Competition dynamic priority grouping and intelligent arrangement system

Abstract

The invention relates to the field of electronic competition and data processing, and discloses a competition dynamic priority grouping and intelligent arrangement system, which comprises an event twinborn body construction module used for aggregating multi-modal data in real time and constructing a competition panoramic digital mirror image; the dynamic narrative position potential field generation module is used for quantifying and modeling multi-dimensional narrative values of all potential games in the future, such as quick cold, perennial enemy and the like, on the basis of event twinborn body data; the intelligent arrangement decision-making module takes the competition state and the narrative position potential field as input, and outputs an optimal arrangement scheme by taking maximization of competition fairness, audience participation degree and narrative value as targets through a reinforcement learning algorithm; and the narrative intervention module is used for identifying narrative opportunity points by analyzing the potential field gradient and generating an executable operation strategy to actively improve the competition attraction. Through the closed-loop design of "intervention-feedback-enhancement", the system can be continuously self-optimized, and the ornamental value and commercial value of the competition are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of e-sports and data processing technology, specifically to a dynamic priority grouping and intelligent scheduling system for competitions. Background Technology

[0002] With the booming development of the global esports industry, the scale, complexity, and commercial value of modern large-scale esports events have reached unprecedented heights. However, the core scheduling and organization methods of current events still largely follow the static model of traditional sports, which contrasts sharply with the highly digital and fast-paced nature of esports, highlighting its shortcomings.

[0003] Existing tournament scheduling technologies often pre-determine most of the schedule, primarily relying on static data such as initial player or team rankings and historical records for grouping and matchups. This approach lacks effective response to dynamic changes during the tournament, failing to capture and utilize real-time information such as fluctuations in player performance, the emergence of new tactics, or changes in audience sentiment. Consequently, it frequently misses potential "golden matches" that could ignite audience enthusiasm and create memorable moments.

[0004] Furthermore, the current extraction of the "storytelling" or "narrative" value of competitions relies heavily on the personal experience and subjective judgment of the competition directors, commentators, and marketing teams. This manually driven model is not only inefficient but also difficult to replicate on a large scale. As a result, many promising narrative threads—such as the "spear versus shield" battle between two players with contrasting styles, a player's "revenge battle" against fate, or the "comeback story" of a rising star—are submerged in the massive amount of competition information and fail to be fully developed and presented. This insufficient extraction of dynamic narrative value directly limits the further enhancement of the competition's content appeal and commercial potential. Summary of the Invention

[0005] Existing event scheduling methods typically employ fixed formats or dynamic matching mechanisms based on a single skill rating. These methods have limitations: 1) rigid schedules, unable to adapt to dynamic changes during the event; 2) singular optimization goals, usually only considering competitive fairness while neglecting the quantification and integration of multi-dimensional factors that significantly impact the overall value of the event, such as its entertainment value and topicality; and 3) a lack of ability to predict future events and proactive guidance mechanisms, hindering the forward-looking planning and cultivation of the event narrative. Therefore, providing an event scheduling system that integrates multi-dimensional dynamic information and collaboratively optimizes and proactively intervenes in the competitive and entertainment value of events is a pressing technical problem that needs to be solved in this field.

[0006] To address the aforementioned technical problems, this invention provides a dynamic priority grouping and intelligent scheduling system for sporting events, comprising:

[0007] The module includes an event twin construction module, a dynamic narrative potential field generation module, an intelligent orchestration decision-making module, and a narrative intervention module.

[0008] The event twin construction module's function is to aggregate multi-source, heterogeneous, multimodal data in real time to construct an event twin synchronized with the actual event progress. This module continuously acquires data through a data interface, and the multimodal data includes, but is not limited to:

[0009] The contestants' real-time physiological indicators and operational data: such as heart rate and stress level collected by sensors, and actions per minute (APM) and critical operation success rate obtained through the competition system interface;

[0010] Social media topic popularity and sentiment scores: such as the amount of discussion about a specific player or potential match on social networks, and the positive or negative sentiment scores of related comments calculated using natural language processing technology; Live streaming platform audience sentiment distribution: such as the real-time statistical distribution of audience sentiment obtained by analyzing live streaming bullet screen data;

[0011] Real-time market odds: Reflecting the market's overall expectations for the match outcome. This module processes and aligns the collected data, constructing and updating a high-dimensional match state vector in real time; this match state vector is a comprehensive digital mapping of the current match ecosystem, constituting the event twin.

