The pre-FOMC announcement drift on SPY (spec pf_spec.md). Run 2026-11-04T11:50:29.634Z; 500 control runs per hold. Trade results in % of the price (total return, close to close), cost 0.04% a round trip.
data: 8489 days 1993-01-29 .. 2026-10-20; 136 dividends; announcement days from data_fomc/fomc_scheduled.csv; cash = 3-month T-bill (FRED DTB3)

=== PF: buy at the close of the day before each scheduled FOMC announcement, sell at the close of the announcement day ===
  DEV      n=  96 net 15.35%  avg 0.160%  PF 1.51 t 1.59 win 58.3% | overnight 0.038% day 0.163% (before costs)
  VAL      n=  80 net 36.56%  avg 0.457%  PF 2.61 t 2.97 win 60.0% | overnight 0.215% day 0.282% (before costs)
  HOLDOUT  n=  85 net -0.82%  avg -0.010%  PF 0.98 t -0.08 win 41.2% | overnight 0.164% day -0.134% (before costs)
  ALL      n=261 net 51.09% (sum of trade %) avg 0.196% (gross 0.236%); PF 1.59 t 2.72; at 0.10% cost: sum 35.43% | overnight 0.133% day 0.103%
  winners 139/261 = 53.3% (95% Wilson 47.2%-59.2%); without the 10 best 18.46%, without the 20 best -0.64%
  10 best: 2008-12-16 4.67%, 2011-08-09 4.61%, 2008-03-18 4.11%, 2009-01-28 3.34%, 2022-05-04 3.01%, 2007-09-18 2.90%, 2008-08-05 2.66%, 2020-04-29 2.58%, 2022-07-27 2.56%, 2009-03-18 2.20%
  10 worst: 2024-12-18 -3.02%, 2011-09-21 -2.99%, 2007-12-11 -2.78%, 2001-03-20 -2.72%, 2022-11-02 -2.55%, 2021-01-27 -2.48%, 1994-02-04 -2.31%, 1997-11-12 -2.11%, 2000-12-19 -2.07%, 1997-02-05 -1.92%
  by year (n, sum %): 1994 8 -0.96 | 1995 8 4.10 | 1996 8 1.97 | 1997 8 -1.69 | 1998 8 3.79 | 1999 8 4.42 | 2000 8 0.24 | 2001 8 -2.26 | 2002 8 0.81 | 2003 8 3.93 | 2004 8 1.67 | 2005 8 -0.67 | 2006 8 0.71 | 2007 8 4.56 | 2008 8 11.33 | 2009 8 8.94 | 2010 8 1.15 | 2011 8 1.37 | 2012 8 4.73 | 2013 8 0.17 | 2014 8 1.18 | 2015 8 2.41 | 2016 8 -1.17 | 2017 8 0.50 | 2018 8 -3.60 | 2019 8 -0.01 | 2020 7 4.64 | 2021 8 0.10 | 2022 8 3.82 | 2023 8 0.01 | 2024 8 -1.41 | 2025 8 1.19 | 2026 6 -4.89
  by weekday of the announcement (n, avg %, PF): Fri 1 -- -- | Tue 85 0.334 2.05 | Wed 166 0.113 1.34 | Thu 9 0.693 6.56
  every other day of the file (same cost) n=8227 avg 0.002% (gross 0.042%) win 51.9%
  control (one-day holds on random days of the same month and weekday, no announcement day, 500 runs, fallbacks 0, 0 s): median avg 0.015%, 5-95% -0.075 .. 0.117%; real avg 0.196% at percentile 99.80
  [PASS] 1 PF >= 1.2 in validation: VAL PF 2.61
  [PASS] 2 >= 50 trades in each period: 96 / 80 / 85
  [FAIL] 3 t >= 2.0 on development: DEV t 1.59
  [PASS] 4 above 95th pct of the control: percentile 99.80
  [PASS] 5 positive without top-10 trades: 18.457%
  [FAIL] 6 holdout net positive (flagged used): HOLDOUT net -0.823% PF 0.98
  VERDICT Buried (4/6 gates)
  whole strategy with cash interest, 1994-02-03 .. 2026-10-20: x3.66, 4.0% a year, worst drawdown -4.1% (2000-12-18 -> 2001-03-20), invested 3.2% of days
  buy-and-hold SPY (total return), same dates: x28.53, 10.8% a year, worst drawdown -55.2%
rule 3: not applicable (the only gaps in the SPY file are real market closures)

=== Neighbouring holds (descriptive, after the gates) ===
  Day before  n=261 (96/80/85) avg 0.097% PF 1.29 DEV t 1.66 VAL PF 1.11 HOLD 3.838% pct 91.00 w/o top10 -8.34% -> 2/6
  Two days    n=261 (96/80/85) avg 0.330% PF 1.80 DEV t 2.67 VAL PF 2.36 HOLD 6.508% pct 100.00 w/o top10 39.28% -> 6/6
  Day after   n=261 (96/80/85) avg -0.061% PF 0.88 DEV t 1.49 VAL PF 0.59 HOLD -10.018% pct 6.80 w/o top10 -41.92% -> 1/6