perform the pitching is Team B. Therefore, the recommended pitching icon 9c indicates a prediction result of pitching advantageous to Team B. That is, a pitch that has a high probability of retiring the batter from Team A by means of a strike, foul, ground out, fly out or the like is predicted and displayed as the recommended pitching icon 9c.

[0046] Note that the content of "advantageous pitching" may change depending on the game situation. For example, pitching that may result in a single hit or a walk is normally disadvantageous pitching, but in a situation where a power hitter is at bat, such pitching may sometimes be considered advantageous. That is, even if pitching is disadvantageous when viewed on an at-bat basis, any pitching that can increase the probability of Team B winning when viewed from the perspective of the entire game can be "advantageous pitching". It can also be said that pitching advantageous to Team B is pitching disadvantageous to Team A.

[0047] In this example, in the next fifth pitch, pitching a slider to the 4-row 2-column course is predicted to be advantageous pitching for Team B. Team B corresponds to an embodiment of the first team that performs pitching according to the present technology.

[0048] The bad pitching icon 9d is an icon that indicates a prediction result of pitching disadvantageous to the team that is to perform the pitching. Therefore, the bad pitching icon 9d indicates a prediction result of pitching disadvantageous to Team B. That is, a pitch that has a high probability of allowing the batter from Team A to get a hit or a home run is predicted and displayed as the bad pitching icon 9d. In this example, in the next fifth pitch, pitching a straight to the 4-row 4-column course is predicted to be disadvantageous pitching for Team B.

[0049] The scoreboard 10 includes displays of the inning number, runner information, the scores of both teams, the ball-strike-out count, the pitcher name, and the batter name.

[0050] The explanatory information display unit 6 displays the predicted pitching sequence explanation 11a, the recommended pitching sequence explanation 11b, and the bad pitching sequence explanation 11c as explanatory information. A checkbox is also placed on the explanatory information display unit 6. Currently, "AI pitching sequence details" is checked, so the predicted pitching sequence explanation 11a, etc. are displayed, but if a check is entered for another item, the corresponding other information will be displayed in the position of the explanatory information display unit 6.

[0051] The predicted pitching sequence explanation 11a, the recommended pitching sequence explanation 11b, and the bad pitching sequence explanation 11c each include the same heatmap 8 and pitching sequence icon 9 as those of the pitching sequence information display unit 5 (symbols are not shown). The contents of these heatmaps 8 may also be switched in conjunction with the contents of the heatmap 8 of the pitching sequence information display unit 5 in response to the operation of the switching button on the pitching sequence information display unit 5.

[0052] In this embodiment, when the pitching sequence that has actually been performed to date is designated as the historical pitching sequence, the pitching sequence icon 9 in the predicted pitching sequence explanation 11a is displayed as the historical pitching sequence that is similar to the pitching sequence icon 9 in the pitching sequence information display unit 5.

[0053] In this example, the first pitch in the history of pitch selections displayed in the predicted pitch selection explanation 11a is on a 4x4 course, and the pitch type is a fastball. On the other hand, the first pitch in the pitch selection information display unit 5 is also on a 4x4 course, and the pitch type is a fastball. In other words, these pitch selections are similar.

[0054] Similarly, the second pitch sequence in the predicted pitch sequence explanation 11a is similar to the fourth pitch sequence in the pitch sequence information display unit 5. The third pitch sequence in the predicted pitch sequence explanation 11a is similar to the third pitch sequence in the pitch sequence information display unit 5. And the fourth pitch sequence in the predicted pitch sequence explanation 11a is similar to the predicted pitch sequence icon 9b in the pitch sequence information display unit 5. Note that the former represents actual past pitch sequences, while the latter represents pitch sequences predicted by the system.

[0055] In this embodiment, the prediction shown by the prediction pitch icon 9b is made more convincing by displaying a historical pitch sequence similar to the pitch sequence status displayed on the pitch sequence information display unit 5 as the predicted pitch sequence explanation 11a. In other words, the predicted pitch sequence explanation 11a is information that indicates the validity of the prediction shown by the predicted pitch sequence icon 9b.

[0056] Below the predicted pitching sequence explanation 11a, the characteristic quantities (pitcher's name, catcher's name, batter's name, both teams' names) and the result of the pitching sequence related to the historical pitching sequence are displayed. Historical pitching sequences may be selected so that the characteristic quantities are similar to the current characteristic quantities. This makes it possible to demonstrate an even higher level of prediction validity.

[0057] Similarly, the pitching icon 9 in the recommended pitching explanation 11b displays a historical pitching sequence similar to the pitching icon 9 in the pitching information display unit 5. In fact, the pitching icon 9 in the recommended pitching explanation 11b is similar to the past pitching icon 9a and the recommended pitching icon 9c in the pitching information display unit 5. Furthermore, the result column in the recommended pitching explanation 11b shows a ground ball to second base, which is a favorable result for Team B. Therefore, based on these factors, the recommended pitching explanation 11b provides information that demonstrates the validity of the prediction shown by the recommended pitching icon 9c.

[0058] Similarly, the pitching icon 9 in the bad pitching explanation 11c displays a historical pitching sequence similar to the pitching icon 9 in the pitching information display unit 5. In fact, the pitching icon 9 in the bad pitching explanation 11c is similar to the past pitching icon 9a and bad pitching icon 9d in the pitching information display unit 5. Also, the result column in the bad pitching explanation 11c shows a home run, which is an unfavorable result for team B. Therefore, based on these factors, the bad pitching explanation 11c provides information that demonstrates the validity of the prediction shown by the bad pitching icon 9d.

[0059] The comment display section 7 shows comments posted by other users. It also displays the icons of other users, like and dislike buttons for comments, replies, and the comment date and time. Users can also post comments themselves. It is also possible to control the display of comments, such as showing only comments from fans of one team or comments from fans of other teams, via checkboxes.

[0060] [Functional Block] Figure 2 is a schematic diagram showing an example configuration of the information processing system 13. The information processing system 13 includes a smartphone 14 and an information processing device 15. The smartphone 14 is a terminal used by the user to use this system, and may be a television, PC, tablet, or a terminal dedicated to this technology. The smartphone 14 has the display device 1 shown in Figure 1, and an operation unit 17 and a communication unit 18.

[0061] Display device 1 is a display device using, for example, liquid crystal, electroluminescence (EL), etc., and displays various images, UI 2, etc. The operation unit 17 is a touch panel and is configured integrally with display device 1. If a television or PC is used instead of a smartphone 14, the operation unit 17 may be a remote control, keyboard, pointing device, or other operating device.

[0062] The communication unit 18 is a communication module for communicating with other devices via a network such as a LAN (Local Area Network) or WAN (Wide Area Network). It may be equipped with a communication module for short-range wireless communication such as Bluetooth®. Alternatively, communication equipment such as a modem or router may be used. In this embodiment, the communication unit 18 performs communication with the information processing device 15, etc.

[0063] Furthermore, the specific configuration of the smartphone 14 is not limited, and for example, it may have a built-in memory unit, speaker, etc.

[0064] The information processing device 15 is, for example, a PC or a server, and is configured on the cloud. The information processing device 15 has a communication unit 19, a storage unit 20, and a controller 21. These blocks are interconnected via a bus 22. Instead of the bus 22, each block may be connected using a communication network or a non-standardized proprietary communication method.

[0065] The communication unit 19 has a configuration similar to, for example, the communication unit 18 of a smartphone 14. In this embodiment, communication with the smartphone 14 and other devices, as well as communication with the baseball stadium 23 where the game is being played, is performed via the communication unit 19.

