Stochastic oscillator 80/20 (14, 3, 3) on GBPUSD, 15-minute bars (spec stoch_spec.md). Run 2026-11-11T06:42:00.953Z; 500 control runs per variant
data: 8297247 minutes 2004-01-01 01:05 .. 2026-09-28 15:46 New York; days and bars: {"forex days":6126,"days with a bar in the window":5903,"bars":563363}; cost 1.5 pips a round trip

=== STO-ZONE ===
  DEV      n= 5142 net -3372.40 pips  avg -0.656  PF 0.96 t -1.28
  VAL      n= 4270 net -7055.70 pips  avg -1.652  PF 0.85 t -4.29
  HOLDOUT  n= 4331 net -6269.50 pips  avg -1.448  PF 0.86 t -4.14
  ALL      n=13743; net -16697.60 pips, avg -1.215 (gross 0.285, cost 1.5); PF 0.90 t -4.84; ended by {"signal":5817,"stop":7498,"close":428}
  in R: median R 20.00 pips; gross 0.014 R, net -0.073 R a trade; cost 0.087 R a trade
  cost scenarios (not gates): 1 pips: avg -0.715, net -9826.10 | 2.5 pips: avg -2.215, net -30440.60
  win rate 5794/13743 = 42.2% (95% Wilson 41.3-43.0%); the 10 best trades 2108.10; drawdown on closed trades -16757.10
  longs n=6993 net -7430.60 gross 3058.90 | shorts n=6750 net -9267.00 gross 858.00
  by year (n, net pips): 2004 543 -1,706 | 2005 526 -1,271 | 2006 521 -1,018 | 2007 541 -456 | 2008 580 1,577 | 2009 620 238 | 2010 615 -132 | 2011 589 397 | 2012 607 -1,000 | 2013 625 -688 | 2014 618 -1,220 | 2015 585 -1,169 | 2016 604 169 | 2017 604 -1,304 | 2018 622 -1,315 | 2019 612 -1,529 | 2020 646 -1,006 | 2021 667 -1,626 | 2022 627 -500 | 2023 652 -1,039 | 2024 635 -163 | 2025 632 -1,319 | 2026 472 -616
  control (the same trades moved to other days at the same London time, 500 runs, 5 s): mean trades 11,796, median avg -1.489, 5-95% -1.829 .. -1.125; real avg -1.215 at percentile 88.00
  [FAIL] 1 PF >= 1.2 in validation: VAL PF 0.85
  [PASS] 2 >= 50 trades in each period: 5142 / 4270 / 4331
  [FAIL] 3 t >= 2.5 on development: DEV t -1.28
  [FAIL] 4 above 95th pct of the control: percentile 88.00
  [FAIL] 5 positive without top-10 trades: -18805.70 pips
  [FAIL] 6 holdout net positive (already used): HOLDOUT net -6269.50 pips PF 0.86
  VERDICT Buried (1/6 gates)
  rule 3: 30 trades open across a suspect gap; without them avg -1.228, gates passed 1/6 -> Buried (verdict unchanged)
  trades per traded day: mean 2.37, max 6; per day with a bar in the window 2.33

