Loading status…

The first version of the bot ran 12 strategies sourced from 10 different traders. After 3 years of backtest data, only 1 had positive expectancy and most were losing money in patterns consistent with structural problems (universe-method mismatch, stripped filters, calendar-based partials that hurt returns).

The narrowed set keeps three traders whose published rules survived re-implementation: Kullamägi (3 swing setups), Tom Hougaard (intraday ORB — currently disabled pending data refresh), and Lance Breitstein (a discipline harness applied to every entry rather than its own signal generator). The other seven trader writeups remain at /ten.html as a record of how the search funnel worked.

~60 parameter variants were tested in parallel via bot/scripts/sweep.py. Four findings stuck:

• 3-of-3 momentum percentile rule (top X% on 1m AND 3m AND 6m simultaneously) beat 2-of-3 by ~0.4 OOS PF on the breakout. This is the corpus's literal "simultaneously" rule and it works.
• Lance's prior-bar ratchet trail at 0.3R arming lifted EP win rate from 67% to 78%. Earlier ratcheting catches more of the small wins that would otherwise revert.
• Looser parabolic-short run thresholds (large-cap 30%, small-cap 150%) caught more daily-bar parabolic candidates than the corpus's literal 50%/300% rule, which was tuned for intraday signals.
• Earnings blackout was a no-op. yfinance has data for most leaders so the gate never blocks anything; the few times it would block are dominated by other gates.

What didn't work: confirmation candle (cuts the best winners that gap and run), wide universe of 1000+ tickers (more noise than signal), strict cap requirement on the parabolic short (kills all trades because most momentum names lack share-count metadata).

• Breitstein's A+ capitulation fade. Pure tape-reading; mechanical attempts produce the canonical "bot shorts the bottom tick" disaster. His teaching is mechanized as the discipline harness instead.
• Hougaard's discretionary size-ups. His edge is psychological — "when I feel conviction, 3x size" cannot be coded.
• Kullamägi's theme/sector pick. The corpus shows ~50% of his winners come from 1-2 dominant themes per period, identified by reading the tape. The leader scanner approximates this but doesn't have a real theme-momentum scorer.
• News and fundamentals. The corpus is explicit: don't predict the news, wait for the price reaction. We follow that.