Prediction device, prediction program, and prediction method

Abstract

To improve the accuracy of predictions for publicly managed sports betting. [Solution] The prediction device 10 calculates a first prediction accuracy, which indicates the accuracy of predictions made using prediction models 2-1, 2-2, ... for public sports competitions held in the past at the first stadium 1a. Based on the first prediction accuracy, the prediction device 10 determines a first application model 3a from prediction models 2-1, 2-2, ... to be used for predicting public sports competitions to be held at the first stadium 1a. The prediction device 10 calculates a second prediction accuracy, which indicates the accuracy of predictions made using prediction models 2-1, 2-2, ... for public sports competitions held in the past at the second stadium 1b. Based on the second prediction accuracy, the prediction device 10 determines a second application model 3b from prediction models 2-1, 2-2, ... to be used for predicting public sports competitions to be held at the second stadium 1b.

Description

Technical Field

[0001] The present invention relates to a prediction device, a prediction program, and a prediction method.

Background Art

[0002] In public competitions such as horse racing, motorcycle racing, boat racing, and auto racing, services using AI (Artificial Intelligence) may be provided. For example, the results of public competitions may be predicted by AI and provided to users.

[0003] As a technology related to public competitions, for example, a prediction system has been proposed that enables prediction not only of behavior but also of the influence on the game by comparing the actions of competitors such as players and starting horses before and during the game with the actions of the same competitors in the past. Also, for example, a prediction device has been proposed that can predict the actual race progress and notify users in an easy-to-understand manner.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0005] When predicting the results of public competitions using AI, the past performance of the players participating in the race is used. Here, when there are few races in the learning data that are the same as the race to be predicted and the venue where the race is held, or when the prediction model does not reflect the characteristics of the stadium, accurate prediction may not be possible.

[0006] In one aspect, the present case aims to improve the prediction accuracy of public competitions.

Means for Solving the Problems

[0007] One proposal provides a prediction device having a processing unit. The processing unit calculates a first prediction accuracy, which indicates the accuracy of predicting public sports events held in the past at the first stadium using multiple prediction models. Based on the first prediction accuracy, it determines a first application model from multiple prediction models to be used for predicting public sports events to be held at the first stadium. It then calculates a second prediction accuracy, which indicates the accuracy of predicting public sports events held in the second stadium using multiple prediction models. Based on the second prediction accuracy, it determines a second application model from multiple prediction models to be used for predicting public sports events to be held at the second stadium. [Effects of the Invention]

[0008] According to one embodiment, the prediction accuracy of public gambling can be improved. [Brief explanation of the drawing]

[0009] [Figure 1] This figure shows an example of a prediction device according to the first embodiment. [Figure 2] This figure shows an example of an information processing system according to the second embodiment. [Figure 3] This figure shows an example of a hardware configuration for a prediction server. [Figure 4] This block shows an example of the functionality of a prediction server. [Figure 5] This figure shows an example of accuracy information. [Figure 6] This figure shows an example of prediction processing. [Figure 7] This figure shows an example of how prediction results are displayed at an off-track betting facility. [Figure 8] This flowchart shows an example of the procedure for predicting competition results. [Modes for carrying out the invention]

[0010] The following description of this embodiment will be made with reference to the drawings. Note that each embodiment can be implemented by combining multiple embodiments within a reasonable scope. [First Embodiment] First, the first embodiment will be described.

[0011] Figure 1 shows an example of a prediction device according to the first embodiment. The first embodiment determines a prediction model to be used for predicting public sports events for each sports venue. The prediction device 10 is a computer that determines a prediction model to be used for predicting public sports events. The prediction device 10 is, for example, a server computer that provides users with AI-based prediction results for public sports events. The prediction device 10 has a processing unit 11. The processing unit 11 controls the prediction device 10 and can execute the required processing. The processing unit 11 is, for example, a processor or arithmetic circuit in the prediction device 10. The processing unit 11 determines a prediction model to be used for predicting public sports events for each sports venue from prediction models 2-1, 2-2, ...

[0012] Predictive models 2-1, 2-2, ... are predictive models that make predictions regarding public sports betting. For example, when information about a public sports race is input to predictive models 2-1, 2-2, ..., it outputs a betting pattern for that race. Race information includes, for example, information about the athletes participating in the race and information about the stadium where the race is held. The betting pattern is, for example, a combination of the type of bet and numbers that identify the athletes (horse number, car number, etc.). For example, predictive models 2-1, 2-2, ... each use different training data for learning.

