# Paper strategy comparison v2 — fixed before evaluation Authorized November 16, 2026. No brokerage orders. Retain existing baseline results and immutable publication archive. Data: Alpaca SIP raw daily OHLCV, January 2023 through latest completed November 16, 2026 session, current 60-stock universe plus SPY. Use provider-verified forward/reverse splits, retaining raw histories. Warm up at least 210 sessions. Periods: development Jan–Dec 2024; validation Jan–Dec 2025; reserved evaluation Jan–Nov 2026. Require entry and final exit inside a period; do not score boundary-open trades as completed. Earlier portions of the existing 2026 history have already been examined in previous baseline tests: reserved evaluation is not a completely untouched historical dataset or prospective validation. Compare these fixed candidates once. No parameter grid or target win-rate optimization: 1. Original range breakout/breakdown (baseline): 20-session range, volume >=1.2x, >=4/5 daily clues, 1.5 ATR initial risk, half at 2R and half at 3R, 10-session timeout and opposing-trend next-open exit. 2. Market-aligned long breakouts: original upside rules plus stock and SPY close > SMA200 and SMA50 > SMA200; RSI <=75. Same exits as baseline. No shorts. 3. Trend pullback recovery: stock and SPY above rising SMA200, stock SMA50 > SMA200; prior close <= EMA20 and current close crosses above EMA20; RSI 45–65. Next-open entry only within 0.5 ATR of signal close; 1.5 ATR stop, full exit at 2R or 10 sessions. 4. Bull-market oversold bounce: stock and SPY close > SMA200; RSI14 crosses back above 35 after <=35; next-open within 0.5 ATR of signal close; 1.5 ATR stop, full exit at 1.5R or 10 sessions. All decisions use closed bars available on signal date. Next-session open fills. Gap skips, gap-through-stop worse opening fill, stop-first ambiguous daily candles. No same-stock overlapping positions within one strategy. Fixed 0.20% round-trip friction; sensitivity 0.50%. No earnings avoidance, dividends, taxes, borrow or portfolio capital limits simulated. Select one candidate using only development completed trades: >=50 completed trades, positive mean net at 0.50% costs, profit factor >=1.1; highest mean net return after 0.20% costs among eligible candidates. Baseline is comparator, not candidate. Freeze selection before opening validation/evaluation summary. Promotion requires >=30 validation and >=30 reserved trades, positive mean net at both cost levels and profit factor >1 in both later periods. Otherwise retain v1 live watch rules and label candidates research only. Do not change thresholds after outcomes are known. Report every strategy, denominator, winner count, mean net, profit factor, worst trade, ambiguous bars, pending/boundary trades, and dated price coverage. Average position returns are not portfolio returns. Cross-stock trades are correlated. Current-universe selection has survivorship bias. No future 90% probability claim.