Source: homily, 'Larry Connors RSI2 mean-reversion strategy', FMZ digest (https://www.fmz.com/digest-topic/5614), sections 'Strategy' steps 1-4 and 'Trading Examples' (DIA). The article describes Connors' RSI(2) rule: take the main trend from a 200-day SMA, buy when RSI(2) falls to 5 or lower above it, and exit on a close above the 5-day SMA. It then ports the rule to OKEx BTC quarterly futures on 15m bars with a 70-bar MA. That port also flips the logic into counter-trend (short RSI2>90 above the MA, long RSI2<10 below it). WHAT IS NOT USED, AND WHY: the 15m crypto port. Our corpus has already falsified crypto RSI(2)/short-horizon oversold fades at fee cost: ADA USD-M 2h RSI-2 long-short (1059 trades, PF 0.78), SOL daily RSI oversold (PF 0.69), BTC COIN-M 1H/4H reversion (PF 0.82/0.95), SOL/XRP/DOGE 15-30m reversion (PF 0.68-0.79). Crypto returns have weak or positive short-horizon autocorrelation, so a 2-bar oversold reading has no reversion to sell. WHAT IS USED: Connors' ORIGINAL mechanism on the instruments it was published for, US broad-index ETFs. These have a documented short-horizon negative autocorrelation at the index level: liquidity providers absorb multi-day index selling from ETF, futures and rebalancing flows and are paid a reversal premium. The rules are Connors' published rules with the article's values frozen: SMA200, RSI(2) (Wilder), entry threshold 5, exit close > SMA5. Each ETF is traded as an independent single-leg long-only sleeve. This is NOT a neutral book, NOT a ranking and NOT a rotation. Instruments: SPY, QQQ, IWM, DIA. These are the four broad US index ETFs, and the list is not a hand-picked survivor set: index ETFs do not delist, and DIA is the source article's own example. Histories run 1993/1998/1999/2000 to 2026, about 26 years on the full four-ETF set. The key test is pre-registered. Connors published this rule in 2008-2009 ('Short Term Trading Strategies That Work'), so every bar from 2009 onward is genuine post-publication out-of-sample data. The analyst should report the 2009-2026 slice separately from 1993-2008. A decay from ~0.8%/trade pre-2009 to ~0.3%/trade post-2009 is the expected result. Sub-0.1%/trade post-2009 falsifies the hypothesis. Selection choices, stated honestly (L122): venue USEQ (the only venue where this mechanism has an economic agent paying it); timeframe 1-DAY (the published frequency, and the only USEQ timeframe with decades of history); instruments = the 4 broad index ETFs (the published domain); no gate beyond Connors' own SMA200, which is part of the published rule and not a fitted addition. A variant WITHOUT the SMA200 filter is pre-registered as secondary, for separate scoring only (L130). Flat state: flat about 80-85% of the time per sleeve; each position lasts about 3-5 sessions.