Hypothesis Pipeline
+ Add hypothesis
Idea Queue
Ranked by IRR/effort ratio. Best ideas at top.
+ Add idea
Factor Zoo (top 20 by |IC|)
| Factor | Family | IC 90d | n | p-value | Verdict |
|---|
Research Log
Capture observations, hypotheses, dis-confirmations, aha moments.
Last 7 days
Backtest Archive
| Strategy | Run At | Sharpe | Total Return | Max DD | Trades |
|---|
Timeframe Agreement (1D / 1W / 1M)
Loading agreement heatmap…
Counter-Thesis Viewer
Enter a ticker to generate a counter-thesis — reasons the thesis might be wrong.
Anomaly WHY Narration
Loading anomaly narrations…
Factor IC Leaderboard
Rolling 90d Spearman IC per factor — ranked by absolute predictive power.
Loading factor IC data…
Methodology & rigor
Hypothesis pipeline: each idea moves through Stage 1 (literature review) → Stage 2 (in-sample backtest) → Stage 3 (walk-forward purged CV) → Stage 4 (live shadow ≥30 days) → Stage 5 (real-money allocation). Promotion requires explicit pre-registered success criteria — no after-the-fact threshold revisions.
Factor Zoo (top 20 by |IC|): information coefficient = rank correlation between factor value at t and forward return at t+h. Each factor's IC reported with its standard error (Fisher transform 95% CI). Factors with ICIR < 0.2 are flagged as noise.
Backtest archive: every published backtest stored with its run-time hash so results aren't quietly tweaked. Sharpe annualized at 252 trading days; max DD computed peak-to-trough.
Bias controls: walk-forward purging (Lopez de Prado 2018 "Advances in Financial Machine Learning") prevents look-ahead leakage; combinatorial cross-validation reduces backtest overfitting; deflated Sharpe ratio (Bailey & Lopez de Prado 2014) penalizes multiple-testing inflation.
Limitations: published Sharpes are gross of slippage + commissions; small-cap and crypto strategies likely overstate live performance by 15-30% due to liquidity assumptions; survivorship bias possible for delisted-during-period names not yet backfilled.
Sources: Lopez de Prado (2018), Bailey & Lopez de Prado (2014) "The deflated Sharpe ratio", Harvey, Liu & Zhu (2016) "...and the cross-section of expected returns" (multiple-testing penalty).
See: AI & methodology hub · Scoring Model Card · Factor Zoo card · Walk-forward CV card