Intelligent transaction decision-making method and system based on hierarchical multi-round adversarial debate

Abstract

The application discloses an intelligent transaction decision-making method and system based on hierarchical multi-round confrontation debate, which is suitable for stock and other financial asset transaction scenarios. The method realizes efficient, low-delay and interpretable intelligent transaction decision-making by constructing a hierarchical multi-round confrontation debate mechanism, solving the limitations of existing technologies in deep reasoning, delay control and decision transparency. The core of the method is to build a complete process of multi-agent confrontation debate, dynamic reputation evaluation and transparent decision-making, divide the multi-role agent into the argument generation group and the refutation group, and execute R rounds of deep debate. In the debate process, the agent reputation is updated by the Bayesian reputation update, and the transaction signal is generated within milliseconds by combining the Soft-Borda dynamic voting mechanism. At the same time, through the transparent decision link of the multi-round debate log and the explanation vector, the requirements of the model interpretability and real-time audit of the financial supervision are met.

Description

Technical Field

[0001] This invention relates to the field of interdisciplinary technology of artificial intelligence and financial technology, specifically to an intelligent trading decision-making method and system based on hierarchical multi-round adversarial debate. Background Technology

[0002] In the field of financial trading, with the increasing complexity of markets and the growing demand for faster trading, intelligent trading decision-making technology has made significant progress in recent years, especially in algorithmic trading and quantitative analysis, leveraging artificial intelligence and machine learning algorithms. However, current technologies still fall short in addressing certain key challenges. In high-frequency trading scenarios, there are stringent millisecond-level latency requirements for trading decisions. However, models that achieve deep inference, involving complex financial market dynamic simulations and multi-factor linkage analysis, often require a significant amount of time for data processing and complex calculations, making it difficult to complete decision outputs in a very short time and failing to meet the real-time requirements of high-frequency trading.

[0003] One-time voting mechanisms reveal significant drawbacks when dealing with conflicting opinions in trading decisions. When trading team members or agents have differing assessments of market trends, simple one-time voting relies solely on the majority opinion, failing to fully leverage the informational value contained within opposing viewpoints.

[0004] With increasingly stringent regulatory environments, transparency and compliance in financial trading decisions have become critical requirements. Regulators expect trading institutions to provide clear explanations of their trading decision-making processes and maintain complete log records for oversight and review. However, many existing intelligent trading decision-making models, especially some deep learning-based black-box models, have internal decision-making logic that is difficult to understand and explain. They cannot intuitively demonstrate how input data is transformed into the final trading decision, nor can they provide detailed logs of the decision-making process.

[0005] A patent search revealed invention patent CN109376922A, which discloses a short-term trading optimization management system and method based on big data. Through big data analysis, it predicts returns and selects the most suitable stocks for daily trading, thereby optimizing and controlling the short-term trading process and improving short-term trading returns. Simultaneously, it tracks market trading changes and adjusts and optimizes stock selection based on market conditions. However, this patent involves a one-time decision-making process, which cannot resolve conflicting viewpoints. It involves a deep debate mechanism among multiple agents, making it difficult to cope with complex and ever-changing market conditions. Furthermore, the decision-making process is opaque and fails to meet regulatory requirements for log recording and explainability.

[0006] In summary, given the problems of the existing technologies, researching an intelligent trading decision-making method and system based on hierarchical multi-round adversarial debate has become a critical task that urgently needs to be addressed. Summary of the Invention

[0007] To address the shortcomings of existing technologies, the purpose of this invention is to provide an intelligent trading decision-making method and system based on hierarchical multi-round adversarial debate.

[0008] The present invention provides an intelligent trading decision-making method based on hierarchical multi-round adversarial debate, comprising the following steps:

[0009] Step S1: Obtain market data and construct input feature vectors, initialize the agent set and divide it into argument generation group and rebuttal group, and assign an initial reputation score to each agent;

[0010] Step S2: Perform R rounds of in-depth debate and dynamically adjust the reputation score;

[0011] Step S3: Based on the adjusted reputation score, perform Soft-Borda dynamic voting and output trading signals.