[0013] The processing unit 11 calculates a first prediction accuracy, which indicates the accuracy of predictions made using prediction models 2-1, 2-2, ... for public sports events held in the past at the first stadium 1a. The first stadium 1a is a stadium where public sports events are held. The first stadium 1a is, for example, a horse racing track, a bicycle racing track, a boat racing track, etc. First, the processing unit 11 obtains information and race results (past data) of public sports events held in the past at the first stadium 1a. Based on the obtained information on public sports events, the processing unit 11 performs predictions using prediction models 2-1, 2-2, ... Then, the processing unit 11 calculates a first prediction accuracy based on the prediction results and race results. Here, the processing unit 11 calculates a prediction accuracy based on the betting amount and payout amount based on the betting patterns based on the predictions made using prediction models 2-1, 2-2, ...

[0014] For example, in calculating the first prediction accuracy for prediction model 2-1, the processing unit 11 uses prediction model 2-1 to output betting patterns for each race based on information from past public sports betting races. Based on past race results, the processing unit 11 identifies the amount of winnings that can be received if bets are placed using the outputted betting patterns. The processing unit 11 calculates the first prediction accuracy as the recovery rate, which represents the ratio of the identified winnings to the total bet amount when betting using the outputted betting patterns.

[0015] The processing unit 11 determines the first applicable model 3a to be used for predicting public sports events held at the first stadium 1a, based on the first prediction accuracy, from among prediction models 2-1, 2-2, .... For example, the processing unit 11 determines the prediction model with the highest first prediction accuracy among prediction models 2-1, 2-2, ... as the first applicable model 3a.

[0016] Further, the processing unit 11 calculates a second prediction accuracy indicating the accuracy of predicting public competitions that have been held in the past in the second stadium 1b using the prediction models 2-1, 2-2, ··· respectively. The second stadium 1b is a stadium where the same public competitions as those in the first stadium 1a are held. First, the processing unit 11 acquires information on the races and race results (past data) of public competitions that have been held in the past in the second stadium 1b. The processing unit 11 executes predictions using the prediction models 2- respectively based on the acquired information on the races of public competitions. Then, the processing unit 11 calculates the second prediction accuracy based on the prediction results and the race results. Here, the processing unit 11 calculates the second prediction accuracy for each of the prediction models 2-1, 2-2, ··· in the same manner as the first prediction accuracy.

[0017] Based on the second prediction accuracy, the processing unit 11 determines a second application model 3b to be used for predicting public competitions to be held in the second stadium 1b from the prediction models 2-1, 2-2, ···. For example, the processing unit 11 determines the prediction model with the highest second prediction accuracy among the prediction models 2-1, 2-2, ··· as the second application model 3b.

[0018] According to the first embodiment, the processing unit 11 of the prediction device 10 calculates a first prediction accuracy indicating the accuracy of predicting public competitions that have been held in the past in the first stadium 1a using the prediction models 2-1, 2-2, ··· respectively. Based on the first prediction accuracy, the processing unit 11 determines a first application model 3a to be used for predicting public competitions to be held in the first stadium 1a from the prediction models 2-1, 2-2, ···. The processing unit 11 calculates a second prediction accuracy indicating the accuracy of predicting public competitions that have been held in the past in the second stadium 1b using the prediction models 2-1, 2-2, ··· respectively. Based on the second prediction accuracy, the processing unit 11 determines a second application model 3b to be used for predicting public competitions to be held in the second stadium 1b from the prediction models 2-1, 2-2, ···.

[0019] In this way, the prediction device 10 can set a prediction model for each stadium. This allows the prediction device 10 to predict public sports events while reflecting the characteristics of the stadium. Therefore, the prediction device 10 can improve the prediction accuracy of public sports events.

[0020] Furthermore, the processing unit 11 calculates the prediction accuracy based on the voting amount and payout amount for voting patterns based on predictions using prediction models 2-1, 2-2, .... This allows the prediction device 10 to calculate the prediction accuracy according to the degree of payout obtained according to the predictions of each prediction model.

[0021] Furthermore, if the first prediction accuracy for the first application model 3a is below a threshold, the processing unit 11 may update the first application model 3a based on the first prediction accuracy for each of the prediction models 2-1, 2-2, ... This allows the prediction device 10 to reduce unnecessary accuracy calculation processing and lessen the processing load.

[0022] [Second Embodiment] Next, a second embodiment will be described. In the second embodiment, when predicting the results of public sports such as horse racing, bicycle racing, boat racing, and auto racing, a prediction model is selected for each sports venue.