[0012] The dynamic narrative potential field generation module, connected to the event twin construction module, calculates all potential future game combinations based on the output match state vector to generate a quantified, multi-dimensional dynamic narrative potential field. This dynamic narrative potential field consists of multiple narrative potential dimensions, specifically including:

[0013] Upset potential (ρ) upset The calculation method is based on the skill gap between the two players and their recent instability, quantifying the potential narrative value of a victory for the weaker player. An example calculation formula is as follows:

[0014] ρ upset (m ij )=f upset (Δr ij ,σ i ,σ j )·wρ upset (m ij )=f upset (Δr ij ,σ i ,σ j )·w story ((underdog i );

[0015] Where, m ij This represents the game between player i and player j, Δr ij It is the difference in the strength ratings of both sides, σ i ,σ j It is the volatility parameter of the recent state of both parties, w story (underdog i () is to assign narrative weight to the weaker party, f upset It is a function used to calculate the final potential value based on the input parameters.

[0016] Nemesis Position (ρ) rivalry The calculation method is based on historical head-to-head records and the popularity of related discussions on social media, quantifying the grudges and antagonism of the game.

[0017] Style clash potential (ρ) clash The calculation method is based on the differences in the players' technical and tactical style vectors, quantifying the entertainment value of the game at the technical and tactical level.

[0018] Narrative arc potential (ρ) storyline The calculation method is based on the matching degree between the player's career trajectory and the preset narrative model, quantifying the player's ability to continue or climax the game's personal storyline.

[0019] In addition, this module also has a dynamic evolution function. When a real match result is generated or new data is received, this module performs a deduction in the event twin, calculates the ripple effect of the event on the entire dynamic narrative potential field, that is, the chain effect of a result on the potential values ​​of all other potential games, thereby realizing the real-time dynamic update of the narrative potential field.

[0020] The intelligent orchestration decision-making module is configured to employ a multi-objective reinforcement learning method. It uses the event state vector output by the event twin construction module and the dynamic narrative potential field output by the dynamic narrative potential field generation module as decision inputs to generate candidate orchestration schemes. To evaluate the quality of the schemes, this module pre-defines a comprehensive reward function. An exemplary function expression is as follows:

[0021] R t =w fair ·F fairness +w eng ·E eng +w npf ·S NPF +w wel ·W wel -C cost ;

[0022] Among them, R t To calculate the overall reward score, Ffairness E scores points for competitive fairness. eng S is the expected audience engagement score. NPF W represents the total narrative position score for all games in the aforementioned tournament format. wel C scores points for player well-being cost For operating costs, w fair w eng w npf w wel These are the weighting coefficients for each item.

[0023] The narrative intervention module analyzes the potential field gradient of the dynamic narrative potential field to identify narrative opportunities with high growth potential. Specifically, this module calculates the gradient by analyzing the sensitivity of the total potential value of the dynamic narrative potential field to external intervention factors, thereby pinpointing the narrative opportunities with the best expected intervention effect. Based on the identified narrative opportunities, the module generates specific narrative guidance strategies. These strategies are a set of executable operational instructions, such as providing event media, marketing, or commentary teams with the narrative themes to be highlighted and related data indexes.

[0024] This invention, through the collaborative work of the aforementioned modules, constitutes a closed-loop feedback and self-reinforcing process. After the narrative guidance strategy generated by the narrative intervention module is executed, the resulting new audience interaction data and media reports are captured by the event twin construction module as new multimodal real-time data and used to update the event state vector. The dynamic narrative potential field generation module then enhances the corresponding potential values ​​within the field, thus completing potential injection. This enhanced dynamic narrative potential field provides the intelligent choreography decision module with higher-value state inputs during reinforcement learning in the next decision cycle, enabling it to generate choreography schemes that maximize the long-term comprehensive value of the event.

[0025] This invention provides a dynamic priority grouping and intelligent scheduling system for competitions. It has the following beneficial effects:

[0026] 1. This invention, through an event twin construction module, aggregates multimodal data in real time, including athlete physiology, audience emotions, and social media buzz. Furthermore, through a dynamic narrative potential field generation module, it transforms previously difficult-to-quantify factors such as the entertainment value and storytelling of a competition into calculable narrative potential. This expands the system's scheduling decisions from the single, static dimension of competitive fairness in existing technologies to a multi-dimensional comprehensive value assessment that reflects the entire competition ecosystem in real time, resulting in a more comprehensive and precise decision-making basis.