[0012] Preferably, in step S1, the argument generation group is used to analyze the input feature vector and propose initial trading opinions, which include buy, sell, and wait and see; the rebuttal group is used to question the opinions proposed by the argument generation group and look for potential risks and logical loopholes.

[0013] Preferably, in step S1, the initial reputation score of each agent is set to an unbiased initial value.

[0014] Preferably, step S2 includes the following sub-steps:

[0015] Step S2.1: Perform the viewpoint generation and rebuttal interaction in round r;

[0016] Step S2.2: Update the agent's reputation score based on the Bayesian mechanism;

[0017] Step S2.3: Construct round debate logs and reputation trajectories.

[0018] Preferably, in step S2.1, in each round of debate, the argument generation group... Parallel generation of trading view sets Refutation Group Received trading opinion set Subsequently, based on its own analytical model, historical data, and information from previous rounds of debate, a set of rebuttal arguments was constructed. Used to generate challenge or falsify arguments The viewpoint; evaluate the effectiveness of the arguments or rebuttals put forward by agent i in the current round r, quantifying it as argument-goal consistency. Argument - Market Fit Two key metrics, where A is the set of agents.

[0019] Preferably, in step S2.2, after each round of debate, a Bayesian update mechanism is used to dynamically adjust the reputation score based on the agent's performance in that round. After round r, the reputation score of agent i is... The update formula is as follows:

[0020]

[0021] in:

[0022] It is the reputation score of agent i in the previous round; The initial reputation score for agent i;

[0023] It is the validity likelihood of the arguments or rebuttals put forward by agent i in the r-th round, which is calculated by weighting performance indicators;

[0024]

[0025] in:

[0026] Indicate the consistency between the argument and the objective;

[0027] Indicates the market fit of the argument;

[0028] 、 These represent the empirical weights for consistency with the target and conformity with history, respectively.

[0029] Preferably, in step S2.3, a structured log is generated and stored in each round of debate to achieve complete transparency and traceability of the decision-making process. The log includes the arguments and rebuttals in natural language format generated by each agent, and the key indicators on which each reputation update is based. and Reputation scores before and after the update And the SHAP-ExConsistency explanatory vector used to explain the basis of model decisions.

[0030] Preferably, in step S3, after all R rounds of debate are completed, the comprehensive score of each trading signal in all rounds is calculated based on the debate results of each round. The comprehensive score is based on the Borda scoring method, that is, it comprehensively considers the ranking of each agent on the signal. At the same time, the dynamic temperature coefficient is calculated, and the comprehensive score and the dynamic temperature coefficient are fused through the Soft-Borda dynamic voting formula to output the trading signal with the highest score.

[0031] Preferably, step S3 includes the following sub-steps:

[0032] Step S3.1, calculate the dynamic temperature coefficient. The formula is:

[0033]

[0034] =0.3； = ； It is the realized volatility over the most recent preset time.

[0035] Step S3.2: Calculate the final score for each trading signal using the Soft-Borda dynamic voting formula.

[0036]

[0037] in For input The original value of the function.

[0038] Output the highest-scoring trading signal. When the score difference between the highest and second-highest-scoring signals is less than a preset threshold, output a "wait and see" signal.

[0039] This invention also provides an intelligent trading decision-making system based on hierarchical multi-round adversarial debate, comprising:

[0040] Module M1 acquires market data and constructs input feature vectors, initializes the agent set and divides it into argument generation group and rebuttal group, and assigns an initial reputation score to each agent;

[0041] Module M2 performs R rounds of in-depth debate and dynamically adjusts the reputation score;

[0042] Module M3 performs Soft-Borda dynamic voting and outputs trading signals based on the adjusted reputation score.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. This invention, through the Hierarchical Multi-Round Adversarial Debate Protocol (H-DAP) and Bayesian reputation update technology, can effectively resolve conflicts of opinion among agents, avoid biased decision-making and overfitting, and improve the accuracy of transaction signals compared to traditional single-round voting.