[0023] Figure 2 shows an example of an information processing system according to the second embodiment. The information processing system according to the second embodiment includes a velodrome terminal 21, an off-track betting terminal 24, and a prediction server 100. The velodrome terminal 21, the off-track betting terminal 24, and the prediction server 100 are connected to a network 30. The network 30 is a wide-area network such as the Internet.

[0024] The prediction server 100 is a server computer for a company that operates a service providing AI-based prediction results for keirin races held at velodromes 20a, 20b, 20c, ... The prediction server 100 uses an applicable model for each velodrome to perform predictions for each race held at each velodrome. The prediction server 100 transmits the prediction results to each velodrome. For example, the prediction server 100 transmits the prediction results for each race held at velodrome 20a to the velodrome terminal 21. The prediction server 100 also transmits the prediction results for each race held at velodromes 20a, 20b, 20c, ... to the off-track betting terminal 24. The prediction server 100 calculates the hit rate and return rate for each applicable model and updates the applicable model with the lowest hit rate and return rate.

[0025] Velodrome terminal 21 is a computer that displays prediction results from prediction server 100 at velodrome 20a. Velodrome terminal 21 is connected to monitor 22. Monitor 22 is, for example, a large LCD television installed in the betting area of ​​velodrome 20a. Velodrome terminal 21 obtains prediction results for each race held at velodrome 20a from prediction server 100. Velodrome terminal 21 displays the obtained prediction results on monitor 22. Velodrome terminal 21 may also display the prediction results on a location other than monitor 22. For example, velodrome terminal 21 may display the prediction results on a monitor installed in the seating area where users of velodrome 20a watch the races. Velodromes 20b, 20c, ... also have velodrome terminals and monitors similar to velodrome terminal 21 and monitor 22.

[0026] The off-track betting terminal 24 is a computer that displays the prediction results from the prediction server 100 at the off-track betting facility 23. The off-track betting facility 23 is a place where betting tickets can be purchased outside of the velodrome. At the off-track betting facility 23, it is possible to place bets on keirin races held at velodromes 20a, 20b, 20c, etc. The off-track betting terminal 24 is connected to a monitor 25. The monitor 25 is, for example, a large LCD television. The off-track betting terminal 24 obtains the prediction results for each race held at velodromes 20a, 20b, 20c, etc. from the prediction server 100. The off-track betting terminal 24 displays the obtained prediction results on the monitor 25.

[0027] Figure 3 shows an example of the hardware configuration of a prediction server. The prediction server 100 is controlled as a whole by a processor 101. The processor 101 is connected to memory 102 and several peripheral devices via a bus 109. The processor 101 may be a multiprocessor. The processor 101 is, for example, a CPU (Central Processing Unit), an MPU (Micro Processing Unit), or a DSP (Digital Signal Processor). At least some of the functions realized by the processor 101 executing a program may be realized by electronic circuits such as an ASIC (Application Specific Integrated Circuit) or a PLD (Programmable Logic Device).

[0028] Memory 102 is used as the main memory of the prediction server 100. Memory 102 temporarily stores at least a portion of the OS (Operating System) program and application programs to be executed by the processor 101. Memory 102 also stores various data used for processing by the processor 101. As memory 102, a volatile semiconductor memory device such as RAM (Random Access Memory) is used.

[0029] Peripheral devices connected to bus 109 include a storage device 103, a GPU (Graphics Processing Unit) 104, an input interface 105, an optical drive device 106, a device connection interface 107, and a network interface 108.

[0030] The storage device 103 electrically or magnetically writes and reads data from its built-in recording medium. The storage device 103 is used as an auxiliary storage device for a computer. The storage device 103 stores the OS program, application programs, and various data. For example, the storage device 103 can be an HDD (Hard Disk Drive) or an SSD (Solid State Drive).

[0031] A monitor 31 is connected to the GPU 104. The GPU 104 displays images on the screen of the monitor 31 according to instructions from the processor 101. The monitor 31 can be an OLED (Electroluminescence) display device or a liquid crystal display device, among others.

[0032] The input interface 105 is connected to a keyboard 32 and a mouse 33. The input interface 105 transmits signals from the keyboard 32 and mouse 33 to the processor 101. Note that the mouse 33 is just one example of a pointing device; other pointing devices can also be used. Other pointing devices include touch panels, tablets, touchpads, and trackballs.