[0027] 2. This invention uses a dynamic narrative potential field generation module to extrapolate potential future matches and simulate the ripple effect of different outcomes on the entire potential field, thus enabling the prediction and evaluation of the future direction of the event narrative. Combined with the multi-objective optimization of the intelligent scheduling decision module, the system can plan a scheduling path that maximizes the entertainment value and topicality of subsequent stages in the early stages of the event, thereby transforming event scheduling from a passive response to past results to an active planning and forward-looking layout of future event value.

[0028] 3. By setting up a narrative intervention module, this invention can proactively identify and generate narrative guidance strategies to guide external operations. The effects of the intervention measures will be captured as new data by the event twin construction module and updated in the form of potential injection, serving as the optimization basis for the next round of decisions by the intelligent orchestration decision module. Attached Figure Description

[0029] Figure 1 This is a functional module structure diagram of the event dynamic priority grouping and intelligent scheduling system of the present invention;

[0030] Figure 2 This is a flowchart illustrating the process of the event twin construction module of the present invention.

[0031] Figure 3 This is a schematic diagram of the present invention;

[0032] Figure 4 A flowchart of the dynamic narrative potential field generation module;

[0033] Figure 5 This is a flowchart of the intelligent orchestration decision module of the present invention. Detailed Implementation

[0034] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] See attached document Figure 1 , Figure 1 This is a functional module structure diagram of a dynamic priority grouping and intelligent scheduling system for competitions according to an embodiment of the present invention. The system provided by the present invention can be deployed on a single server, a distributed server cluster, or a cloud computing platform in terms of hardware. At the software level, the system may include: an event twin construction module 10, a dynamic narrative potential field generation module 20, an intelligent scheduling decision module 30, and a narrative intervention module 40.

[0036] The event twin construction module 10 is the system's data input and state representation unit. This module is configured with multiple data interfaces for real-time aggregation of multimodal data from different data sources. After being cleaned, aligned, and normalized by the module's internal data processing pipeline, this data is constructed into a unified, high-dimensional event state vector. This event state vector is a digital mapping of the entire real event ecosystem at time t, forming the event twin. This module outputs the generated event state vector to the dynamic narrative potential field generation module 20 and the intelligent staging decision module 30.

[0037] The dynamic narrative potential field generation module 20 receives the tournament state vector output from the event twin construction module 10. This module's function is to calculate all potential future match combinations based on the real-time data contained in the tournament state vector, generating a quantified dynamic narrative potential field covering the entire future tournament schedule. This dynamic narrative potential field consists of multiple preset narrative potential dimensions and is output in the form of a data structure to the intelligent matchmaking decision module 30 and the narrative intervention module 40.

[0038] The intelligent staging decision module 30 has its data input terminals connected to the event twin construction module 10 and the dynamic narrative potential field generation module 20, respectively. This module uses the received event state vector and dynamic narrative potential field as the basis for decision-making and employs a multi-objective reinforcement learning method to generate candidate staging schemes. Internally, this module is configured with a comprehensive reward function to evaluate the expected total value of each candidate staging scheme. This comprehensive reward function uses the quantification result of the dynamic narrative potential field as a core evaluation indicator. Through optimization calculation using a reinforcement learning algorithm, this module selects the scheme with the highest reward function evaluation value from all candidate staging schemes and outputs it as the final optimal staging scheme.

[0039] The narrative intervention module 40 receives the dynamic narrative potential field output by the dynamic narrative potential field generation module 20. The function of this module is to analyze the internal structure and numerical distribution of the field, specifically analyzing the potential field gradient to identify narrative opportunities with high growth potential. Based on the identified narrative opportunities, this module generates a set of narrative guidance strategies to guide external operational activities of the event.