[0045] 2. This invention utilizes a Soft-Borda dynamic voting and temperature adaptive mechanism to overcome the shortcomings of fixed-weight voting in terms of insufficient sensitivity under highly volatile market conditions, effectively reducing risk exposure and maximum drawdown.

[0046] 3. This invention leverages millisecond-level parallel debate and log-based explanation chain technology to achieve a balance between low latency in deep inference and high interpretability of models in high-frequency trading, thus meeting regulatory transparency requirements.

[0047] 4. This invention employs multimodal data fusion technology to integrate information from multiple sources, such as market data and news, eliminating information blind spots in single-source data and enhancing the completeness and transparency of decision-making. Attached Figure Description

[0048] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0049] Figure 1 This is a flowchart of an intelligent trading decision-making method based on hierarchical multi-round adversarial debate in an embodiment of the present invention. Detailed Implementation

[0050] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0051] This invention discloses an intelligent trading decision-making method and system based on hierarchical multi-round adversarial debate, applicable to stock and other financial asset trading scenarios. This method achieves efficient, low-latency, and interpretable intelligent trading decisions by constructing a hierarchical multi-round adversarial debate mechanism, overcoming the limitations of existing technologies in deep reasoning, latency control, and decision transparency. The core of this method lies in constructing a complete process of multi-agent adversarial debate, dynamic reputation assessment, and transparent decision-making. Multi-role agents are divided into an argument generation group and a rebuttal group, executing R rounds of deep debate. During the debate, Bayesian reputation is used to update agent reputations, combined with a Soft-Borda dynamic voting mechanism to generate trading signals within millisecond-level latency. Simultaneously, the transparent decision-making chain through multi-round debate logs and explanatory vectors meets the requirements of financial regulation for model interpretability and real-time auditing.

[0052] Example 1:

[0053] Figure 1 This is a flowchart of an intelligent trading decision-making method based on hierarchical multi-round adversarial debate in an embodiment of the present invention.

[0054] like Figure 1 As shown, this embodiment 1 provides an intelligent trading decision-making method based on hierarchical multi-round adversarial debate, including the following steps:

[0055] Step S1: Obtain market data and construct input feature vectors, initialize the agent set and divide it into argument generation group and rebuttal group, and assign an initial reputation score to each agent.

[0056] Specifically, market data includes market quotes, company announcements, and social media sentiment. The argument generation group analyzes the input feature vector and proposes initial trading opinions, including buy, sell, and wait / see. The refutation group challenges the opinions proposed by the argument generation group and identifies potential risks and logical flaws. Each agent's initial reputation score is set to an unbiased initial value. In this embodiment, all agents have an initial reputation score of 0.5.

[0057] Step S2: Perform R rounds of in-depth debate and dynamically adjust the reputation score.

[0058] In this embodiment, R is an integer greater than 3, and the single-round delay of R rounds of debate does not exceed 50ms.

[0059] Specifically, step S2 includes the following sub-steps:

[0060] Step S2.1: Perform round r of viewpoint generation and rebuttal interaction, where r is an integer from 1 to R.

[0061] Specifically, in each round of debate, the argument generation group Parallel generation of trading view sets Refutation Group Received trading opinion set Subsequently, based on its own analytical model, historical data, and information from previous rounds of debate, a set of rebuttal arguments was constructed. Used to generate challenge or falsify arguments The viewpoint; evaluate the effectiveness of the arguments or rebuttals put forward by agent i in the current round r, quantifying it as argument-goal consistency. Argument - Market Fit Two key metrics, where A is the set of agents.

[0062] Argument-goal consistency measures the alignment between the agent's viewpoint and the pre-defined objective function, which is to maximize the expected return.

[0063] Argument-market fit is used to measure the degree of fit or correlation between an agent's views and historical market trends.

[0064] Step S2.2: Update the agent's reputation score based on the Bayesian mechanism.

[0065] Specifically, after each round of debate, a Bayesian update mechanism is used to dynamically adjust the reputation score based on the agent's performance in that round (i.e., the effectiveness of the arguments). After round r, the reputation score of agent i is... The update formula is as follows:

[0066]

[0067] in:

[0068] It is the reputation score of agent i in the previous round; Let i be the initial reputation score of agent i.