[0033] The optical drive device 106 reads data recorded on the optical disc 34 using laser light or the like. The optical disc 34 is a portable recording medium on which data is recorded in a way that makes it readable by the reflection of light. Examples of optical discs 34 include DVD (Digital Versatile Disc), DVD-RAM, CD-ROM (Compact Disc Read Only Memory), and CD-R (Recordable) / RW (ReWritable).

[0034] The device connection interface 107 is a communication interface for connecting peripheral devices to the prediction server 100. For example, a memory device 35 and a memory reader / writer 36 can be connected to the device connection interface 107. The memory device 35 is a recording medium equipped with a communication function with the device connection interface 107. The memory reader / writer 36 is a device that writes data to or reads data from the memory card 37. The memory card 37 is a card-type recording medium.

[0035] The network interface 108 is connected to the network 30. The network interface 108 transmits and receives data to and from other computers or communication devices via the network 30.

[0036] The prediction server 100 can realize the processing functions of the second embodiment with the hardware configuration described above. The prediction device 10 shown in the first embodiment can also be realized with the same hardware as the prediction server 100 shown in Figure 3. Furthermore, the velodrome terminal 21 and the off-track betting terminal 24 can also be realized with the same hardware as the prediction server 100. The processor 101 is an example of the processing unit 11 shown in the first embodiment.

[0037] The prediction server 100 implements the processing functions of the second embodiment by executing a program recorded on a computer-readable recording medium, for example. The program describing the processing content to be executed by the prediction server 100 can be recorded on various recording media. For example, the program to be executed by the prediction server 100 can be stored in the storage device 103. The processor 101 loads at least a portion of the program in the storage device 103 into the memory 102 and executes the program. Alternatively, the program to be executed by the prediction server 100 can be recorded on a portable recording medium such as an optical disc 34, a memory device 35, or a memory card 37. The program stored on the portable recording medium becomes executable after being installed in the storage device 103, for example, under control from the processor 101. The processor 101 can also directly read and execute the program from the portable recording medium. Next, the functions of the prediction server 100 will be described in detail.

[0038] Figure 4 is a block diagram illustrating an example of the functions of a prediction server. The prediction server 100 includes a storage unit 110, a prediction unit 120, a communication unit 130, and a model setting unit 140. The storage unit 110 is implemented using the storage area of ​​memory 102 or storage device 103. The prediction unit 120, communication unit 130, and model setting unit 140 are implemented by the processor 101 executing a program stored in memory 102.

[0039] The memory unit 110 stores prediction models 111-1, 111-2, ... and accuracy information 112. The prediction models 111-1, 111-2, ... accept race information as input and output promising betting patterns for the race. The prediction models 111-1, 111-2, ... are generated by learning information and results from past races. For example, the training data used for each prediction model 111-1, 111-2, ... is different.

[0040] For example, prediction models 111-1, 111-2, ... include prediction models trained using information and results of all races held at velodromes 20a, 20b, 20c, ... during a predetermined period as training data. Also, for example, prediction models 111-1, 111-2, ... include prediction models trained using information and results of all races held at a specific velodrome among velodromes 20a, 20b, 20c, ... during a predetermined period as training data. Also, for example, prediction models 111-1, 111-2, ... include prediction models trained using information and results of all races held at a specific velodrome among velodromes 20a, 20b, 20c, ... with specific characteristics (for example, an indoor velodrome) during a predetermined period as training data. Accuracy information 112 is information that records the hit rate and return rate when predictions are made using the applied model for each velodrome. The hit rate is the percentage of times the prediction was correct. The recovery rate is the ratio of the amount of money recovered to the amount of money spent on voting.

[0041] The race information used to train the prediction models 111-1, 111-2, etc. includes information about the participating athletes and information about the velodrome where the race is held. Athlete information includes, for example, their recent performance results.

[0042] Information about a velodrome includes, for example, the bank's inclination angle, bank's circumference, straightaway distance, curve shape and radius, track material, track condition, past average race times, past finishing order distribution, and average racing line in past races. The bank's inclination angle is numerical data indicating the angle of inclination between the inside and outside of the bank. The bank's circumference is numerical data indicating the total length of the bank. The straightaway distance is numerical data indicating the length of the final straight. The curve shape and radius are numerical data indicating the radius of the curve or a numerical representation of its shape.