[0066] Figure 3 is a schematic diagram showing an example of information stored in the memory unit 20. The memory unit 20 is a storage device such as non-volatile memory, for example, an HDD (Hard Disk Drive) or SSD (Solid State Drive) may be used. In addition, any non-transient storage medium that can be read by a computer may be used. As shown in Figure 3, the memory unit 20 stores a pitching report log 25, a pitching log 26, a predicted pitching model 27a, a recommended pitching model 27b, a bad pitching model 27c, a model output 28, and explanatory information 29. In addition, programs and algorithms for executing this technology are also stored.

[0067] The controller 21 controls the operation of each block in the information processing device 15. The controller 21 has hardware circuits necessary for a computer, such as a CPU and memory (RAM, ROM). Various processes are executed when the CPU executes the program related to this technology stored in the storage unit 20 and memory. As the controller 21, devices such as FPGAs (Field Programmable Gate Arrays) or other PLDs (Programmable Logic Devices) or ASICs (Application Specific Integrated Circuits) may be used.

[0068] In this embodiment, the CPU of the controller 21 executes a program related to this technology (for example, an application program), thereby realizing the following functional blocks: a pitching log acquisition unit 32, a predicted pitching learning unit 33a, a recommended pitching learning unit 33b, a bad pitching learning unit 33c, a predicted pitching prediction unit 34a, a recommended pitching prediction unit 34b, a bad pitching prediction unit 34c, an explanatory information generation unit 35, and a display control unit 36.

[0069] The information processing method according to this embodiment is then executed by these functional blocks. Dedicated hardware such as ICs (integrated circuits) may be used as appropriate to implement each functional block. The processing performed by each functional block will be described below, but the details of the processing for some functional blocks may be explained in more detail later.

[0070] Figure 4 is a schematic diagram showing the contents of the pitching sequence log 25 and the pitching sequence log 26. The pitching sequence log acquisition unit 32 acquires the pitching sequence log 25 and the pitching sequence log 26 from the baseball field 23. Specifically, a server or the like is located on the baseball field 23 side, and the pitching sequence log 25 and the pitching sequence log 26 are stored on the server or the like. The pitching sequence log acquisition unit 32 receives the pitching sequence log 25 and the pitching sequence log 26 by communicating with the server or the like via the communication unit 19.

[0071] Figure 4 shows one pitching sequence log 25 and four pitching sequence logs 26 (26a to 26d) line by line. Each pitching sequence log 25 and each of the four pitching sequence logs 26 corresponds to one pitching sequence. The pitching sequence logs 26 are logs related to past pitching sequences that have already been performed, while the pitching sequence log 25 is a log related to the next pitching sequence. In this example, only the four pitching sequence logs 26 for the same day are shown, but in reality, for example, a large number of past pitching sequence logs 26 are stored in the storage unit 20. The pitching sequence log 26 includes feature quantities 39, pitching sequence information 40, and results 41. The pitching sequence log 25 includes only feature quantities 39.

[0072] Feature quantity 39 is typically a quantity related to a game that can change from game to game. As shown in the figure, in this embodiment, feature quantity 39 includes the scores of Team A and Team B when the pitches were thrown, the pitcher of Team A, the batter of Team B, the date, the stadium, the strike count, the ball count, the out count, and the wind direction.

[0073] Other parameters such as the earned run average of the pitchers of Team A, the batting average of the batters of Team B, defensive shifts, wind speed, number of innings, weather, and temperature may also be included in feature 39. Furthermore, some of the above parameters may not be included in feature 39. In this example, Team A corresponds to one embodiment of the first team, and Team B corresponds to one embodiment of the second team, which is the opponent of the first team.

[0074] The pitching information 40 is information that indicates the content of the pitches the pitcher has made. In this embodiment, the pitching information 40 includes information on the pitching course and type of pitch, similar to the pitching icon 9 in Figure 1. The pitching course is expressed in horizontal and vertical coordinates. The pitching information 40 also includes ball speed. In addition, any other information related to the pitching, such as spin rate or spin axis, may be included in the pitching information 40.

[0075] Result 41 is information indicating the result of the pitch selection. In this embodiment, Result 41 includes pitch selection results such as called strikes and ground balls to second base, similar to the results displayed on the history pitch selection display unit 4 in Figure 1.

[0076] A pitching log 26 is generated each time a pitch is thrown and the result is obtained. In other words, one row of the pitching log 26 is added to the table for each pitch. In this example, pitching logs 26a, 26b, 26c, and 26d represent the states in which the respective pitches were thrown.

[0077] On the other hand, the pitching sequence log 25 is a log related to the next pitching sequence, but the next pitching sequence has not yet taken place. Therefore, the pitching sequence log 25 contains only the current features and does not include the pitching sequence information 40 and the result 41. When the next pitching sequence is later taken, the pitching sequence information 40 and the result 41 will be added to the pitching sequence log 25, and the pitching sequence log 25 will be treated as a regular pitching sequence log 26.

[0078] The pitching log acquisition unit 32 generates a real-time pitching log 25 and stores it in the storage unit 20. When a pitch is made and a result is obtained, the pitching information 40 and the result 41 are added to the real-time pitching log 25 to generate a pitching log 26, which is also stored in the storage unit 20.

[0079] Although not shown in the diagrams, the pitching log 25 and pitching log 26 may include information on strategies or tactics such as pinch hitters, pitching changes, and stolen bases.

[0080] The predictive pitching sequence learning unit 33a trains a predictive pitching sequence model 27a to predict the actual pitching sequence. The recommended pitching sequence learning unit 33b trains a recommended pitching sequence model 27b to predict a pitching sequence that is advantageous to the team making the pitches. The bad pitching sequence learning unit 33c trains a bad pitching sequence model 27c to predict a pitching sequence that is disadvantageous to the team making the pitches. In other words, the predictive pitching sequence model 27a, the recommended pitching sequence model 27b, and the bad pitching sequence model 27c are all machine learning models, but there is no particular limitation on what type of machine learning model each of them is.

[0081] The predicted pitching sequence unit 34a predicts the actual pitching sequence using the predicted pitching sequence model 27a. The recommended pitching sequence unit 34b predicts favorable pitching sequences using the recommended pitching sequence model 27b. The unfavorable pitching sequence prediction unit 34c predicts unfavorable pitching sequences using the unfavorable pitching sequence model 27c. These prediction results are appropriately stored in the storage unit 20 as model output 28.

[0082] The explanatory information generation unit 35 generates explanatory information 29 that indicates the validity of the pitch selection prediction. The generated explanatory information 29 is stored in the storage unit 20 as appropriate.

[0083] The display control unit 36 ​​generates the pitching icons 9 shown in Figure 1 based on the predicted pitching and explanatory information 29, and transmits them to the smartphone 14 via the communication unit 19. As a result, the pitching icons 9 are displayed on the display device 1, and the display of the predicted pitching and explanatory information 29 to the user is controlled. The display control unit 36 ​​may also be configured in the smartphone 14 itself.

[0084] The predicted pitching sequence unit 34a, the recommended pitching sequence unit 34b, and the bad pitching sequence unit 34c correspond to one embodiment of the prediction unit according to this technology. The explanatory information generation unit 35 corresponds to one embodiment of the generation unit according to this technology.