[0040] The aforementioned modules work together to form the overall data processing and control flow of this invention. The event twin construction module 10 provides the system with basic data and state input; the dynamic narrative potential field generation module 20 quantifies and models the narrative value based on this; the intelligent orchestration decision module 30 uses the outputs of the former two to make optimal decisions; and the narrative intervention module 40 analyzes the potential field and generates external intervention commands. After the narrative guidance strategy output by the narrative intervention module 40 is executed externally, its resulting effects (such as new social media hotspots) will be captured again by the event twin construction module 10 as new multimodal data, thereby forming a closed loop between the data flow and control flow of the entire system, and continuously performing iterative calculations in subsequent operating cycles.

[0041] See attached document Figure 2 , Figure 2 This is a flowchart of the event twin construction module according to an embodiment of the present invention. The specific execution steps of this module 10 may include:

[0042] S21, through the data interface configured in the module, multi-source heterogeneous multimodal data is aggregated in real time. This module obtains raw data streams from multiple independent external data sources. In one embodiment, the multimodal data includes:

[0043] Real-time physiological indicators and operational data of the athletes: By connecting to the sensor devices worn by the athletes that conform to the communication protocol, real-time physiological indicators such as heart rate and skin conductance are obtained; at the same time, by making API calls to the competition system's server, operational data such as effective actions per minute (APM) and hit rate of specific skills during the competition are obtained.

[0044] Social media topic popularity and sentiment score: By calling the public APIs of mainstream social media platforms, data such as the number of discussions and reposts of keywords related to the event, players or specific matches are obtained to calculate the topic popularity; and a natural language processing model is applied to the collected text comment data to calculate the sentiment score.

[0045] Viewer Emotion Distribution on Live Streaming Platforms: By collecting and analyzing real-time bullet screen data from live streaming platforms that are officially partnered with the event, and using preset keyword and emoji classification rules, the quantity and proportion of various emotions (such as anticipation, surprise, dissatisfaction, etc.) expressed by viewers within a specific time window are statistically analyzed to form the viewer emotion distribution.

[0046] Real-time market odds: Obtain real-time odds data for future match results by connecting to one or more legitimate, publicly available betting data providers.

[0047] S22 processes the aggregated raw data. The module's internal data processing pipeline performs a series of standardized operations on the various heterogeneous data acquired in step S21, specifically including:

[0048] Data cleaning: Removing invalid values, missing values, or obviously abnormal data points from a data stream.

[0049] Data normalization: Scaling data of different types and dimensions (e.g., heart rate values, social media discussion counts, odds values) to a uniform numerical range, such as [0, 1], through mathematical transformations (e.g., min-max normalization or Z-score normalization) to eliminate dimensional differences.

[0050] Timestamp alignment: Attach precise timestamps to data from all sources and align data streams with different frequencies and delays according to a unified time base to ensure data synchronization at any given time.

[0051] S23, Constructing a competition state vector based on the processed data. The standardized multidimensional data output from step S22 is organized at each time point t to construct a unified, high-dimensional competition state vector, S. t This vector is a comprehensive digital snapshot of the entire event ecosystem at time t. In one embodiment, the structure of this vector can be defined as:

[0052] S t =[V P (t),V A (t),V M (t)]S t =[V P (t),V A (t),V M (t)];

[0053] in:

[0054] V P (t) is the set of state sub-vectors of all contestants at time t, where each contestant's vector contains their static strength score, dynamic physiological indicators, and operational data.

[0055] V A (t) is the audience interaction state subvector at time t, which includes social media topic popularity, sentiment score, and audience emotion distribution on the live streaming platform.

[0056] V M (t) is the media and environment state subvector at time t, which includes market odds, the stage of the event, and other relevant business or environmental parameters.

[0057] S24, output and continuously update the event state vector. The event twin construction module 10 outputs the constructed event state vector to the dynamic narrative potential field generation module 20 and the intelligent scheduling decision module 30. This module repeats steps S21 to S23 at a preset frequency (e.g., per second or per minute) to realize the event state vector S t Continuous and real-time updates ensure that it remains synchronized with the progress of the actual event.

[0058] See attached document Figure 3 , Figure 3 This is a flowchart of the dynamic narrative potential field generation module according to an embodiment of the present invention. The specific execution steps of this module 20 may include:

[0059] S31, Receive the real-time event state vector S output by the event twin construction module 10. t The event state vector provides the data foundation for all subsequent calculations in this module.

