Algorithms, Data & Models
Quantitative trading models for prediction markets and traditional assets. Equip up to five to an agent, and combine them to change which signals it trades on.
Probability Market Models
Ten models built for prediction markets, where every price is a probability between 0 and 1. Ordinary chart indicators assume a price that can run to any number, so they misread a market that cannot pass 1: these work in log-odds instead, on Bayesian, information-theoretic and microstructure foundations.
Metrics legend: eff. = effectiveness rating (share of evaluation-set signals that were directionally correct); +WR = measured win-rate improvement vs. an unassisted baseline over the same period, before fees. Badges appear only once a model has independently measured live or paper-forward results — models without a badge have not yet been measured.
Biomimetic Models
Biology-inspired models for traditional asset price action. Equip them to an agent and combine them to change which signals it trades on.
Named Combinations
Sets of 3+ models designed to complement each other. The figure on each card is the combination's design target — the interaction it was built to produce. None has been measured yet, and no combination applies a bonus to your trade outcomes: your agent's results come from the models' own signals. Treat these as hypotheses to test with a backtest or paper-forward run, not as expected returns.
How Combinations Work
A combination is a named set of models designed to work together, drawing on how the corresponding biological systems interact. Equipping one changes which signals your agent acts on. Nothing more. There is no multiplier applied on top.
Central Dogma (HELIX + ENZYME + CRISPR) mirrors the DNA-to-RNA-to-Protein information flow: momentum cascades feed mean reversion timing, filtered through precision execution. Its design target is a +8% win rate versus the same models used separately. That target is a hypothesis about how they interact. It has not been measured, and no dataset, period, sample size, baseline or fee basis stands behind it yet.
Test loadouts against your own market and period before trusting any of them. A backtest or a paper-forward run is the evidence, not the label.