[0085] [Learning Process] Figure 5 is a flowchart showing an example of the learning process. The process in this example involves pre-training the predictive pitching model 27a, the recommended pitching model 27b, and the bad pitching model 27c. Typically, the process in this example is executed when the user is not using the information processing system 13.

[0086] The pitching log 26 is saved (step 101). Specifically, the pitching log acquisition unit 32 acquires the pitching log 26 and stores it in the storage unit 20.

[0087] Figure 6 is a schematic diagram showing the learning process of the predictive pitching model 27a. The predictive pitching model 27a is trained (step 102). First, the predictive pitching learning unit 33a reads the pitching log 26 from the storage unit 20. Then, as shown in Figure 6, the feature quantities 39 contained in the pitching log 26 are used as input data and the pitching information 40 as ground truth data to perform supervised learning of the predictive pitching model 27a.

[0088] Since the pitching log 26 is a log of past pitching sequences, this learning process generates a predictive pitching model 27a that can predict the actual pitching sequences using the feature vector 39 as input. Note that the learning method in this example is just one example, and other learning methods may be used.

[0089] Figure 7 is a schematic diagram showing the learning process of the recommended pitching model 27b. The recommended pitching model 27b is trained (step 103). For example, the recommended pitching learning unit 33b performs offline reinforcement learning of the recommended pitching model 27b.

[0090] Specifically, feature quantities 39 are input to the recommended pitching model 27b, and pitching information 40 is output. The recommended pitching learning unit 33b includes an error calculation unit 44, which compares the outputted pitching information 40 with "past pitching information 40 in the situation of the feature quantities 39 input to the recommended pitching model 27b" using a loss function and calculates the error. Then, the recommended pitching model 27b is retrained based on the error.

[0091] This generates a recommended pitching model 27b that can predict advantageous pitching sequences using feature 39 as input. Note that the learning method in this example is just one example, and other learning methods such as offline contextual bandit may be used.

[0092] Next, the bad pitching model 27c is trained (step 104). The content of this training is generally the same as that shown in Figure 7, so the illustration is omitted. That is, the bad pitching learning unit 33c performs offline reinforcement learning using past pitching information 40, and generates a bad pitching model 27c that can predict unfavorable pitching using the feature quantities 39 as input.

[0093] In this example, steps 101 to 104 are executed sequentially, but these processes may be executed in a different order. The frequency of each process may also differ. For example, the ball distribution log may be saved each time a ball is distributed, and the learning process may be executed as a batch process, such as once a day. Furthermore, the frequency of the learning process may differ between models.

[0094] [Prediction Processing] Figure 8 is a flowchart showing an example of prediction processing and explanatory information generation processing. The processing in this example is executed once each time a pitch is thrown. A pitching report log 25 is saved (step 201). Specifically, the pitching report log 25 is acquired by the pitching log acquisition unit 32 and stored in the storage unit 20.

[0095] Figure 9 is a schematic diagram showing the content of the predicted pitching sequence. The predicted pitching sequence is predicted (step 202). Specifically, first the predicted pitching sequence prediction unit 34a reads the pitching sequence log 25 from the storage unit 20. Then, as shown in Figure 9, the feature quantities 39 contained in the pitching sequence log 25 are input to the trained predicted pitching sequence model 27a. As a result, the predicted pitching sequence model 27a outputs pitching sequence information 40 related to the predicted pitching sequence.

[0096] Recommended pitching sequences are predicted (step 203). The content of this process is generally the same as the process in Figure 9, so it is omitted from the illustration. That is, the recommended pitching sequence prediction unit 34b inputs the feature quantity 39 into the recommended pitching sequence model 27b, and the pitching sequence information 40 related to the recommended pitching sequence is output.

[0097] A bad pitch selection is predicted (step 204). The details of this process are also omitted from the illustration. The bad pitch selection prediction unit 34c inputs the feature quantity 39 into the bad pitch selection model 27c, and pitch selection information 40 related to the bad pitch selection is output.

[0098] [Description Information Generation Process] Figure 10 is a schematic diagram showing the content of the generation of the description information 29. The predicted pitching explanation 11a is generated (step 205). Specifically, the description information generation unit 35 generates the predicted pitching explanation 11a based on the predicted pitching predicted by the predicted pitching prediction unit 34a. More specifically, if the pitching that has actually been performed up to the present is considered the historical pitching, the predicted pitching explanation 11a is generated by using historical pitching that is similar to the predicted pitching as the predicted pitching explanation 11a.

[0099] In Figure 10A, the horizontal axis of the graph represents feature vector 39, and the vertical axis represents pitch selection. Predicted pitch selections are shown as black circles, and historical pitch selections are shown as white circles. Specifically, the position on the horizontal axis for predicted pitch selections represents the current feature vector 39, and the position on the vertical axis represents the content of the predicted pitch selection. Similarly, the position on the horizontal axis for historical pitch selections represents the feature vector 39 at the time the historical pitch selection occurred, and the position on the vertical axis represents the content of the historical pitch selection.

[0100] In this example, a two-dimensional value is calculated by converting each of the feature quantities 39 and the pitching sequence into a one-dimensional value according to some criterion. The calculation of the value is performed sequentially by the explanatory information generation unit 35 each time the pitching sequence log 25 or pitching sequence log 26 is acquired.

[0101] If a predicted ball distribution is predicted, in step 205, the explanatory information generation unit 35 calculates the distance between the predicted ball distribution and each historical ball distribution. Here, distance refers to the distance between the two-dimensional values ​​calculated above. In other words, the distance is calculated by summing the square of the difference between the values ​​related to feature quantity 39 and the square of the difference between the values ​​related to ball distribution, and taking the square root.

[0102] The calculated distances are then sorted, and the historical pitch sequence that minimizes the distance (illustrated by arrows) is selected as the predicted pitch sequence explanation 11a, thereby generating the predicted pitch sequence explanation 11a. In this example, historical pitch sequences similar to the predicted pitch sequence are selected in this way.

[0103] A recommended pitching sequence explanation is generated (step 206). The explanation information generation unit 35 generates a recommended pitching sequence explanation 11b based on the predicted recommended pitching sequence. Specifically, the recommended pitching sequence explanation 11b is generated by using a historical pitching sequence similar to the recommended pitching sequence as the recommended pitching sequence explanation 11b.

[0104] Figure 10B shows the contents of the generation process. In this example, similar to Figure 10A, the historical pitch selection that minimizes the distance from the recommended pitch selection is chosen. In this process, only historical pitch selections that resulted in a favorable outcome for the team making the pitch selection are extracted. For example, only historical pitch selections that resulted in an out are extracted. The specific content of a favorable outcome is not limited, and historical pitch selections that resulted in a strike, etc., may also be extracted. In Figure 10B, extracted historical pitch selections are shown with solid white circles, and non-extracted historical pitch selections are shown with dashed white circles.

[0105] In this example, there is a non-extracted historical pitch sequence that is closer to the recommended pitch sequence than any of the extracted historical pitch sequences. However, this non-extracted historical pitch sequence is not selected, and the extracted historical pitch sequence with the closest distance to the recommended pitch sequence is selected.

[0106] A bad pitching sequence explanation is generated (step 207). The explanation information generation unit 35 generates a bad pitching sequence explanation 11c based on the predicted bad pitching sequence. Specifically, the bad pitching sequence explanation 11c is generated by using a historical pitching sequence similar to a bad pitching sequence as the bad pitching sequence explanation 11c.

