RSI(2) on the DAX, daily closes, Connors' rules replicated from SPY (spec rdx_spec.md). Run 2026-11-12T07:53:45.160Z; 500 control runs per variant. Results in net % per trade, 0.04% cost per round trip, fills at the close.
data: 9814 days, 1987-12-30 .. 2026-10-20; signals from 1988-12-27; performance index (dividends inside the closes); gap days (rule 3): 1988-12-30, 1989-12-29

=== RD2-10: buy at the close when close > SMA200 and RSI(2) < 10; sell at the first close above SMA5 ===
  DEV      n= 119 net 38.33 %  avg 0.322 %  PF 1.58 t 1.80
  VAL      n=  75 net -9.79 %  avg -0.131 %  PF 0.87 t -0.36
  HOLDOUT  n=  76 net 12.95 %  avg 0.170 %  PF 1.30 t 0.78
  ALL      n=270 net 41.49 % avg 0.154 % (gross 0.194 %); PF 1.22 t 1.08; with 0.10% cost avg 0.094 %
  win rate 179/270 = 66.3% (95% Wilson 60.5%-71.7%); trading days held median 3 max 17; open at the end: 0
  the 10 best trades 34.80 %; net without the 10 best 6.69 %, without the 20 best -19.99 %
  worst open loss at a close (open positions included): -21.93 % (trade entered 2011-07-28, exited 2011-08-12)
  next-open fills (descriptive, trades entered from 1994-01-01): n=228 avg 0.253 % PF 1.36 t 1.50; the same entries at the close: n=228 avg 0.174 %; next-open by period avg 0.697 / -0.174 / 0.225 %
  whole strategy from 1988-12-27 with no interest while out: growth x1.40 (0.9% a year), max drawdown -28.5% (2005-10-17 to 2011-08-10), volatility 7.6%, invested 11.2% of days; buy and hold x18.85 (8.1% a year), max drawdown -72.7%, volatility 21.7%
  crashes, SPY's calendar windows (strategy / buy and hold): 2000-02 dot-com 1.1% / -67.2%; 2008-09 financial crisis -14.5% / -53.7%; 2020 Covid -7.0% / -36.6%; 2022 rate hikes 1.5% / -24.0%
  shared entries: of the 236 entries from 1993-01-29 (SPY's first day), 45 on a SPY RS2-10 entry date and 27 on the next DAX day after one; 30 of all 270 on a QQQ RQ2-10 entry date
  by year (n, net %): 1989 12 -0.35 | 1990 9 -1.31 | 1991 7 -0.96 | 1992 6 -2.04 | 1993 8 6.37 | 1994 4 -4.96 | 1995 5 3.32 | 1996 8 1.14 | 1997 9 17.17 | 1998 6 1.64 | 1999 7 -2.71 | 2000 11 7.10 | 2002 2 -0.99 | 2003 4 4.91 | 2004 9 3.54 | 2005 12 6.48 | 2006 5 -2.18 | 2007 8 -2.73 | 2008 1 -12.74 | 2009 6 5.04 | 2010 12 7.74 | 2011 6 -17.37 | 2012 9 12.69 | 2013 12 12.48 | 2014 10 -7.51 | 2015 6 -5.22 | 2016 4 -0.32 | 2017 11 3.40 | 2018 1 -6.55 | 2019 6 0.91 | 2020 7 -7.71 | 2021 7 2.52 | 2022 3 2.98 | 2023 10 6.68 | 2024 8 1.78 | 2025 9 5.62 | 2026 10 3.62
  control (random entries on dips in an uptrend: close > SMA200 and < SMA5; p 1.203e-1; 500 runs, 0 s): mean trades 263.71, median avg 0.168 %, 5-95% 0.006 .. 0.320 %; real avg 0.154 % at percentile 45.00
  [FAIL] 1 PF >= 1.2 in validation: VAL PF 0.87
  [PASS] 2 >= 50 trades in each period: 119 / 75 / 76
  [FAIL] 3 t >= 2.5 on development: DEV t 1.80
  [FAIL] 4 above 95th pct of the control: percentile 45.00
  [PASS] 5 positive without top-10 trades: 6.69 %
  [PASS] 6 holdout net positive (partly used): HOLDOUT net 12.95 % PF 1.30
  rule 3: 0 trade(s) open across a gap day
  VERDICT Buried (3/6 gates)