[0060] S32, based on the received event state vector S t For all potential future match combinations that have not yet occurred during the tournament schedule. ij (Here, i and j represent the indices of different players), calculate their basic narrative potential vector N(m) one by one. ij This vector consists of multiple independent narrative potential dimensions, and in one embodiment, its calculation includes:

[0061] Upset potential (ρ) upset This value quantifies the potential story value of a player with a lower skill rating defeating a player with a higher skill rating. Its calculation formula can be defined as:

[0062] ρ upset (m ij )=f upset (Δr ij ,σ i ,σ j )·w story (underdog i );

[0063] Where, Δr i j = |r i -r j | represents the absolute value of the difference in skill ratings between players i and j; σ i and σ j These parameters represent the volatility of the two athletes' recent performance. These parameters can be calculated based on the variance of recent competition results or the fluctuation range of physiological data; w story (underdog i(underdog) is the weaker side in a match. i A pre-defined narrative weight is assigned to it; f upset It is a pre-defined nonlinear function whose function value increases with the increase of and the increase of the weak square value.

[0064] Nemesis Position (ρ) rivalry ): Used to quantify the historical rivalries and antagonism in matches. Its calculation formula can be expressed as:

[0065] v i s j ocial ));

[0066] Among them, H ij It is a historical confrontation index calculated based on a historical confrontation database; From the event state vector S t Extracted from social media discussions that simultaneously mention players i and j; w h and w s These are the weighting coefficients corresponding to historical confrontation and social popularity, respectively; tanh is the hyperbolic tangent function used to normalize the results.

[0067] Style clash potential (ρ) clash This value quantifies the degree of difference in the technical and tactical styles of the two players in a match. Its calculation formula can be defined as:

[0068]

[0069] Among them, T i and T j This is a feature vector extracted from the event twins, representing the tactical styles of player i and player j. This formula calculates the cosine distance between the two style vectors, with a value range of [0, 2]. The larger the value, the more significant the style difference.

[0070] Narrative arc potential (ρ) storyline The calculation of ). This value is used to quantify the ability of the match to match a preset narrative model. Its calculation formula can be defined as:

[0071]

[0072] Among them, L i (t) and L j (t) represents the career trajectory sequences of players i and j up to the current time t, respectively; k is an index in the preset narrative model library, each model containing two character templates A. k and B k; g(·) is a matching degree calculation function used to calculate the degree of fit between a player's career trajectory and a specific role template. The formula aims to find the value that gives the highest matching degree to a certain model in the game and narrative model library formed by the two players.

[0073] S33, Construct a dynamic narrative potential field. This involves calculating all future potential games m as shown in step S32. ij Narrative potential vector N(m) ij These data are integrated to form a unified data structure stored in matrix or tensor form. This data structure is the dynamic narrative potential field Φ at time t. t .

[0074] S34 performs dynamic evolution calculations of the dynamic narrative potential field. When the system receives a new real match result O... t When dealing with updated multimodal data, this module performs inference within the event twin environment, recalculating the impact of the event on the entire dynamic narrative potential field Φ. t The ripple effect is the chain reaction generated by the narrative potential values ​​of other related potential matches. For example, an upset result will directly change the skill rating r of the relevant players, and thus change the upset potential ρ of all their future potential matches. upset Meanwhile, the shift in social media discussion sparked by this result will also alter the power dynamics of potential rivalries in related matchups. rivalry This evolutionary process can be formally represented as:

[0075] Φ t+1 =F evolve (Φ t O t );

[0076] Where, Φ t+1 It is the updated dynamic narrative potential field, F evolve It is an evolution function, specifically implemented as an evolution function for Φ. t All results of the match O t The affected items are recalculated according to the formula in step S32.

[0077] S35 will update the dynamic narrative potential field Φ t+1 The data is output to the intelligent orchestration decision module 30 and the narrative intervention module 40 as the basis for their subsequent operations.