[0107] The content of this process is generally the same as that of the process shown in Figure 10B, so the illustration is omitted. In this process, only historical pitch selections that resulted in unfavorable outcomes for the team making the pitch selections are extracted. For example, only historical pitch selections that resulted in singles, doubles, triples, and home runs are extracted. The specific content of other unfavorable outcomes is not limited.

[0108] The method for generating the explanatory information 29 shown in Figures 10A and 10B is merely an example, and similar historical pitch selections may be performed by other methods. Furthermore, since the explanatory information 29 only needs to indicate the validity of the prediction of the pitch selection information 40, the explanatory information 29 may be generated by methods other than selecting historical pitch selections similar to the pitch selection information 40.

[0109] The display control unit 36 ​​controls the display of the predicted pitch distribution information 40 and the generated explanatory information 29 (step 208). As a result, the display shown in Figure 1 is produced.

[0110] If the next pitching sequence is performed, the pitching information 40 and result 41 are added to the pitching log 25 and saved as pitching log 26. Then, the same process is started again. Note that the order of steps 201-203 and steps 204-206 can be changed as needed.

[0111] In the information processing system 13 according to this embodiment, pitching information 40 regarding the next pitching sequence is predicted based on the feature quantities 39 of the baseball game up to the present. Furthermore, explanatory information 29 indicating the validity of the prediction is generated based on the predicted pitching information 40. This makes it possible to provide high-quality information regarding pitching sequence prediction.

[0112] Technologies that handle "if" scenarios (where A actually occurred, but hypothetically B occurred) are used in various fields. For example, in coupon distribution on e-commerce sites, there are times when it is necessary to estimate whether distributing coupon B to a particular user would have had a positive effect, even though coupon A was actually distributed to that user. Similarly, on video streaming sites, there are times when it is necessary to estimate how the viewing time would have changed if video B had been displayed to a particular user, even though video A was recommended to them. As such, technologies that use data to estimate "if" scenarios have a great deal of demand in the real world.

[0113] Methods for estimation include "causal inference," "counterfactual machine learning," and "reinforcement learning." These differ from methods that use machine learning models to predict general actions (clicks, purchases, etc.). For example, a method is known in which a model that selects the optimal action (type of coupon, type of video, etc.) for each user is directly trained from past log data. Alternatively, a method is known in which a model that simulates the environment (user behavior on an e-commerce site, user behavior on a video streaming site) is trained from log data, and then a model that selects the optimal action for each user within that simulation is trained.

[0114] To provide technology capable of handling such "what-if" scenarios as entertainment for sports fans (especially professional baseball fans), a suitable system is necessary. However, conventional technologies have various limitations and have not adequately met user needs.

[0115] In this technology, in addition to predicting pitching information 40, explanatory information 29 is generated to demonstrate the validity of the prediction. This improves the persuasiveness of the pitching information 40, allowing users to understand the pitching information 40 with a sense of conviction. As a result, high-quality information is provided to users, improving the quality of the viewing experience.

[0116] Furthermore, this technology predicts pitching information 40 regarding the next pitching sequence based on the feature quantities 39 obtained so far. This makes it possible to predict the pitching information 40 with high accuracy.

[0117] Furthermore, this technology predicts the pitch sequence as pitch sequence information 40. For core fans who think about the pitch sequence while watching a game, there may be a need to compare their own thoughts with the AI's, or to have the AI ​​provide additional information when considering the pitch sequence, which can be enjoyable. Predicting the pitch sequence satisfies these needs.

[0118] Furthermore, this technology predicts whether a pitch sequence is advantageous or disadvantageous to the team making the pitch selection, using pitch selection information 40. This makes it possible to entertain users even if they are casual fans who don't think about pitch selection, as long as the AI's predictions are correct regarding whether the batter is kept at bay or hit. Even if the AI's predictions are wrong, it can still provide a different kind of excitement, showing that data isn't everything.

[0119] Furthermore, in this technology, the pitching information 40 includes at least one of the pitching course, pitch type, ball speed, spin rate, or spin axis. This makes it possible to provide the user with even more detailed information.

[0120] Furthermore, in this technology, the feature vector 39 includes at least one of the following: the score of the team making the pitch, the score of the opposing team, the pitcher or their earned run average, the batter or their batting average, the date, the stadium, the strike count, the ball count, the out count, the wind direction, or the defensive shift. This makes it possible to make predictions even more accurate.

[0121] Furthermore, in this technology, explanatory information 29 is generated by using historical pitch sequences similar to the predicted pitch sequence information 40 as explanatory information 29. This further improves the validity of the prediction and makes it possible to give users a sense of satisfaction.

[0122] Furthermore, this technology controls the display of pitching information 40 and explanatory information 29 to the user. This allows the user to intuitively grasp the content of the information.

[0123] Furthermore, this technology uses a machine learning model to predict the pitching information 40. This makes it possible to make predictions with even greater accuracy.

[0124] <Second Embodiment> A more detailed embodiment of the information processing system 13 according to this technology will be described as a second embodiment. In the following description, parts that are the same as the configuration and operation of the information processing system 13 described in the above embodiment may be omitted or simplified.

[0125] [Prediction of Tactics] Figure 11 is a schematic diagram showing an example of UI 47 related to pitching changes. Figure 12 is a schematic diagram showing an example of UI 48 related to pinch hitting. In this embodiment, it is predicted whether or not a predetermined tactic should be executed at the timing of the next pitching sequence. In this example, pitching changes and pinch hitting correspond to "predetermined tactics," and it is predicted whether or not a pitching change should be made and whether or not a pinch hitter should be sent in, respectively. In addition, it may be predicted whether or not any other tactics such as bunting, pinch running, and stealing bases should be executed.

[0126] In Figure 1, the "AI Pitching Details" checkbox in the explanatory information display unit 6 was checked, but in the example in Figure 11, the "Reserve Pitcher" checkbox is checked. Consequently, the area that was the explanatory information display unit 6 has become the reserve pitcher display unit 49.

[0127] The reserve pitcher display unit 49 shows a list of the names, wins, losses, and earned run averages of the reserve pitchers for both teams. In addition, for Team B, the defensive team, the recommendation level (if-relief) for each reserve pitcher's relief pitching is displayed. This value indicates how advantageous it would be for Team B to use that reserve pitcher in relief, and is a value predicted by the information processing system 13.

[0128] In the example shown in Figure 12, the "Reserve Batter" option is checked. Consequently, the area that was previously the explanatory information display unit 6 has become the reserve batter display unit 50.

[0129] The reserve batter display unit 50 shows a list of the names, batting averages, home run totals, and RBI totals of the reserve batters for both teams. In addition, for the offensive team A, the recommendation level (if pinch hitter) for each reserve batter is displayed. This value indicates how advantageous it would be for team A to use that reserve batter as a pinch hitter, and is a value predicted by the information processing system 13.

[0130] Figure 13 is a schematic diagram showing an example of the configuration of a functional block according to this embodiment. In addition to the functional block shown in Figure 2 (not shown), the controller 21 of the information processing device 15 implements the if-succession learning unit 53a, the if-substitute learning unit 53b, the if-succession prediction unit 54a, and the if-substitute prediction unit 54b.

[0131] Figure 14 is a schematic diagram showing an example of information stored in the memory unit 20. In addition to the information shown in Figure 3 (not shown), the memory unit 20 also stores the if-replacement base model 55a, the if-pinch-hitter base model 55b, the if-replacement detail model 56a, and the if-pinch-hitter detail model 56b.