[0069] It is the validity likelihood of the arguments or rebuttals put forward by agent i in the r-th round, which is calculated by weighting performance indicators;

[0070]

[0071] in:

[0072] Indicates the degree of consistency between argument and objective (range).

[0073] Indicates the degree of market fit of the argument (range).

[0074] 、 These represent the empirical weights for consistency with the target and historical conformity, respectively. They are configurable parameters used to balance forward-looking perspectives and historical experience. In this embodiment... =0.6, =0.4.

[0075] Step S2.3: Construct round debate logs and reputation trajectories.

[0076] Specifically, in each round of debate, structured logs are generated and stored to achieve complete transparency and traceability of the decision-making process. The logs include the arguments and rebuttals in natural language format generated by each agent, as well as the key metrics on which each reputation update is based. and Reputation scores before and after the update In addition, the SHAP-ExConsistency interpretation vector is used to explain the basis of the model's decisions, ensuring that the financial regulatory requirements for model interpretability are met.

[0077] Step S3: Based on the adjusted reputation score, perform Soft-Borda dynamic voting and output trading signals.

[0078] Specifically, after completing all R rounds of debate, a comprehensive score is calculated for each trading signal (buy, sell, or wait / see) across all rounds based on the debate results. The comprehensive score is based on the Borda scoring method. This involves comprehensively considering the ranking of each agent's signal; simultaneously calculating the dynamic temperature coefficient; and then using the Soft-Borda dynamic voting formula to fuse the comprehensive score and the dynamic temperature coefficient to output the trading signal with the highest score.

[0079] In this embodiment, step S3 includes the following sub-steps:

[0080] Step S3.1, calculate the dynamic temperature coefficient. The formula is:

[0081]

[0082] =0.3； = ； This is the realized volatility over the last 60 minutes. The greater the market volatility, The smaller the value, the more sensitive the voting results are to high-score signals.

[0083] Step S3.2: Calculate the final score for each trading signal using the Soft-Borda dynamic voting formula.

[0084]

[0085] in For input The original value of the function.

[0086] Output the highest-scoring trading signal. To control risk, when the score difference between the highest and second-highest-scoring signals is less than a preset threshold, it indicates weak market consensus, and a "wait and see" signal is output.

[0087] In this embodiment, the preset threshold is 0.05.

[0088] Example 2:

[0089] The present invention also provides an intelligent trading decision system based on hierarchical multi-round adversarial debate. The intelligent trading decision system based on hierarchical multi-round adversarial debate can be implemented by executing the process steps of the intelligent trading decision method based on hierarchical multi-round adversarial debate. That is, those skilled in the art can understand the intelligent trading decision method based on hierarchical multi-round adversarial debate as a preferred implementation of the intelligent trading decision system based on hierarchical multi-round adversarial debate.

[0090] This intelligent trading decision-making system, based on hierarchical multi-round adversarial debate, includes:

[0091] Module M1 acquires market data and constructs input feature vectors, initializes the agent set and divides it into argument generation group and rebuttal group, and assigns an initial reputation score to each agent;

[0092] Module M2 performs R rounds of in-depth debate and dynamically adjusts the reputation score;

[0093] Module M3 performs Soft-Borda dynamic voting and outputs trading signals based on the adjusted reputation score.

[0094] Specifically, module M2 includes the following sub-modules:

[0095] Module M2.1 performs the interaction of generating and refuting viewpoints in round r, where the value of round r is 1, 2, 3...R.

[0096] Specifically, in module M2.1, in each round of the debate, the argument generation group... Parallel generation of trading view sets Refutation Group Received trading opinion set Subsequently, based on its own analytical model, historical data, and information from previous rounds of debate, a set of rebuttal arguments was constructed. Used to generate challenge or falsify arguments The viewpoint; evaluate the effectiveness of the arguments or rebuttals put forward by agent i in the current round r, quantifying it as argument-goal consistency. Argument - Market Fit Two key metrics, where A is the set of agents.

