Supertrend (ATR 10, x3) on XAUUSD 1-hour bars (spec stau_spec.md). Run 2026-11-06T12:15:06.257Z; 500 control runs per variant. Results in R (1.5 x ATR14 of the 1-hour bars at the signal bar).
data: 7067434 one-minute bars, 2004-06-11 00:19 .. 2026-10-07 16:57 New York (minute close stamps)
1-hour bars: 129857; positions from bar 500 (2004-07-14 23:00); direction changes after that: 3163

=== ST-LS: 1-hour bars, long and short (always in the market) ===
  DEV      n= 1215 net 206.76 R  avg 0.170 R  PF 1.16 t 1.92
  VAL      n= 1005 net 54.21 R  avg 0.054 R  PF 1.05 t 0.51
  HOLDOUT  n=  944 net 151.00 R  avg 0.160 R  PF 1.15 t 1.46
  ALL      n=3164 net 411.97 R avg 0.130 R (gross 0.198 R, cost 0.068 R a trade); $ per ounce 2558.61; PF 1.12 t 2.24
  win rate 1195/3164 = 37.8% (95% Wilson 36.1-39.5%); ended by {"reversal":3163,"end of data":1}; hours held median 44.00 max 437.00; time in the market 100.0%
  the 10 best trades 247.48 R; net without the 10 best 164.49 R, without the 20 best 27.87 R
  worst open loss: -16.51 R; trades that were 3 R or more against the position at some point: 262
  longs n=1582 net 547.23 R gross 654.47 R | shorts n=1582 net -135.26 R gross -28.04 R
  made across market closures of 36 h or more (weekends, holidays): 35.29 R (356.96 $ per ounce) of a gross 626.43 R
  drawdown on closed trades -75.28 R; including open losses -76.00 R; financing: scenario A (3% a year) net 244.33 R, scenario B (today's swap) net 208.39 R
  by year (n, net R): 2004 67 -14.63 | 2005 137 56.24 | 2006 157 34.12 | 2007 145 34.46 | 2008 144 66.89 | 2009 146 17.87 | 2010 146 -32.50 | 2011 127 52.69 | 2012 146 -8.39 | 2013 152 36.94 | 2014 131 14.87 | 2015 144 -11.75 | 2016 148 -24.45 | 2017 141 6.24 | 2018 148 -7.10 | 2019 141 39.45 | 2020 109 108.61 | 2021 146 -1.56 | 2022 148 3.52 | 2023 149 -14.51 | 2024 143 16.20 | 2025 145 17.92 | 2026 104 20.81
  control (direction series shifted in time, 500 runs, 4 s): mean trades 3164.94, median avg -0.060 R, 5-95% -0.162 .. 0.039 R; real avg 0.130 R at percentile 99.80
  [FAIL] 1 PF >= 1.2 in validation: VAL PF 1.05
  [PASS] 2 >= 50 trades in each period: 1215 / 1005 / 944
  [FAIL] 3 t >= 2.5 on development: DEV t 1.92
  [PASS] 4 above 95th pct of the control: percentile 99.80
  [PASS] 5 positive without top-10 trades: 164.49 R
  [PASS] 6 holdout net positive (flagged used): HOLDOUT net 151.00 R PF 1.15
  VERDICT Buried (4/6 gates)
  rule 3: 917 trades are open across a suspect or unclear gap; without them avg -0.190 R, gates passed 2/6 -> Buried (verdict unchanged)

=== ST-L: 1-hour bars, long while green, flat while red ===
  DEV      n=  608 net 304.93 R  avg 0.502 R  PF 1.56 t 3.78
  VAL      n=  502 net 11.07 R  avg 0.022 R  PF 1.02 t 0.16
  HOLDOUT  n=  472 net 231.23 R  avg 0.490 R  PF 1.51 t 2.84
  ALL      n=1582 net 547.23 R avg 0.346 R (gross 0.414 R, cost 0.068 R a trade); $ per ounce 3132.76; PF 1.35 t 4.08
  win rate 639/1582 = 40.4% (95% Wilson 38.0-42.8%); ended by {"turned red":1582}; hours held median 47.00 max 340.00; time in the market 52.8%
  the 10 best trades 200.24 R; net without the 10 best 346.99 R, without the 20 best 222.57 R
  worst open loss: -16.51 R; trades that were 3 R or more against the position at some point: 126
  longs n=1582 net 547.23 R gross 654.47 R | shorts n=0 net 0.00 R gross 0.00 R
  made across market closures of 36 h or more (weekends, holidays): 86.18 R (441.01 $ per ounce) of a gross 654.47 R
  drawdown on closed trades -73.91 R; including open losses -75.60 R; financing: scenario A (3% a year) net 458.29 R, scenario B (today's swap) net 7.21 R
  by year (n, net R): 2004 34 0.54 | 2005 68 66.67 | 2006 79 49.63 | 2007 72 57.34 | 2008 72 35.89 | 2009 73 37.27 | 2010 73 19.88 | 2011 64 44.52 | 2012 73 -6.80 | 2013 76 -35.30 | 2014 65 -7.65 | 2015 72 -18.90 | 2016 74 5.69 | 2017 71 25.70 | 2018 74 -9.38 | 2019 70 50.91 | 2020 55 91.69 | 2021 73 -10.42 | 2022 74 6.63 | 2023 74 11.32 | 2024 72 50.91 | 2025 72 75.49 | 2026 52 5.61
  control (direction series shifted in time, 500 runs, 2 s): mean trades 1582.46, median avg 0.171 R, 5-95% 0.070 .. 0.270 R; real avg 0.346 R at percentile 99.60
  [FAIL] 1 PF >= 1.2 in validation: VAL PF 1.02
  [PASS] 2 >= 50 trades in each period: 608 / 502 / 472
  [PASS] 3 t >= 2.5 on development: DEV t 3.78
  [PASS] 4 above 95th pct of the control: percentile 99.60
  [PASS] 5 positive without top-10 trades: 346.99 R
  [PASS] 6 holdout net positive (flagged used): HOLDOUT net 231.23 R PF 1.51
  VERDICT Buried (5/6 gates)
  rule 3: 461 trades are open across a suspect or unclear gap; without them avg -0.023 R, gates passed 2/6 -> Buried (verdict unchanged)

buy and hold over the same data, for reference: 384 -> 4109.91