[0132] Figure 15 is a flowchart showing an example of the learning process. The process in this example involves pre-training the if-replacement base model 55a, the if-pinch-hitting base model 55b, the if-replacement detail model 56a, and the if-pinch-hitting detail model 56b.

[0133] The pitching change log is saved (step 301). The pitching change log is acquired from the baseball field 23 via the communication unit 19 by a pitching change log acquisition unit (not shown), etc. Specifically, the pitching change log includes a pitching change base log 57a and a pitching change detail log 62a.

[0134] Figure 16 is a schematic diagram showing the contents of the pitching change base log 57a and the pitching change base quick report log 58a. The pitching change base log 57a includes feature quantities 39, pitching change actions 60, and results 41. In this example, feature quantities 39 include the date, batter, stadium, out count, ball count, strike count, wind direction, scores for both teams, pitcher, relief pitcher, and the result of the pitching sequence. The pitching change action 60 is information indicating whether or not a pitching change occurred. The result 41 is the game result and includes information such as the win / loss record and scores for both teams.

[0135] Figure 17 is a schematic diagram showing the contents of the pitching details log 62a and the pitching details quick report log 63a. The pitching details log 62a includes feature quantities 39, pitcher actions 61, and results 41. In this example, feature quantities 39 include the date, batter, stadium, out count, ball count, strike count, wind direction, and scores for both teams. Pitcher actions 61 is information showing the statistics of the pitcher who actually made the pitches, and includes the pitcher's name, average fastball velocity, and throwing arm. Results 41 are the results of the pitching sequence.

[0136] The if-reliever base model 55a is trained (step 302). For example, the if-reliever learning unit 53a performs offline reinforcement learning of the if-reliever base model 55a. Specifically, the feature quantities 39 from Figure 16 are input to the if-reliever base model 55a, and a pitching change action 60 is output. The output pitching change action 60 is compared with "a pitching change action 60 that actually resulted in a favorable game outcome in the situation of the feature quantities 39 input to the if-reliever base model 55a" using a loss function, and the error is calculated. Then, the if-reliever base model 55a is retrained based on the error. As a result, an if-reliever base model 55a is generated that outputs the probability that it is better to make a pitching change using the feature quantities 39 as input.

[0137] The if-referencing detail model 56a is trained (step 303). For example, the if-referencing learning unit 53a performs offline reinforcement learning of the if-referencing detail model 56a. Specifically, the feature quantities 39 and pitcher actions 61 from Figure 17 are input to the if-referencing detail model 56a, and the appropriateness of the pitcher performing the pitcher change related to the pitcher action 61 is output. The outputted appropriateness is compared with the "results 41 of the feature quantities 39 and pitcher actions 61 input to the if-referencing detail model 56a" using a loss function, and the error is calculated. Then, the if-referencing detail model 56a is retrained based on the error. As a result, the if-referencing detail model 56a is generated that takes the feature quantity 39 as input and outputs the appropriateness of the pitcher change for each pitcher.

[0138] The pinch hitter log is saved (step 304). Specifically, the pinch hitter base log 57b and the pinch hitter detail log 62b are obtained as the pinch hitter log.

[0139] Figure 18 is a schematic diagram showing the contents of the pinch-hitting base log 57b and the pinch-hitting base quick report log 58b. The pitching change base log 57a includes feature quantities 39, pinch-hitting actions 65, and results 41. In this example, feature quantities 39 include the date, batter, reserve batter, and the result of the pitch sequence. The pinch-hitting action 65 is information indicating whether or not a pinch hitter was used. The result 41 is the game result.

[0140] Figure 19 is a schematic diagram showing the contents of the pinch hitter details log 62b and the pinch hitter details quick report log 63b. The pinch hitter details log 62b includes feature quantities 39, batter actions 66, and results 41. The batter actions 66 are information showing the batter's stats, including the batter's name, batting average, and batting dominant arm.

[0141] The if-substitute base model 55b is trained (step 305). For example, the if-substitute learning unit 53b performs offline reinforcement learning of the if-substitute base model 55b. Specifically, the feature quantities 39 from Figure 18 are input to the if-substitute base model 55b, and a substitute action 65 is output. The output substitute action 65 is compared with "a substitute action 65 that actually resulted in a favorable game outcome in the situation of the feature quantities 39 input to the if-substitute base model 55b" using a loss function, and the error is calculated. Then, the if-substitute base model 55b is retrained based on the error. As a result, an if-substitute base model 55b is generated that outputs the probability that it is better to select a substitute using the feature quantities 39 as input.

[0142] The if-substitute-hitting detail model 56b is trained (step 306). For example, the if-substitute-hitting learning unit 53b performs offline reinforcement learning of the if-substitute-hitting detail model 56b. Specifically, the feature quantities 39 and batter actions 66 from Figure 19 are input to the if-substitute-hitting detail model 56b, and the appropriateness of the batter related to the batter action 66 to perform as a substitute hitter is output. The outputted appropriateness and the "results 41 of the feature quantities 39 and batter actions 66 input to the if-substitute-hitting detail model 56b" are compared using a loss function to calculate the error. Then, the if-substitute-hitting detail model 56b is retrained based on the error. As a result, the if-substitute-hitting detail model 56b is generated that takes the feature quantity 39 as input and outputs the appropriateness of each batter to perform as a substitute hitter.

[0143] Figure 20 is a flowchart showing an example of the prediction process related to pitcher changes if. A pitcher change log is saved (step 401). Specifically, the pitcher change log includes the pitcher change base log 58a shown in Figure 16 and the pitcher change detailed log 63a shown in Figure 17.

[0144] The probability that a pitching change is advisable is predicted (step 402). The if-pitching change prediction unit 54a inputs the features contained in the pitching change base quick report log 58a into the if-pitching change base model 55a, and outputs the probability that a pitching change is advisable.

[0145] The suitability of each pitcher for pitching changes is predicted (step 403). The if-pitching-change prediction unit 54a inputs the feature quantities contained in the pitching-change detailed report log 63a into the if-pitching-change detailed model 56a, and the suitability of each pitcher is output.

[0146] The suitability of each pitcher for pitching changes is displayed (step 404). Under the control of the display control unit 36, the suitability is displayed in the if pitching change column in Figure 11.

[0147] Step 405 determines whether the probability of making a pitching change is 0.5 or greater. If the probability is 0.5 or greater (Yes in Step 405), highlighting of UI 47 is performed (Step 406). In the example in Figure 11, the "Reserve Pitcher" checkbox and the if pitching change column are highlighted, but the content of the highlighting can be arbitrary. If the probability is less than 0.5 (No in Step 405), no highlighting is performed and the process ends. Note that the probability threshold is not limited to 0.5 and can be changed as appropriate.

[0148] Figure 21 is a flowchart showing an example of the prediction process related to pinch hitting. A pinch hitter quick report log is saved (step 501). Specifically, the pinch hitter quick report log includes the pinch hitter base quick report log 58b shown in Figure 18 and the pinch hitter detailed quick report log 63b shown in Figure 19.

[0149] The probability that it would be better to use a pinch hitter is predicted (step 502). The if-pinch hitter prediction unit 54b inputs the features contained in the pinch hitter base quick report log 58b into the if-pinch hitter base model 55b, and outputs the probability that it would be better to use a pinch hitter.

[0150] The suitability of each batter as a pinch hitter is predicted (step 503). The if-pinch hitter prediction unit 54b inputs the feature quantities contained in the pinch hitter detailed report log 63b into the if-pinch hitter detailed model 56b, and the suitability of each batter is output.