[0078] See attached document Figure 4 , Figure 4 This is a flowchart of the intelligent orchestration decision module according to an embodiment of the present invention. The specific execution steps of this module 30 may include:

[0079] S41, the event scheduling problem is constructed as a Markov Decision Process (MDP). This module receives the event state vector S from the event twin construction module 10. t and the dynamic narrative potential field Φ from the dynamic narrative potential field generation module 20 t As input, the tuple (S, A, P, R, γ) of this Markov decision process is defined as follows:

[0080] State space (S): states s t ∈S is the event state vector S at time t. t ;

[0081] Action space (A): Action a t ∈A is a candidate game arrangement that can be executed at time t. Each candidate game arrangement is a specific set of game arrangements that conforms to the official rules of the tournament;

[0082] State transition function (P): State transition P(s) t+1 |s t ,a t The probability of ) is not calculated by this module, but by the execution of orchestration scheme a. t Subsequently, the actual physical progression of the event and the results of the new data collection by the event twin construction module 10 are jointly determined;

[0083] Reward function (R): When in state s t Choose action a t Then, the system calculates the instant reward value based on a preset comprehensive reward function, the specific design of which is detailed in S42;

[0084] Discount factor (γ): A preset constant between 0 and 1 used to adjust the weight of future rewards in the current decision.

[0085] S42, for each candidate orchestration scheme a t A value assessment is performed. This assessment is conducted using a pre-defined, weighted reward function that calculates a comprehensive reward based on multiple optimization objectives. In one embodiment, the expression for this function is:

[0086] R t =w fair ·F fairness +w eng ·E eng +w npf ·SR t =w fair ·F fairness +w eng ·E eng +

[0087] wnpf ;

[0088] The definitions of each item are as follows:

[0089] R t The overall reward score is the module's evaluation of candidate orchestration scheme a. t The quantitative assessment results of its immediate value.

[0090] F fairness : Competitive fairness score, the value of which is based on scheme a t The difference distribution of the strength ratings r of both sides in all matches was calculated, and the smaller the strength gap between the two sides, the higher the fairness score of the match combination.

[0091] E eng The expected audience engagement score is calculated using a pre-trained prediction model that predicts a based on historical data. t The plan brought about changes in metrics such as viewership and increased social media discussion.

[0092] S NPF : The total score of the narrative potential field, whose value is scheme a t All games included m i The narrative potential vector N(m) of j i The norm sum of j), i.e. This value is directly derived from the dynamic narrative potential field Φ. t The data.

[0093] W wel Athlete well-being score: This score is used to assess the rationality of the competition schedule. It is calculated by factors such as the density of consecutive competitions for athletes and the travel burden of cross-regional travel. The lower the burden, the higher the score.

[0094] C cost Operating costs, used to quantify the implementation of plan a t Estimated costs of required resources such as venue and manpower.

[0095] w fair w eng w npf w wel : These are the weighting coefficients for each of the above scoring items. These coefficients are hyperparameters that can be set and adjusted by external users (such as the event organizers) according to the strategic goals of the event (e.g., focusing on fairness in the early stages of the event and on entertainment value in the final stage).

[0096] S43, Policy Optimization via Reinforcement Learning Algorithm. This module employs a reinforcement learning algorithm (e.g., Q-learning based on value functions or policy gradient based on policy functions) to learn within the MDP environment defined in S41. The goal of the algorithm is to learn an optimal policy π. * Such that from any state s t Initially, the long-term cumulative discount rewards that can be obtained The goal is to maximize the expected value. The learning process involves iteratively updating the strategy through simulation or interaction with the real environment until convergence.

[0097] S44 outputs the optimal tournament arrangement. After policy learning is complete, for the current input tournament state vector s t The module applies the learned optimal policy π * (s t That is, choosing the action value function Q that yields the optimal action value. * (s t a) Maximize the action a t argmax a∈A Q * (s t ,a). This action a t The corresponding candidate orchestration scheme is the optimal orchestration scheme output by this module.

[0098] See attached document Figure 5 , Figure 5 This is a flowchart of a narrative intervention module according to an embodiment of the present invention. The specific execution steps of module 40 may include:

[0099] S51, Receive the real-time dynamic narrative potential field Φ output from the dynamic narrative potential field generation module 20. t This dynamic narrative potential field forms the data foundation for the analysis in this module.

[0100] S52, Analyzing the dynamic narrative potential field Φ t The potential field gradient is used to identify narrative opportunities with high growth potential. The potential field gradient is defined here as the sensitivity of the total potential value of the dynamic narrative potential field to external intervention factors. In a specific embodiment, this gradient is calculated by performing a small perturbation simulation in an event twin environment. The specific process is as follows:

[0101] From the dynamic narrative potential field Φ t Choose one potential future match m ij .

[0102] Select an external intervention factor e that is relevant to this game. ij For example, media exposure or social media discussion surrounding the match.