[0097] Argument-goal consistency measures the alignment between an agent's viewpoint and a predefined objective function (such as "maximizing expected returns") in terms of direction.

[0098] Argument-market fit is used to measure the degree of fit or correlation between an agent's views and historical market trends.

[0099] Module M2.2 updates the agent's reputation score based on a Bayesian mechanism.

[0100] Specifically, in module M2.2, a Bayesian update mechanism is used in each round of debate to dynamically adjust the reputation score based on the agent's performance in that round (i.e., the effectiveness of the arguments). After round r, the reputation score of agent i is... The update formula is as follows:

[0101]

[0102] in:

[0103] It is the reputation score of agent i in the previous round; Let i be the initial reputation score of agent i.

[0104] It is the validity likelihood of the arguments or rebuttals put forward by agent i in the r-th round, which is calculated by weighting performance indicators;

[0105]

[0106] in:

[0107] Indicates the degree of consistency between argument and objective (range).

[0108] Indicates the degree of market fit of the argument (range).

[0109] 、 These represent the empirical weights for target consistency and historical conformity, respectively, and are configurable parameters. In this embodiment... =0.6, =0.4, used to balance forward-looking perspectives with historical experience.

[0110] Module M2.3 constructs round debate logs and reputation tracks.

[0111] Specifically, in module M2.3, a structured log is generated and stored in each round of debate to achieve complete transparency and traceability of the decision-making process. The log includes the arguments and rebuttals in natural language format generated by each agent, as well as the key indicators on which each reputation update is based. and Reputation scores before and after the update In addition, the SHAP-ExConsistency interpretation vector is used to explain the basis of the model's decisions, ensuring that the financial regulatory requirements for model interpretability are met.

[0112] Specifically, module M3 includes the following sub-modules:

[0113] Module M3.1 calculates the dynamic temperature coefficient to adapt to market volatility. The formula is:

[0114]

[0115] =0.3； = ； It is the realized volatility over the most recent preset time (60 minutes in this embodiment). The greater the market volatility, The smaller the value, the more sensitive the voting results are to high-score signals.

[0116] Module M3.2 uses the Soft-Borda dynamic voting formula to calculate the final score for each trading signal:

[0117]

[0118] The highest-scoring trading signal is output. To control risk, when the score difference between the highest and second-highest-scoring signals is less than a preset threshold (the threshold is 0.05 in this embodiment), it indicates that the market consensus is not strong, and a "wait and see" signal is output.

[0119] To verify the actual effectiveness of the present invention, a control experiment was constructed, the details of which are as follows:

[0120] The experimental subjects were selected as the traditional single-round majority voting trading model (Baseline) and compared with the method of this invention. Market data was selected from 5-minute candlestick charts of CSI 300 constituent stocks, company announcements, and financial news headlines, covering the time period from January 2022 to December 2024. A 60-day sliding window rolling backtesting method was adopted, with the test period set from January 2023 to December 2024. To ensure the fairness of the experiment, both models were limited to a maximum of one transaction per day, a holding period of no more than T+3 days, and a uniform transaction fee of 0.2%.

[0121] Key Indicator Comparison

[0122]

[0123] Experimental data show that the method of this invention has a decision latency of 43 milliseconds, which is within the acceptable range for millisecond-level transactions. The annualized return reaches 16.8%, the annualized Sharpe ratio is 2.03, and the maximum drawdown is -6.1%. It achieves 100% interpretability log coverage, recording structured debate logs and explanatory vectors, meeting the financial regulatory requirements for model traceability and interpretability.

[0124] Experimental results show that, compared to existing single-round voting mechanisms, this invention, by introducing multi-round debate and reputation update mechanisms, can more effectively identify noise signals and enhance cross-validation of market opinions. It significantly improves the Sharpe ratio while controlling maximum drawdown, achieving a better balance between return and risk.