=== RD2-5: buy at the close when close > SMA200 and RSI(2) < 5; sell at the first close above SMA5 ===
  DEV      n=  63 net 29.50 %  avg 0.468 %  PF 1.90 t 1.90
  VAL      n=  39 net 10.62 %  avg 0.272 %  PF 1.34 t 0.54
  HOLDOUT  n=  36 net 8.50 %  avg 0.236 %  PF 1.44 t 0.74
  ALL      n=138 net 48.63 % avg 0.352 % (gross 0.392 %); PF 1.58 t 1.77; with 0.10% cost avg 0.292 %
  win rate 100/138 = 72.5% (95% Wilson 64.5%-79.2%); trading days held median 3 max 11; open at the end: 0
  the 10 best trades 30.25 %; net without the 10 best 18.38 %, without the 20 best -4.04 %
  worst open loss at a close (open positions included): -21.59 % (trade entered 2011-07-29, exited 2011-08-12)
  next-open fills (descriptive, trades entered from 1994-01-01): n=116 avg 0.601 % PF 2.03 t 2.43; the same entries at the close: n=116 avg 0.426 %; next-open by period avg 0.790 / 0.375 / 0.630 %
  whole strategy from 1988-12-27 with no interest while out: growth x1.56 (1.2% a year), max drawdown -23.1% (2007-10-25 to 2011-08-10), volatility 5.1%, invested 5.4% of days; buy and hold x18.85 (8.1% a year), max drawdown -72.7%, volatility 21.7%
  crashes, SPY's calendar windows (strategy / buy and hold): 2000-02 dot-com 3.1% / -67.2%; 2008-09 financial crisis 0.9% / -53.7%; 2020 Covid -7.0% / -36.6%; 2022 rate hikes 0.0% / -24.0%
  shared entries: of the 121 entries from 1993-01-29 (SPY's first day), 20 on a SPY RS2-10 entry date and 26 on the next DAX day after one; 21 of all 138 on a QQQ RQ2-10 entry date
  by year (n, net %): 1989 5 0.66 | 1990 5 -2.73 | 1991 3 -0.82 | 1992 4 -0.61 | 1993 5 2.67 | 1994 2 -4.84 | 1995 4 3.37 | 1996 4 -0.00 | 1997 5 14.33 | 1998 4 2.07 | 1999 5 -0.16 | 2000 6 7.72 | 2002 1 1.60 | 2003 1 0.64 | 2004 3 0.28 | 2005 6 5.34 | 2006 3 -4.42 | 2007 7 7.18 | 2009 2 -2.19 | 2010 2 1.66 | 2011 5 -16.78 | 2012 5 6.56 | 2013 6 9.39 | 2014 5 4.30 | 2015 4 4.91 | 2016 2 -0.86 | 2017 5 -0.59 | 2018 1 -5.15 | 2019 2 1.39 | 2020 3 -2.30 | 2021 3 1.35 | 2023 5 3.28 | 2024 4 3.53 | 2025 5 1.86 | 2026 6 6.00
  control (random entries on dips in an uptrend: close > SMA200 and < SMA5; p 5.301e-2; 500 runs, 0 s): mean trades 128.63, median avg 0.182 %, 5-95% -0.064 .. 0.422 %; real avg 0.352 % at percentile 87.60
  [PASS] 1 PF >= 1.2 in validation: VAL PF 1.34
  [FAIL] 2 >= 50 trades in each period: 63 / 39 / 36
  [FAIL] 3 t >= 2.5 on development: DEV t 1.90
  [FAIL] 4 above 95th pct of the control: percentile 87.60
  [PASS] 5 positive without top-10 trades: 18.38 %
  [PASS] 6 holdout net positive (partly used): HOLDOUT net 8.50 % PF 1.44
  rule 3: 0 trade(s) open across a gap day
  VERDICT Buried (3/6 gates)

buy and hold from 1988-12-27: x18.85, 8.1% a year, max drawdown -72.7% (2000-03-07 to 2003-03-12)