=== STO-KD ===
  DEV      n= 5377 net -8363.40 pips  avg -1.555  PF 0.89 t -3.41
  VAL      n= 4366 net -7732.00 pips  avg -1.771  PF 0.83 t -5.03
  HOLDOUT  n= 4573 net -6281.60 pips  avg -1.374  PF 0.85 t -4.28
  ALL      n=14316; net -22377.00 pips, avg -1.563 (gross -0.063, cost 1.5); PF 0.86 t -6.90; ended by {"stop":9715,"signal":4203,"close":398}
  in R: median R 14.00 pips; gross -0.002 R, net -0.137 R a trade; cost 0.134 R a trade
  cost scenarios (not gates): 1 pips: avg -1.063, net -15219.00 | 2.5 pips: avg -2.563, net -36693.00
  win rate 4369/14316 = 30.5% (95% Wilson 29.8-31.3%); the 10 best trades 2010.80; drawdown on closed trades -22382.00
  longs n=7335 net -12028.20 gross -1025.70 | shorts n=6981 net -10348.80 gross 122.70
  by year (n, net pips): 2004 557 -1,789 | 2005 529 -1,585 | 2006 534 -1,260 | 2007 552 -1,190 | 2008 637 -610 | 2009 677 -1,193 | 2010 677 -382 | 2011 628 745 | 2012 586 -1,100 | 2013 636 -1,416 | 2014 602 -1,392 | 2015 623 -962 | 2016 597 845 | 2017 597 -1,748 | 2018 681 -1,408 | 2019 630 -1,650 | 2020 684 -988 | 2021 678 -1,623 | 2022 674 -1,071 | 2023 712 -620 | 2024 656 53 | 2025 680 -1,509 | 2026 489 -525
  control (the same trades moved to other days at the same London time, 500 runs, 5 s): mean trades 12,360, median avg -1.445, 5-95% -1.757 .. -1.112; real avg -1.563 at percentile 28.60
  [FAIL] 1 PF >= 1.2 in validation: VAL PF 0.83
  [PASS] 2 >= 50 trades in each period: 5377 / 4366 / 4573
  [FAIL] 3 t >= 2.5 on development: DEV t -3.41
  [FAIL] 4 above 95th pct of the control: percentile 28.60
  [FAIL] 5 positive without top-10 trades: -24387.80 pips
  [FAIL] 6 holdout net positive (already used): HOLDOUT net -6281.60 pips PF 0.85
  VERDICT Buried (1/6 gates)
  rule 3: 29 trades open across a suspect gap; without them avg -1.579, gates passed 1/6 -> Buried (verdict unchanged)
  trades per traded day: mean 2.54, max 7; per day with a bar in the window 2.43

=== STO-2R ===
  DEV      n= 4388 net -1257.40 pips  avg -0.287  PF 0.98 t -0.46
  VAL      n= 3655 net -5147.90 pips  avg -1.408  PF 0.89 t -3.03
  HOLDOUT  n= 3584 net -6839.30 pips  avg -1.908  PF 0.84 t -4.24
  ALL      n=11627; net -13244.60 pips, avg -1.139 (gross 0.361, cost 1.5); PF 0.92 t -3.66; ended by {"target":3099,"stop":6768,"close":1760}
  in R: median R 20.20 pips; gross 0.008 R, net -0.078 R a trade; cost 0.086 R a trade
  cost scenarios (not gates): 1 pips: avg -0.639, net -7431.10 | 2.5 pips: avg -2.139, net -24871.60
  win rate 4265/11627 = 36.7% (95% Wilson 35.8-37.6%); the 10 best trades 2126.30; drawdown on closed trades -13836.60
  longs n=5979 net -5336.10 gross 3632.40 | shorts n=5648 net -7908.50 gross 563.50
  by year (n, net pips): 2004 489 -1,134 | 2005 478 -1,186 | 2006 455 -1,066 | 2007 468 159 | 2008 482 1,627 | 2009 520 434 | 2010 496 277 | 2011 489 773 | 2012 511 -1,141 | 2013 544 -1,593 | 2014 529 -355 | 2015 513 -1,414 | 2016 505 32 | 2017 522 -452 | 2018 520 37 | 2019 522 -1,403 | 2020 534 -593 | 2021 547 -2,026 | 2022 512 -246 | 2023 547 -798 | 2024 526 -566 | 2025 528 -1,406 | 2026 390 -1,205
  control (the same trades moved to other days at the same London time, 500 runs, 6 s): mean trades 9,750, median avg -1.439, 5-95% -1.882 .. -0.939; real avg -1.139 at percentile 84.80
  [FAIL] 1 PF >= 1.2 in validation: VAL PF 0.89
  [PASS] 2 >= 50 trades in each period: 4388 / 3655 / 3584
  [FAIL] 3 t >= 2.5 on development: DEV t -0.46
  [FAIL] 4 above 95th pct of the control: percentile 84.80
  [FAIL] 5 positive without top-10 trades: -15370.90 pips
  [FAIL] 6 holdout net positive (already used): HOLDOUT net -6839.30 pips PF 0.84
  VERDICT Buried (1/6 gates)
  rule 3: 49 trades open across a suspect gap; without them avg -1.164, gates passed 1/6 -> Buried (verdict unchanged)
  trades per traded day: mean 2.01, max 5; per day with a bar in the window 1.97