[0043] Track material is a dummy variable indicating whether the track is made of cement or wood. Track condition is a dummy variable indicating the dryness or wetness of the track on the day of the race. Past average race times are records of past average race times on the track, broken down by vehicle type and weather. Past finishing order distribution is records of past finishing order distribution on the track, broken down by vehicle number and race development such as leading or overtaking. Average racing line in past races is a numerical representation of which line—inside, middle, or outside—was the most advantageous.

[0044] The prediction unit 120 makes predictions for keirin races using prediction models 111-1, 111-2, ... Before the opening of each keirin track 20a, 20b, 20c, ... the prediction unit 120 makes predictions for each race held at each keirin track on that day. For example, the prediction unit 120 inputs information on each race held at each keirin track into the applicable model and outputs the betting patterns for each race. In addition, when the applicable model is updated, the prediction unit 120 makes predictions for past races at the keirin track whose applicable model is being updated using prediction models 111-1, 111-2, ... For example, the prediction unit 120 inputs information on races from a predetermined period in the past at the keirin track whose applicable model is being updated into each of the prediction models 111-1, 111-2, ... and outputs the betting patterns for each race for each prediction model.

[0045] The communication unit 130 transmits the prediction results from the prediction unit 120 to the velodromes 20a, 20b, 20c, ... and the off-track betting facility 23. For example, when velodrome 20a opens for business, the communication unit 130 transmits the betting patterns for each race held at velodrome 20a on that day, output by the prediction unit 120, in response to a request from the velodrome terminal 21. Also, when the off-track betting facility 23 opens for business, the communication unit 130 transmits the betting patterns for each race held at velodrome 20a, 20b, 20c, ..., in response to a request from the off-track betting facility terminal 24, output by the prediction unit 120.

[0046] The model setting unit 140 sets the applicable model. After the closing of business for each velodrome 20a, 20b, 20c, ..., the model setting unit 140 aggregates the return rate and hit rate for each velodrome if bets were placed using the betting patterns output by the prediction unit 120. The model setting unit 140 updates the applicable model for velodromes where the return rate and hit rate values ​​are below the lower limit. The model setting unit 140 calculates the return rate and hit rate for each prediction result of prediction models 111-1, 111-2, ... for past races at the velodrome whose applicable model is being updated. The model setting unit 140 sets the prediction model with the highest return rate and hit rate as the applicable model.

[0047] Note that the lines connecting each element shown in Figure 4 represent only a part of the communication path, and other communication paths besides those shown can also be set. Next, the accuracy information 112 stored in the memory unit 110 will be explained.

[0048] Figure 5 shows an example of accuracy information. Accuracy information 112 is information that records the hit rate and return rate when predictions are made using the applied model for each velodrome. Accuracy information 112 has the following items set for each velodrome: No., Venue, Applied model, Current return rate, Current hit rate, Lower limit of return rate, Lower limit of hit rate, and Display priority. The No. item is set as a number that identifies the corresponding velodrome. The Venue item is set as the name of the corresponding velodrome. The Applied model item is set as the applied model for the corresponding velodrome.

[0049] The "Current Recovery Rate" field displays the current recovery rate of the prediction results for races at the corresponding velodrome using the applied model. The recovery rate is calculated as follows: Recovery Rate [%] = (Total Payout Amount ÷ Total Purchase Amount) × 100. The "Current Hit Rate" field displays the current hit rate of the prediction results for races at the corresponding velodrome using the applied model. The hit rate is calculated as follows: Hit Rate [%] = (Number of Hits ÷ Number of Purchases) × 100. The "Current Recovery Rate" and "Current Hit Rate" fields are reset, for example, for each velodrome race series.

[0050] A minimum return rate is set for the "Lower Return Rate" item. When the value of the "Current Return Rate" item falls below the value of the "Lower Return Rate" item, the applicable model for the corresponding velodrome is updated. A minimum hit rate is set for the "Lower Hit Rate" item. When the value of the "Current Hit Rate" item falls below the value of the "Lower Hit Rate" item, the applicable model for the corresponding velodrome is updated. The values ​​for the "Lower Return Rate" and "Lower Hit Rate" items may be the same for all velodromes, or they may differ from velodrome to velodrome.

[0051] The display priority setting determines the degree to which prediction results for multiple velodromes are displayed when they are shown. For example, the display priority setting will be higher for each track with a higher value in the "Current Return Rate" field, and if the "Current Return Rate" values ​​are the same, the display priority setting will be higher for each track with a higher value in the "Current Hit Rate" field. Alternatively, the display priority setting may also be set so that a higher value in the "Current Hit Rate" field results in a higher rank. Next, we will explain the prediction process using the prediction model.

