Primitive Documentation

Algorithm Library:
Methods, Not “Strategies”

AlgoBoy maintains a library focused on mechanisms — not trading tips or packaged strategies. This library exists to document how algorithms are built, why they behave the way they do, and where they fail.

✓ What This Is

  • A structured catalogue of algorithmic methods
  • A reference for assumptions & trade-offs
  • A technical resource for builders
  • A foundation for reproducible research

× What This Is Not

  • A list of profitable strategies
  • A signal selling service
  • A set of black-box models
  • A promise of financial returns

If a method cannot be clearly described, parameterized, and stress-tested, it does not belong here.

Method-First Thinking

We approach algorithms as composable components. Most real-world systems are built from layers: data conditioning, feature construction, signal generation, and risk constraints. We isolate these layers to allow methods to be compared across markets and reused without narrative bias.

Core Method Categories

1

Mean Reversion

Methods that assume price, odds, or probabilities revert toward a reference level.

Key Considerations

Time horizon sensitivity, liquidity dependence, and regime fragility.

Common Failure: Reversion assumptions breaking during structural shifts or crowding.
2

Momentum & Trend

Methods that exploit directional persistence over defined horizons.

Key Considerations

Entry timing vs continuation, volatility-adjusted momentum, signal decay.

Common Failure: Momentum becoming noise in low-participation or range-bound conditions.
3

Relative Value

Algorithms that compare entities against each other rather than in isolation (e.g. Runner vs Field).

Key Considerations

Reference group stability, hidden correlations, cross-contamination.

Common Failure: Benchmark drift invalidating comparisons.
4

Market Microstructure

Exploiting structural properties: order book imbalance, queue position, and spread compression.

Key Considerations

Venue-specific rules, latency sensitivity, data completeness.

Common Failure: Assuming identical behavior across different venues.
5

Probabilistic & Odds

Primarily for betting exchanges. Implied probability drift, overround normalization, and price efficiency.

Key Considerations

Liquidity timing, participant behavior, market closure dynamics.

Common Failure: Treating odds as stable probabilities instead of negotiated prices.
6

Regime & State-Based

Methods that adapt behavior based on detected market conditions (Volatility regimes, etc).

Key Considerations

Detection lag, false regime switches, overfitting definitions.

Common Failure: Defining more regimes than the data can statistically support.
7

Anomaly Detection

Flagging deviations rather than predicting direction. Volume anomalies, probability compression.

Key Considerations

False positives, data artifacts, reaction vs anticipation.

Common Failure: Confusing noise with signal under sparse data conditions.

Cross-Market Applicability

A core design goal of the library is cross-market relevance. We evaluate methods across Traditional Financial Markets, Betting Exchanges, and select Crypto venues. This often reveals uncomfortable truths: methods that "work" in one market fail elsewhere because structural differences matter more than indicators.

Failure Is Part of the Documentation

AlgoBoy explicitly documents when methods stop working, how performance decays, and what conditions invalidate results. Negative results are not hidden. If a method fails consistently under certain conditions, that failure is part of the method’s definition.