[0151] The suitability level for pinch-hitting for each batter is displayed (step 504). Under the control of the display control unit 36, the suitability level is displayed in the if pinch-hitting column in Figure 12.

[0152] It is determined whether the probability of using a pinch hitter is 0.5 or greater (step 505). If the probability is 0.5 or greater (Yes in step 505), highlighting of UI48 is performed (step 506). If the probability is less than 0.5 (No in step 505), the process ends without highlighting.

[0153] In this embodiment, when a user is watching from the perspective of the defensive team, they can understand when it would be better to make a pitching change rather than considering the current pitcher's pitching strategy, and which player would be recommended as a replacement. Similarly, when a user is watching from the perspective of the offensive team, they can understand when it would be better to send in a pinch hitter, and which player would be recommended as a pinch hitter. This makes it possible to provide users with an even higher quality experience.

[0154] Furthermore, in this embodiment, since a machine learning model is used for prediction, it is possible to resolve the cold start problem, which is inability to handle newly recruited players for whom there is no data, or combinations of pitchers and batters for whom there is no head-to-head record. In addition, it is possible to make predictions that reflect the detailed game situation.

[0155] In this embodiment, explanatory information indicating the validity of predictions related to pitching changes and pinch hitters may be generated.

[0156] <Third Embodiment> [Simulation] Figure 22 is a schematic diagram showing an example of UI 70 related to simulation. In this embodiment, at the timing after the end of the game, the user's specification of pitch distribution is accepted, and the result is predicted if the specified pitch distribution had been carried out during the game. The timing after the end of the game corresponds to one embodiment of the third timing related to this technology.

[0157] After the game ends, the user can specify the scenario and pitching sequence they want to simulate by operating the UI 70 shown in Figure 22. Specifically, the scenario is specified by operating the checkboxes displayed in the scenario specification unit 71. The pitching sequence (course and pitch type) is also specified on the heatmap 8 in the pitching sequence specification unit 72.

[0158] When a pitch sequence is specified, the simulation results are displayed on the simulation result display unit 73. That is, the results of a hypothetical pitch sequence performed in a scenario specified by the user are simulated and displayed. The simulation results include results such as swings and misses and outs, and their probabilities. The simulation results may also include called strikes and called balls.

[0159] Furthermore, the pitch selection unit 72 displays, for reference, the actual pitch selections, recommended pitch selections, and ineffective pitch selections used in the situation specified by the user. The simulation result display unit 73 displays explanatory information 29 indicating the validity of the recommended and ineffective pitch selections. In addition, explanatory information 29 indicating the validity of the simulation results may also be displayed. Moreover, it may be possible to perform simulations by specifying the situation up to the present, such as between innings, rather than after the end of the game.

[0160] Furthermore, in this embodiment, turning points during the game are extracted by the system and presented to the user. A turning point is a moment in the game when the flow of the game changes significantly. For example, a turning point is extracted when the team making the pitches gives up a run and the recommended pitches and the actual pitches are dissimilar. In this example, the scene selection unit 71 presents the bottom of the 3rd inning as a turning point. The user can also select a turning point from the presented options instead of specifying a scene themselves.

[0161] Figure 23 is a schematic diagram showing an example of the configuration of a functional block according to this embodiment. The controller 21 of the information processing device 15 realizes the simulation learning unit 75, the simulation prediction unit 76, and the turning point extraction unit 77.

[0162] Figure 24 is a schematic diagram showing an example of information stored in the memory unit 20. The memory unit 20 stores a simulation model 79.

[0163] Figure 25 is a flowchart showing an example of the learning process. The process in this example is to pre-train the simulation model 79. First, the pitching log 26 shown in Figure 4 is acquired and stored in the storage unit 20 (step 601).

[0164] The simulation learning unit 75 performs training on the simulation model 79 (step 602). Specifically, supervised learning is performed using the feature quantities 39 and pitching information 40 contained in the pitching log 26 as input data and the result 41 as ground truth data. This generates a simulation model 79 that predicts the result 41 based on the feature quantities 39 and pitching information 40.

[0165] Figure 26 is a flowchart showing an example of the simulation process. At-bats in which the defensive team gave up runs are extracted (step 701). The turning point extraction unit 77 acquires the pitching log 26, and from the pitching log 26, the at-bats in which the defensive team gave up runs are extracted.

[0166] The distances of the actual and recommended pitch sequences are calculated (step 702). Turning points are extracted (step 703). Specifically, the turning point extraction unit 77 extracts as turning points at at-bats in which runs were scored that the type of pitch in the actual and recommended pitch sequences differed, or where the distance to the pitch sequence was large. Such turning points can also be described as situations in which the recommended pitch sequence and the actual pitch sequence were dissimilar.

[0167] A turning point is displayed (step 704). The display control unit 36 ​​controls the display of the turning point to the scene selection unit 71. The scene and ball sequence entered by the user are accepted (step 705). The user can specify the scene themselves, or select a displayed turning point and specify the ball sequence. The scene and ball sequence are accepted when the user presses the simulation button or otherwise.

[0168] The simulation is performed (step 706). The simulation prediction unit 76 inputs the scene and ball distribution to the simulation model 79, and outputs the result 41. The display control unit 36 ​​displays the result 41 on the simulation result display unit 73 (step 707).

[0169] When watching a game, users sometimes think, "What if I had thrown the ball the way I wanted in that situation?" if their favorite team gets hit hard. In this embodiment, it is possible to simulate the result, so users can try out their preferred pitching strategy and gain a sense of satisfaction, thinking, "I should have thrown it the way I wanted." Alternatively, even if they think, "I got hit hard, but what if I had thrown the ball the way the AI ​​recommended?", they can simulate the recommended pitching strategy after the game and gain a sense of satisfaction, thinking, "I guess I could have held them off if I had thrown the ball the way the AI ​​recommended." This makes it possible to provide users with a new kind of enjoyment, even in a losing game.

[0170] Furthermore, in this embodiment, turning points are presented to the user, making it easier for the user to perform simulations and improving convenience.

[0171] <Other Embodiments> This technology is not limited to the embodiments described above, and various other embodiments can be realized.

[0172] In the example shown in Figure 1, pitching information 40 regarding the pitching sequence at the second timing, which is after the first timing, is predicted based on the feature quantities 39 of the game up to the first timing. The first timing is the present, and the second timing is the timing of the next pitching sequence. However, the specific timings of the first and second timings are not limited, as long as they are within the scope of what this technology can implement. For example, the first timing could be a time slightly before the present, and pitching information 40 regarding the next pitching sequence could be predicted based on the feature quantities 39 up to that time slightly before the present.

[0173] In each prediction process of this technology, predictions may be made without using machine learning models. In other words, the specific prediction method is not limited.

[0174] While models such as the predictive pitching model 27a were trained based on the pitching log 26, training may also be performed on the simulation model 79. In other words, model-based reinforcement learning may be performed.

[0175] The explanatory information 29 may be calculated using a data-driven method. For example, the importance of the pitching logs 26 may be calculated using a method such as an influence function, and the pitching logs 26 with high importance may be used to generate the explanatory information 29.

[0176] In calculating the feature quantities 39 and pitch selection values ​​in Figure 10A, weights may be assigned to the content included in the feature quantities 39 and pitch selections. For example, relatively important pitcher names and pitch types may be given greater weight, while less important elements such as dates may be given less weight. This generates useful explanatory information 29 for users who feel more satisfied when the pitcher name and pitch type match. Also, in generating explanatory information 29 for bad pitch selections, pitch selections that result in extra-base hits may be given greater weight. Furthermore, the method of calculating specific values ​​is not limited.