[0103] In the event twin environment, regarding the intervention factor e ij Apply a known small increment Δe ij This forms a simulated, perturbed state vector of the competition.

[0104] Based on the disturbed event state vector, the dynamic narrative potential field generation module 20 is invoked to recalculate a simulated dynamic narrative potential field Φ. ′ t And obtain its total potential value S. ′ NPF

[0105] The formula for calculating the estimated potential field gradient corresponding to this disturbance is as follows:

[0106]

[0107] in, For the game m ij The potential field gradient, S N PF is the original total potential value.

[0108] Repeat steps 1 to 5 above to iterate through all key potential matches and identify the potential match with the highest calculated gradient value as the current narrative opportunity point.

[0109] S53, based on the narrative opportunities identified in step S52, generates a structured and specific narrative guidance strategy. This strategy is a set of operational instructions that can be directly delivered to the event operations team (including media, marketing, or commentary teams) for execution. In a specific embodiment,

[0110] The narrative guidance strategy is a structured dataset containing the following fields:

[0111] Target game identifier: Clearly indicates the potential future game that this strategy is targeting.

[0112] Core narrative theme: Define the core story points that need to be highlighted and communicated, such as "rivalry showdown," "style restraint," or "revenge battle."

[0113] Specific operational instructions: Provide clear and actionable operational suggestions, such as "create a 1-minute warm-up video for the target match" or "start a discussion on social media about 'whether player XX can continue their winning streak with a specific tactic'."

[0114] Data Index: Provides an index or identifier that points to relevant data in the event twin building block 10, which the operations team can consult to support their content creation, such as analysis reports pointing to historical head-to-head data or quantitative data on players' technical and tactical styles.

[0115] S54 outputs the generated narrative guidance strategy to an external interface for the event operations team to receive and execute. The output of this module directly serves the event's operational activities, aiming to enhance specific potential values ​​within the dynamic narrative potential field through external intervention.

[0116] See attached document Figure 1 , Figure 1 This is a system overall workflow diagram according to an embodiment of the present invention. A complete end-to-end workflow of the system may include the following steps:

[0117] S11, the system starts and continuously collects data. The event twin building module 10 continuously aggregates multimodal real-time data from external data sources through its configured multiple data interfaces. This data includes athletes' physiological and operational data, audience interaction data from social media and live streaming platforms, and event environment data such as market odds.

[0118] S12 involves constructing the event twin and generating the dynamic narrative potential field. The event twin construction module 10 preprocesses the raw data aggregated in step S11 and constructs the latest event state vector S at each decision time point t. t Subsequently, the state vector S of the event... t It is transmitted to the dynamic narrative potential field generation module 20. The dynamic narrative potential field generation module 20 is based on S t The data is used to calculate all potential future game combinations, generating a quantitative, multi-dimensional dynamic narrative potential field Φ. t .

[0119] S113, execute intelligent staging decision. The intelligent staging decision module 30 simultaneously receives the event state vector S from the event twin construction module 10. t and the dynamic narrative potential field Φ from the dynamic narrative potential field generation module 20 t This module, in its internal Markov decision process, will use S t and Φ t As input to the current state, multiple candidate tournament arrangements that conform to the tournament rules are generated. Then, the module utilizes a pre-defined comprehensive reward function R. t Each candidate scheme is evaluated, and the optimal scheme is selected and output through reinforcement learning algorithm to maximize the long-term cumulative reward.

[0120] S114, Execute the generation of the narrative intervention strategy. The narrative intervention module 40 receives the dynamic narrative potential field Φ generated by the dynamic narrative potential field generation module 20. t This module analyzes Φ tThe system identifies narrative opportunities with high growth potential by analyzing the potential field gradient and generates a structured narrative guidance strategy containing specific operational instructions and data indexes based on these opportunities. This strategy is then output to the external event operations team. Step S315 completes the feedback data feedback. After the external event operations team executes the corresponding operational activities based on the narrative guidance strategy output in step S114, the real-world effects of these activities, such as changes in audience discussion and increased media coverage, are captured again as new multimodal data by the event twin construction module 10 at the next time point. This feedback data is then integrated into the processes of steps S111 and S112 to update the event state vector and dynamic narrative potential field, thereby providing updated and higher-value state inputs for the system's next round of scheduling decisions and intervention strategy generation. These steps are repeated cyclically, constituting the complete workflow of this system.

