Skip to main content

114. Inverted Opportunity Discovery

Status: Accepted Date: 2025-07-06

Context​

A traditional automated trading or analysis system operates in a linear fashion:

  1. Look at a specific market (e.g., BTC/USDT).
  2. Analyze its data (price, indicators).
  3. Based on the analysis, decide on an action (buy, sell, wait).
  4. Repeat for the next market.

This approach is market-centric. The core question is "What should I do in this market?" The strategy is encoded in the decision logic.

Decision​

The Tyche module will implement an inverted opportunity discovery model.

Instead of starting with the market, we start with a well-defined opportunity pattern, which we call a "spell". A spell is a description of a desirable market setup, expressed in natural language (e.g., "a market that is currently overextended and showing signs of RSI divergence, making it a good candidate for a short-term reversal").

The system's core function is not to analyze a given market, but to take a given spell and search the entire universe of available markets to find the one that best matches that pattern.

This is achieved by using the AI-powered pairwise comparison engine (adr://ai-powered-comparisons) within the Dike tournament structure (adr://tournament-integration). The spell is used as the core of the prompt for the AI, defining the criteria for what makes one market "better" than another.

Consequences​

Positive:

  • Strategy-Centric Design: This approach puts the trading strategy (the "spell") front and center. It allows us to think about and define our strategies first, and then let the system find the best place to apply them.
  • Efficient Discovery: Instead of analyzing every market in detail, we use the efficient elimination process of a tournament to quickly zero in on the most promising candidate.
  • Scalability: This model scales well. To add a new strategy, we simply write a new spell (a new prompt), we don't need to change the underlying analysis code. To cover more markets, we just add them to the tournament pool.
  • Reduces Noise: It focuses our attention only on the highest-probability setups that match our predefined criteria, rather than getting distracted by marginal setups in every market.

Negative:

  • Requires Predefined Patterns: This model is only as good as the spells we define. It cannot discover entirely new types of opportunities that don't fit into one of our existing patterns.
  • Potential for Overfitting: There is a risk of crafting spells that are too specific or "overfit" to past market behavior, which may not perform well in the future.

Mitigation:

  • Portfolio of Spells: We will develop and maintain a diverse portfolio of spells, ranging from very specific to more general patterns, to ensure broad market coverage.
  • Continuous Spell Evaluation: The performance of trades originating from each spell will be tracked rigorously. Underperforming spells will be refined or retired, just like any other part of the trading strategy.
  • Complementary Analysis: This inverted discovery model is one tool among many. It can be complemented by other, more traditional market-centric analysis tools and dashboards for human traders.