[0177] A "satisfied" button may be displayed on the UI 70 related to the simulation, allowing the user to provide feedback indicating their satisfaction. Based on this feedback, the weighting used in calculating the feature vectors 39 and the pitch distribution values ​​may be optimized.

[0178] In the simulation, the trends of the scenarios specified by the user are analyzed, and the criteria for extracting turning points may be changed according to the analysis results. This allows for personalized extraction criteria, making it possible to present turning points that are more likely to interest the user.

[0179] The term "baseball" in this technology may include softball. Furthermore, this technology may be applied to other sports besides baseball and softball. For example, it may predict the placement of passes in soccer (trajectory, spin, etc.). It may also predict whether or not to implement tactics such as player substitutions.

[0180] Some or all of the functions of the information processing device 15 shown in Figure 2 may be incorporated into the smartphone 14. In other words, the information processing system 13 may be implemented by multiple computers or by a single computer.

[0181] Figure 27 is a block diagram showing an example of the hardware configuration of a computer 500 capable of realizing the information processing device 15. The computer 500 includes a CPU 501, ROM 502, RAM 503, an input / output interface 505, and a bus 504 connecting these to each other. A display unit 506, an input unit 507, a storage unit 508, a communication unit 509, and a drive unit 510 are connected to the input / output interface 505.

[0182] The display unit 506 is a display device using, for example, liquid crystal, EL, etc. The input unit 507 is, for example, a keyboard, pointing device, touch panel, or other operating device. If the input unit 507 includes a touch panel, the touch panel may be integrated with the display unit 506. The storage unit 508 is a non-volatile storage device, for example, an HDD, flash memory, or other solid memory. The drive unit 510 is a device capable of driving the removable recording medium 511, for example, an optical recording medium or magnetic recording tape. The communication unit 509 is a modem, router, or other communication device for communicating with other devices, which can be connected to a LAN, WAN, etc. The communication unit 509 may communicate using either wired or wireless methods. The communication unit 509 is often used separately from the computer 500.

[0183] Information processing by the computer 500 having the hardware configuration described above is realized through the cooperation of software stored in the memory unit 508 or ROM 502, etc., and the hardware resources of the computer 500. Specifically, the information processing method related to this technology is realized by loading the programs that constitute the software, stored in the ROM 502, etc., into the RAM 503 and executing them.

[0184] The program is installed on the computer 500, for example, via a removable recording medium 511. Alternatively, the program may be installed on the computer 500 via a global network or the like. In addition, any non-transient storage medium that the computer 500 can read may be used.

[0185] In this disclosure, "system" means a collection of multiple components (devices, modules (parts), etc.), regardless of whether all components are located in the same enclosure. Therefore, multiple devices housed in separate enclosures and connected via a network, and a single device containing multiple modules in one enclosure, are both considered systems.

[0186] The execution of the information processing method related to this technology by a computer system includes both cases where the following are performed by a single computer, and cases where each process is performed by different computers: for example, model learning, prediction of pitch selection, generation of explanatory information, prediction of whether or not to make a pitching change or pinch hitter, prediction of pitching change suitability and pinch hitter suitability, and simulation. Furthermore, the execution of each process by a predetermined computer includes having another computer execute part or all of the process and obtaining the results. In other words, the information processing method related to this technology can also be applied to cloud computing configurations in which a single function is shared and processed jointly by multiple devices via a network.

[0187] The UI, information processing device, smartphone, various log configurations, and processing flows described with reference to each drawing are merely embodiments and can be modified as needed without departing from the spirit of this technology. In other words, other arbitrary configurations and algorithms may be adopted to implement this technology.

[0188] Where the word "abbreviated" is used in this disclosure, it is used solely to facilitate understanding of the explanation, and there is no special meaning in the use or non-use of the word "abbreviated." In other words, in this disclosure, concepts that define shape, size, positional relationships, states, etc., such as "center," "central," "uniform," "equal," "same," "orthogonal," "parallel," "symmetric," "extending," "axial," "cylindrical," "cylindrical shape," "ring shape," and "annular shape," include concepts such as "substantially centered," "substantially central," "substantially uniform," "substantially equal," "substantially the same," "substantially orthogonal," "substantially parallel," "substantially symmetric," "substantially extending," "substantially axial," "substantially cylindrical," "substantially cylindrical shape," "substantially ring shape," and "substantially annular shape." For example, states that fall within a predetermined range (e.g., a range of ±10%) based on "perfectly centered," "perfectly central," "perfectly uniform," "perfectly equal," "perfectly the same," "perfectly orthogonal," "perfectly parallel," "perfectly symmetric," "perfectly extending," "perfectly axial," "perfectly cylindrical," "perfectly cylindrical shape," "perfectly ring shape," and "perfectly annular shape" are also included. Therefore, even if the word "abbreviated" is not added, the concept may still be included in what is typically expressed with "abbreviated" added. Conversely, the state expressed with "abbreviated" does not mean that the complete state is excluded.

[0189] In this disclosure, expressions using "greater than A" such as "greater than A" and "less than A" are expressions that comprehensively include both concepts that include cases where something is equivalent to A and concepts that do not include cases where something is equivalent to A. For example, "greater than A" is not limited to cases where something is not equivalent to A, but also includes "greater than or equal to A". Similarly, "less than A" is not limited to "less than A", but also includes "less than or equal to A". When implementing this technology, you may appropriately adopt specific settings from the concepts included in "greater than A" and "less than A" so that the effects described above are achieved.

[0190] It is also possible to combine at least two of the feature features of the present technology described above. In other words, the various feature features described in each embodiment may be combined arbitrarily, regardless of the specific embodiment. Furthermore, the various effects described above are merely examples and not limiting, and other effects may also be exhibited.