[0121] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A dynamic priority grouping and intelligent scheduling system for competitions, characterized in that, include: The event twin construction module is used to aggregate multimodal data in real time, construct and synchronously reflect the event state vector of the current event ecosystem, and form the event twin of the event; The dynamic narrative potential field generation module is used to calculate potential game combinations based on the event twin through the game state vector, quantify and generate a dynamic narrative potential field that covers future potential games and consists of multiple narrative potential dimensions. The intelligent choreography decision module is configured to use a multi-objective reinforcement learning method to generate candidate choreography schemes based on the event state vector and the dynamic narrative potential field, evaluate the candidate choreography schemes by a preset comprehensive reward function, and optimize them by reinforcement learning algorithm to select the optimal choreography scheme. The narrative intervention module is used to analyze the potential field gradient of the dynamic narrative potential field to identify narrative opportunity points with high growth potential, and generate narrative guidance strategies based on the narrative opportunity points to guide the operation of the event, thereby improving the potential value in the dynamic narrative potential field through external intervention.

2. The event dynamic priority grouping and intelligent scheduling system according to claim 1, characterized in that, The multimodal data includes: The athlete's real-time physiological indicators and operational data; Social media topic popularity and sentiment score; The distribution of audience sentiment on live streaming platforms; Real-time market odds.

3. The event dynamic priority grouping and intelligent scheduling system according to claim 1, characterized in that, The dynamic narrative potential field includes: Upset Potential: Based on the skill rating gap between the two sides and their recent instability, quantify the potential tournament value of the weaker side winning; Rivalry Position: Based on historical head-to-head records and related social media discussion, quantify the grudges and rivalries between matches; Style clash and positioning: Based on the differences in the technical and tactical style vectors of the players, quantify the entertainment value of the game at the technical and tactical level; Narrative Arc Potential: Based on the match between a player's career trajectory and a pre-defined narrative model, quantify the ability of a match to continue or escalate a player's personal storyline.

4. The event dynamic priority grouping and intelligent scheduling system according to claim 1, characterized in that, The dynamic evolution of the dynamic narrative potential field includes: Based on the actual match results and newly incoming multimodal data, the outcome of a potential game is deduced in the event twin, and the ripple effect of the potential game outcome on the dynamic narrative potential field is calculated to realize the dynamic evolution of the narrative potential field.

5. The event dynamic priority grouping and intelligent scheduling system according to claim 1, characterized in that, The evaluation metrics for the comprehensive reward function include: A score for fairness in competition based on player skill ratings; Expected audience engagement score based on model prediction; The total score of the narrative potential field based on the dynamic narrative potential field; A player welfare score is used to assess the rationality of the competition schedule.

6. The event dynamic priority grouping and intelligent scheduling system according to claim 5, characterized in that, The comprehensive reward function is defined as a weighted sum of multiple objectives, and its expression is: R t =w fair ·F fairness +w eng ·E eng +w npf ·S NPF +w wel ·W wel -C cost Among them, R t To calculate the overall reward score, F f Airness is a score for fair play in competition, E eng S is the expected audience engagement score. N PF represents the total narrative position score for all games in the aforementioned tournament format, W wel C scores points for player well-being cost For operating costs, w fair ,w eng ,w npf ,w wel These are the weighting coefficients for each item.

7. The event dynamic priority grouping and intelligent scheduling system according to claim 1, characterized in that, The narrative intervention module identifies the narrative opportunity point with the highest potential for growth by analyzing the sensitivity of the total potential value of the dynamic narrative potential field to external intervention factors.

8. The event dynamic priority grouping and intelligent scheduling system according to claim 1, characterized in that, The dynamic narrative potential field is a narrative guidance strategy that identifies narrative opportunity points and generates a set of specific, executable operational instructions for event media, marketing, or commentary teams based on these opportunities.

9. The event dynamic priority grouping and intelligent scheduling system according to claim 1, characterized in that, The audience interaction data and new information from media reports triggered by the narrative guidance strategy are captured by the event twin construction module as new multimodal real-time data; thereby completing the potential injection into the dynamic narrative potential field and providing higher-value state input for the intelligent orchestration decision module in the next decision cycle.

Images