[0052] Figure 6 shows an example of the prediction process. The prediction unit 120 of the prediction server 100 performs predictions for races held at a certain velodrome (for example, velodrome 20a) using an applicable model for that velodrome (for example, prediction model 111-1). The prediction unit 120 inputs information about the races held at velodrome 20a into the prediction model 111-1. The race information includes information about the athletes participating in the race and information about the velodrome where the race is held. The prediction unit 120 outputs a predetermined number of betting patterns using the prediction model 111-1. The number of points indicates the number of combinations of betting types and rider numbers.

[0053] As an example, the prediction unit 120 outputs betting patterns for two-car exacta and trifecta bets where car number 1 is in first place, car number 4 or 7 is in second place, and car numbers 7, 2 or 4 are in third place (represented as 1-4, 7-7, 2, 4). The prediction unit 120 also outputs betting patterns for two-car exacta and trifecta bets where car number 4 is in first place, car number 7 or 1 is in second place, and car numbers 7, 2 or 1 are in third place (represented as 4-7, 1-7, 2, 1). The communication unit 130 of the prediction server 100 transmits the prediction results to a velodrome terminal (for example, velodrome terminal 21) at the velodrome where the predicted race is being held.

[0054] When the velodrome terminal 21 obtains prediction results from the prediction server 100, it displays screen 41 on the monitor 22. Screen 41 is a screen that shows the obtained prediction results. Screen 41 includes the display of the velodrome name and race number. Screen 41 also includes a display of the betting pattern obtained as a prediction result.

[0055] For example, screen 41 displays two exacta and trifecta bets, represented by 1-4,7-7,2,4 and 4-7,1-7,2,1, as betting patterns. Screen 41 also shows that the two exacta bet represented by 1-4,7-7,2,4 is worth 2 points and the trifecta bet is worth 4 points. Furthermore, screen 41 displays symbols indicating the likelihood of each rider's success. By displaying such screen 41, the velodrome terminal 21 allows users of the velodrome 20a to place bets while viewing AI-generated predictions. Next, the display of prediction results at the off-track betting facility 23 will be explained.

[0056] Figure 7 shows an example of how prediction results are displayed at an off-track betting facility. The communication unit 130 of the prediction server 100 transmits the prediction results for races held at velodromes 20a, 20b, 20c, ... and the display priority for each velodrome to the off-track betting terminal 24. In this case, the communication unit 130 transmits the prediction results for races held at velodromes ranked 1st to 7th in display priority to the off-track betting terminal 24.

[0057] When the off-track betting terminal 24 obtains prediction results from the prediction server 100, it displays the obtained prediction results on the monitor 25. Here, the off-track betting terminal 24 highlights and displays the prediction results of the velodrome with the highest display priority. For example, the off-track betting terminal 24 displays the prediction results in screen arrangement 51.

[0058] Screen layout 51 shows the on-screen arrangement of prediction results for velodromes with higher display priority. Screen layout 51 indicates that the prediction results for the velodrome with the 1st display priority are placed in the upper left, the prediction results for the 2nd display priority are placed in the upper right, the prediction results for the 3rd display priority are placed in the lower left, and the prediction results for the 4th display priority are placed in the lower right. The content displayed as the prediction results for each velodrome is the same as in screen 41.

[0059] The off-track betting terminal 24 displays the prediction results for the velodromes with higher display priority in screen layout 51 for a predetermined time, and then displays the prediction results in screen layout 52. Screen layout 52 is the on-screen arrangement of the prediction results for the velodromes with lower display priority. Screen layout 52 indicates that the prediction results for the velodrome with the 5th display priority are placed in the upper left, the prediction results for the velodrome with the 6th display priority are placed in the upper right, and the prediction results for the velodrome with the 7th display priority are placed in the lower left. After the off-track betting terminal 24 displays the prediction results in screen layout 52 for a predetermined time, it displays the prediction results in screen layout 51.

[0060] In this way, the off-track betting terminal 24 displays the prediction results for races held at each velodrome on the monitor 25 so that the prediction results for velodromes with high display priority are highlighted. Here, the prediction server 100 sets a high display priority for velodromes whose results were predicted using an application model with a high recovery rate. This makes it easier for users of the off-track betting terminal 23 to see prediction results with high prediction accuracy when placing bets while viewing the AI-generated prediction results. Next, the procedure for race prediction processing by the prediction server 100 will be explained.