[0125] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0126] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A smart trading decision-making method based on hierarchical multi-round adversarial debate, characterized in that, Includes the following steps: Step S1: Obtain market data and construct input feature vectors, initialize the agent set and divide it into argument generation group and rebuttal group, and assign an initial reputation score to each agent; Step S2: Perform R rounds of in-depth debate and dynamically adjust the reputation score; Step S2 includes the following sub-steps: Step S2.1: Perform the viewpoint generation and rebuttal interaction in round r; In step S2.1, in each round of debate, the argument generation group... Parallel generation of trading view sets Refutation Group Received the set of trading viewpoints Subsequently, based on its own analytical model, historical data, and information from previous rounds of debate, a set of rebuttal arguments was constructed. A group used to challenge or disprove the stated arguments. The viewpoint; evaluate the effectiveness of the arguments or rebuttals put forward by agent i in the current round r, quantifying it as argument-goal consistency. Argument - Market Fit Two key metrics, where A represents the set of agents; Step S2.2: Update the agent's reputation score based on the Bayesian mechanism; In step S2.2, after each round of debate, a Bayesian update mechanism is used to dynamically adjust the reputation score based on the agent's performance in that round. After round r, the reputation score of agent i is... The update formula is as follows: in: It is the reputation score of agent i in the previous round; The initial reputation score for agent i; It is the validity likelihood of the arguments or rebuttals put forward by agent i in the r-th round, which is calculated by weighting performance indicators; in: Indicate the consistency between the argument and the objective; Indicates the market fit of the argument; 、 These represent the empirical weights for consistency with the objective and conformity with historical data, respectively. Step S2.3: Construct round debate logs and reputation trajectories; Step S3: Based on the adjusted reputation score, perform Soft-Borda dynamic voting and output trading signals.

2. The intelligent trading decision-making method based on hierarchical multi-round adversarial debate according to claim 1, characterized in that, In step S1, the argument generation group is used to analyze the input feature vector and propose initial trading opinions, which include buying, selling, and waiting; the rebuttal group is used to question the opinions proposed by the argument generation group and look for potential risks and logical loopholes.

3. The intelligent trading decision-making method based on hierarchical multi-round adversarial debate according to claim 1, characterized in that, In step S1, the initial reputation score of each agent is set to an unbiased initial value.

4. The intelligent trading decision-making method based on hierarchical multi-round adversarial debate according to claim 1, characterized in that, In step S2.3, a structured log is generated and stored in each round of debate to achieve complete transparency and traceability of the decision-making process. The log includes the arguments and rebuttals in natural language format generated by each agent, and the key indicators on which each reputation update is based. and Reputation scores before and after the update And the SHAP-ExConsistency explanatory vector used to explain the basis of model decisions.

5. The intelligent trading decision-making method based on hierarchical multi-round adversarial debate according to claim 1, characterized in that, In step S3, after all R rounds of debate are completed, the comprehensive score of each trading signal in all rounds is calculated based on the debate results of each round. The comprehensive score is based on the Borda scoring method, which comprehensively considers the ranking of each agent on the signal. At the same time, the dynamic temperature coefficient is calculated, and the comprehensive score and the dynamic temperature coefficient are fused through the Soft-Borda dynamic voting formula to output the trading signal with the highest score.

6. The intelligent trading decision-making method based on hierarchical multi-round adversarial debate according to claim 5, characterized in that, Step S3 includes the following sub-steps: Step S3.1, calculate the dynamic temperature coefficient. The formula is: =0.3； = ； It is the realized volatility over the most recent preset time. Step S3.2: Calculate the final score for each trading signal using the Soft-Borda dynamic voting formula. Output the highest-scoring trading signal. When the score difference between the highest and second-highest-scoring signals is less than a preset threshold, output a "wait and see" signal.

7. An intelligent trading decision-making system based on hierarchical multi-round adversarial debate, employing the intelligent trading decision-making method based on hierarchical multi-round adversarial debate as described in any one of claims 1-6, characterized in that, include: Module M1 acquires market data and constructs input feature vectors, initializes the agent set and divides it into argument generation group and rebuttal group, and assigns an initial reputation score to each agent; Module M2 performs R rounds of in-depth debate and dynamically adjusts the reputation score; Module M3 performs Soft-Borda dynamic voting and outputs trading signals based on the adjusted reputation score.

Images