[0191] Furthermore, this technology can also be configured as follows: (1) An information processing method that predicts pitching information relating to pitching at a second timing after the first timing based on the characteristics of the game up to a first timing in baseball, and generates explanatory information indicating the validity of the prediction of the pitching information based on the predicted pitching information. (2) An information processing method according to (1), wherein the first timing is the present, and the second timing is the timing at which the next pitching will be performed, based on the present. (3) An information processing method according to (1) or (2), wherein the pitching information predicts the pitching that will actually be performed. (4) An information processing method according to any one of (1) to (3), wherein the pitching information predicts whether the pitching is advantageous or disadvantageous to the first team performing the pitching. (5) An information processing method according to any one of (1) to (4), wherein the pitching information includes at least one of the course, type of pitch, speed, rotation speed, or axis of rotation of the pitching. (6) An information processing method according to any one of (1) to (5), wherein the feature quantity includes at least one of the following: the score of the first team that performs the pitching sequence, the score of the second team that is the opponent of the first team, the pitcher of the first team or his earned run average, the batter of the second team or his batting average, the date, the stadium, the strike count, the ball count, the out count, the wind direction, or the defensive shift. (7) An information processing method according to any one of (1) to (6), wherein, when the pitching sequence actually performed up to the first timing is designated as the historical pitching sequence, the information processing method generates the explanatory information by designating the historical pitching sequence that is similar to the predicted pitching sequence information as the explanatory information. (8) An information processing method according to any one of (1) to (7), wherein the information processing method predicts whether or not a predetermined tactic should be executed at the second timing. (9) An information processing method according to (8), wherein the predetermined tactic is a pitching change or a pinch hitter. (10) An information processing method described in (9), wherein the information processing method predicts the degree of suitability for each player to execute the aforementioned pitching change or pinch-hitting.(11) An information processing method according to any one of (1) to (10), wherein at a third timing after the second timing, the user specifies the pitching sequence, and the method predicts the result if the pitching sequence specified at the second timing had been performed. (12) An information processing method according to (11), wherein the pitching sequence information predicts a pitching sequence that is advantageous to the first team performing the pitching sequence, and extracts a timing from the second timing in which the first team concedes a run and the predicted advantageous pitching sequence is dissimilar to the actual pitching sequence. (13) An information processing method according to any one of (1) to (12), wherein the method controls the display of the predicted pitching sequence information and the generated explanatory information to the user. (14) An information processing method according to any one of (1) to (13), wherein the pitching sequence information is predicted using a machine learning model. (15) An information processing system comprising: a prediction unit that predicts pitching information relating to pitching at a second timing after the first timing based on the characteristics of the game up to a first timing in baseball; and a generation unit that generates explanatory information indicating the validity of the prediction of the pitching information based on the pitching information predicted by the prediction unit. (16) An information processing system according to (15), wherein the prediction unit predicts the pitching that will actually be performed as the pitching information. (17) An information processing system according to (15) or (16), wherein the prediction unit predicts the pitching that will be advantageous or disadvantageous to the first team performing the pitching as the pitching information. (18) An information processing system according to any one of (15) to (17), wherein the prediction unit predicts whether or not a predetermined tactic should be executed at the second timing. An information processing system according to any one of (19)(15) to (18), wherein the prediction unit receives a specification of the pitch distribution by the user at a third timing after the second timing, and predicts the result if the pitch distribution specified at the second timing had been carried out.An information processing system according to any one of (20)(15) to (19), further comprising a display control unit that controls the display to the user of the ball distribution information predicted by the prediction unit and the explanatory information generated by the generation unit.

[0192] 1…Display device 5…Pitching information display unit 6…Explanation information display unit 9…Pitching icon 9a…Past pitching icon 9b…Predicted pitching icon 9c…Recommended pitching icon 9d…Bad pitching icon 11a…Predicted pitching explanation 11b…Recommended pitching explanation 11c…Bad pitching explanation 13…Information processing system 14…Smartphone 15…Information processing device 21…Controller 25…Pitching live report log 26, 26a-26d…Pitching log 27a…Predicted pitching model 27b…Recommended pitching model 27c…Bad pitching model 28…Model output 29…Explanation information 32…Pitching log acquisition unit 33a…Predicted pitching learning unit 33b…Recommended pitching learning unit 33c…Bad pitching learning unit 34a…Predicted pitching prediction unit 34b…Recommended pitching prediction unit 34c…Bad pitching prediction unit 35...Explanation Information Generation Unit 36...Display Control Unit 39...Feature Quantity 40...Pitching Information 41...Result 44...Error Calculation Unit 49...Reserve Pitcher Display Unit 50...Reserve Batter Display Unit 53a...Pitching Relay Learning Unit 53b...Pinch Hitter Learning Unit 54a...Pitching Relay Prediction Unit 54b...Pinch Hitter Prediction Unit 55a...Pitching Relay Base Model 55b...Pinch Hitter Base Model 56a...Pitching Relay Details Model 56b...Pinch Hitter Details Model 57a...Pitching Relay Base Log 57b...Pinch Hitter Base Log 58a...Pitching Relay Base Quick Report Log 58b...Pinch Hitter Base Quick Report Log 60...Pitching Relay Action 61...Pitcher Action 62a...Pitching Relay Details Log 62b...Pinch Hitter Details Log 63a...Pitching Relay Details Quick Report Log 63b...Pinch Hitter Details Quick Report Log 65...Pinch Hitter Action 66...Batter Action 71...Scene Specification Unit 72...Pitching selection unit 73...Simulation result display unit 75...Simulation learning unit 76...Simulation prediction unit 77...Turning point extraction unit 79...Simulation model

Claims

1. An information processing method that predicts pitching information regarding pitching at a second timing after the first timing, based on the characteristics of the game up to the first timing in baseball, and generates explanatory information indicating the validity of the prediction of the pitching information based on the predicted pitching information.

2. An information processing method according to claim 1, wherein the first timing is the present, and the second timing is the timing at which the next ball distribution occurs, with reference to the present.

3. An information processing method according to claim 1, wherein the pitching information predicts the actual pitching sequence to be performed.

4. An information processing method according to claim 1, wherein the pitching information predicts whether the pitching sequence is advantageous or disadvantageous to the first team that performs the pitching sequence.

5. An information processing method according to claim 1, wherein the pitching information includes at least one of the course, type of pitch, speed, rotation speed, or axis of rotation of the pitching.

6. An information processing method according to claim 1, wherein the feature quantity includes at least one of the following: the score of the first team that performs the pitching, the score of the second team that is the opponent of the first team, the pitcher of the first team or his earned run average, the batter of the second team or his batting average, the date, the stadium, the strike count, the ball count, the out count, the wind direction, or the defensive shift.

7. An information processing method according to claim 1, wherein, when the pitches actually performed up to the first timing are designated as historical pitches, the information processing method generates the explanatory information by designating historical pitches similar to the predicted pitch information as explanatory information.

8. An information processing method according to claim 1, the method for predicting whether or not a predetermined tactic should be executed at the second timing.

9. The information processing method according to claim 8, wherein the predetermined tactic is a pitching change or a pinch hitter.

10. An information processing method according to claim 9, the method for predicting the degree of suitability for each player to execute the pitching change or pinch-hitting.

11. An information processing method according to claim 1, wherein, at a third timing after the second timing, the method receives a specification of the ball distribution by a user and predicts the result if the ball distribution specified at the second timing had been carried out.

12. An information processing method according to claim 11, wherein the pitching information includes predicting a pitching sequence that is advantageous to the first team performing the pitching sequence, and extracting from the second timings a timing in which the first team concedes a run and the predicted advantageous pitching sequence is dissimilar to the actual pitching sequence.

13. An information processing method according to claim 1, comprising controlling the display of the predicted ball distribution information and the generated explanatory information to the user.

14. An information processing method according to claim 1, wherein the prediction of the pitch distribution information is performed using a machine learning model.

15. An information processing system comprising: a prediction unit that predicts pitching information regarding pitching at a second timing after the first timing, based on the characteristics of the game up to the first timing in baseball; and a generation unit that generates explanatory information indicating the validity of the prediction of pitching information based on the pitching information predicted by the prediction unit.

16. An information processing system according to claim 15, wherein the prediction unit predicts the actual pitching sequence as the pitching sequence information.

17. An information processing system according to claim 15, wherein the prediction unit predicts, as pitching information, a pitching sequence that is advantageous or disadvantageous to the first team that performs the pitching sequence.

18. An information processing system according to claim 15, wherein the prediction unit predicts whether or not a predetermined tactic should be executed at the second timing.

19. An information processing system according to claim 15, wherein the prediction unit receives a specification of the pitch distribution by a user at a third timing after the second timing, and predicts the result if the pitch distribution specified at the second timing had been carried out.

20. An information processing system according to claim 15, further comprising a display control unit that controls the display to the user of the ball distribution information predicted by the prediction unit and the explanatory information generated by the generation unit.

Images