[0061] Figure 8 is a flowchart showing an example of the procedure for predicting competition results. The process shown in Figure 8 will be explained below according to the step numbers. [Step S11] The prediction unit 120 selects one velodrome from velodrome 20a, 20b, 20c, ... Note that the processes in steps S11 to S13 are executed before the opening of business for velodrome 20a, 20b, 20c, ...

[0062] [Step S12] The prediction unit 120 performs predictions using the applicable model for each race held on the day at the velodrome selected in step S11. For example, the prediction unit 120 refers to the accuracy information 112 and identifies the applicable model for the velodrome selected in step S11 from among prediction models 111-1, 111-2, ... The prediction unit 120 inputs the race information into the applicable model and outputs the betting patterns for each race held on the day at the velodrome selected in step S11.

[0063] [Step S13] The prediction unit 120 determines whether all velodromes have been selected. If the prediction unit 120 determines that all velodromes have been selected, it proceeds to step S14. If the prediction unit 120 determines that there are still velodromes that have not been selected, it proceeds to step S11.

[0064] [Step S14] The communication unit 130 transmits the prediction results to the velodromes 20a, 20b, 20c, ... and the off-track betting facility 23. For example, when the velodromes 20a, 20b, 20c, ... open for business, the communication unit 130 transmits the betting patterns for each race held at the receiving velodromes on that day in response to requests from the velodrome terminals at the velodromes 20a, 20b, 20c, ... Also, when the off-track betting facility 23 opens for business, the communication unit 130 transmits the betting patterns for each race held at the velodromes 20a, 20b, 20c, ... in response to requests from the off-track betting facility terminal 24. The communication unit 130 may also transmit prediction results for the velodromes requested by the velodromes 20a, 20b, 20c, ... and the off-track betting facility 23, respectively.

[0065] [Step S15] The model setting unit 140 selects one velodrome from velodrome 20a, 20b, 20c, ... Note that the processing in steps S15 to S21 is executed after the closing of business for velodrome 20a, 20b, 20c, ...

[0066] [Step S16] The model setting unit 140 calculates the return rate and the hit rate and updates the accuracy information 112. Based on the betting patterns and results output in Step S12 for each race held on the day at the velodrome (selected venue) selected in Step S15, the model setting unit 140 calculates the total purchase amount, total payout amount, number of hits, and number of purchases if bets were made using that betting pattern. The model setting unit 140 adds the calculated values ​​to the total purchase amount, total payout amount, number of hits, and number of purchases aggregated for the selected venue in the current round. The model setting unit 140 then calculates the return rate [%] = (total payout amount ÷ total purchase amount) × 100 and sets the calculated return rate in the accuracy information 112's current return rate item for the selected venue. The model setting unit 140 also calculates the hit rate [%] = (number of hits ÷ number of purchases) × 100 and sets the calculated hit rate in the accuracy information 112's current hit rate item for the selected venue.

[0067] [Step S17] The model setting unit 140 determines whether the current recovery rate and hit rate values ​​are below the lower limit. For example, the model setting unit 140 determines that the current recovery rate and hit rate values ​​are below the lower limit if the current recovery rate item value for the selection field of the accuracy information 112 is below the lower limit of the recovery rate item, or if the current hit rate item value is below the lower limit of the hit rate item. If the model setting unit 140 determines that the current recovery rate and hit rate values ​​are below the lower limit, it proceeds to step S18. If the model setting unit 140 determines that the current recovery rate and hit rate values ​​are not below the lower limit, it proceeds to step S21.

[0068] [Step S18] The prediction unit 120 performs predictions for past races at the selected venue using each of the prediction models 111-1, 111-2, ... For example, the prediction unit 120 inputs information on past races for a predetermined period at the selected venue into each of the prediction models 111-1, 111-2, ... and outputs the betting patterns for each race for each prediction model.

[0069] [Step S19] The model setting unit 140 calculates the return rate and hit rate for each prediction model 111-1, 111-2, ... Based on past race results, the model setting unit 140 aggregates the total purchase amount, total payout amount, number of hits, and number of purchases for each prediction model, assuming bets were made using the betting patterns output in Step S18. For each prediction model, the model setting unit 140 calculates the return rate [%] = (total payout amount ÷ total purchase amount) × 100. The model setting unit 140 also calculates the hit rate [%] = (number of hits ÷ number of purchases) × 100 for each prediction model.

[0070] [Step S20] The model setting unit 140 sets the prediction model with the highest recovery rate and accuracy rate as the applicable model for the selected field. For example, the model setting unit 140 determines that the prediction model with the highest recovery rate calculated in step S19 is the applicable model for the selected field. If there are multiple prediction models with the highest recovery rate, the model setting unit 140 determines that the prediction model with the highest accuracy rate among the multiple prediction models with the highest recovery rates is the applicable model for the selected field. The model setting unit 140 sets the prediction model that has been determined as the applicable model in the item for the applicable model for the selected field in the accuracy information 112.

[0071] [Step S21] The model setting unit 140 determines whether all velodromes have been selected. If the model setting unit 140 determines that all velodromes have been selected, it terminates the process. If the model setting unit 140 determines that there are still velodromes that have not been selected, it proceeds to step S15.

[0072] In this way, the prediction server 100 performs predictions for keirin races using an application model set for each keirin track. Here, the prediction server 100 performs predictions for past races held at the keirin track using prediction models 111-1, 111-2, ..., and sets the prediction model with the highest return rate and accuracy rate as the application model for that keirin track. As a result, the prediction server 100 can set a prediction model that can predict public sports events while reflecting the characteristics of the keirin track as its application model. Therefore, the prediction server 100 can improve the prediction accuracy of public sports events.

[0073] Furthermore, when determining the applicable model, the process of performing predictions for past races for all prediction models 111-1, 111-2, ... would be computationally intensive. Therefore, the prediction server 100 updates the applicable model when the recovery rate and accuracy rate fall below a lower limit. This allows the prediction server 100 to reduce unnecessary processing and alleviate its processing load.

[0074] The prediction server 100 outputs voting patterns based on prediction models 111-1, 111-2, ... The prediction server 100 calculates the return rate and hit rate based on the total purchase amount, total payout amount, number of correct predictions, and number of purchases when voting is made according to the outputted voting patterns, and determines the applicable model based on the return rate and hit rate. This allows the prediction server 100 to set the prediction model that yields the largest payout according to the prediction as the applicable model.

[0075] Although embodiments have been illustrated above, the configurations of each part shown in the embodiments can be replaced with others having similar functions. Furthermore, other arbitrary components or processes may be added. Moreover, any two or more configurations (features) from the embodiments described above may be combined. [Explanation of Symbols]

[0076] 1a 1st stadium 1b 2nd stadium 2-1, 2-2, ... Predictive Model 3a First Application Model 3b Second Application Model 10 Prediction device 11 Processing Section

Claims

1. A processing unit that calculates a first prediction accuracy, which indicates the accuracy of predicting public sports events held in the past at the first stadium using each of several prediction models; determines a first application model from the several prediction models to be used for predicting public sports events to be held at the first stadium based on the first prediction accuracy; calculates a second prediction accuracy, which indicates the accuracy of predicting public sports events held in the past at the second stadium using each of the several prediction models; and determines a second application model from the several prediction models to be used for predicting public sports events to be held at the second stadium based on the second prediction accuracy. A prediction device having the following features.

2. The processing unit updates the first application model based on the first prediction accuracy for each of the plurality of prediction models if the first prediction accuracy for the first application model is less than a threshold. The prediction device according to claim 1.

3. The processing unit calculates the prediction accuracy based on the voting amount and payout amount for voting patterns based on predictions using the plurality of prediction models. The prediction device according to claim 1.

4. On the computer, The first prediction accuracy is calculated, which shows the accuracy of predicting past public sports events held at the first stadium using multiple prediction models. Based on the first prediction accuracy, a first application model to be used for predicting the public sports competition to be held at the first stadium is determined from the plurality of prediction models. A second prediction accuracy is calculated, which indicates the accuracy of predicting the public sports competitions held in the past at the second stadium using each of the multiple prediction models. Based on the second prediction accuracy, a second application model to be used for predicting the public sports competition to be held at the second stadium is determined from the plurality of prediction models. A predictive program that executes a process.

5. Computers The first prediction accuracy is calculated, which shows the accuracy of predicting past public sports events held at the first stadium using multiple prediction models. Based on the first prediction accuracy, a first application model to be used for predicting the public sports competition to be held at the first stadium is determined from the plurality of prediction models. A second prediction accuracy is calculated, which indicates the accuracy of predicting the public sports competitions held in the past at the second stadium using each of the multiple prediction models. Based on the second prediction accuracy, a second application model to be used for predicting the public sports competition to be held at the second stadium is determined from the plurality of prediction models. Prediction method.

Images