Failed breakdown
Older words on this page, kept as written: the “unseen years” are the years before 2023 and the “seen era” the years since; a “gate” is a stop; “null p” is the share of the 5,000 coin flips that did as well; the “walk-forward windows” are the year-by-year check.
Failed11 /100Mark yesterday's regular-hours low and high on ES. If a one-minute bar trades at least 0.5 points under the low and a bar then closes back across it within 60 minutes, buy at the next open. Stop one tick past the lowest low of the flush; flat at 15:30. From 9:30, last entry 15:00, one trade a day, every weekday.
Real, or luck?
Not real, or not worth its costs. 1,362 trades over 18.6 years. Net after costs: -$80,708. On the 15 years before 2023: -$46,336, and 1,216 of 5,000 coin flips did as well or better; 3 of those 15 years made money.
Where the money was
From the record, not from memory.
The stop is the strategy: 1015 of 1362 exits were the stop (-$275,838.00), and the trades that were not stopped made $195,130.10. The numbers are a spike: a 10% nudge either way keeps 0% of the result. Longest wait for a new equity high: 207 months, median year -$2,652.60.
Can you live on it?
Passing and being able to hold it are different things. These numbers are the difference.
Longest wait for a new high: 207 months, ending 2026-08. Worst twelve months: -$27,648, ending 2022-05. A typical year: -$2,653. One contract needs about $94,009 of room to get through its worst stretch (its 95th-percentile drawdown, from the record): an estimate, not a guarantee and not broker margin.
What it costs
Every trade above already paid one tick each way plus commission.
With costs doubled: -$114,758. Tripled: -$148,808. It already loses at the one tick a side every trade paid: there is no fill quality at which this comes out ahead.
If you trade this by eye
The level did not hold: 75% of trades ended at the stop, which sits where the setup says price should not go.
Against 5,000 coin flips the written strategy is a coin (1216 of 5,000 matched or beat it on the unseen years); whatever you do on top of the written steps is where any edge is, and your fills are how to find out which part.
That is what the record says about the strategy as written. It is not what next year says.
What to try next
The one change the record points at, if it points at one.
The strategy's main setting moved 10% either way. The record predicts it fails the same way, because a 10% nudge keeps 0% of the result; the run would show it. Want it run? Put it on the form; it counts as version #2 of this strategy.
The score, in six parts
11 of 100. Each part is a question; the number is how well the record answered it.
The record's own line: Sharpe retention 0.0 (per-trade) -> 0; absolute OOS -> 0
The record's own line: windows profitable 0.2 -> 0; efficiency 1.2 -> 0; window concentration -> 0
The record's own line: null p 0.2432 (5000 seeds) -> 0; DD95/BT 1.06 -> 4; top trade None -> 0
The record's own line: retention@1x 0.0 -> 0; PF@2x 0.62 -> 0
The record's own line: plateau 0.0 -> 0; parsimony 5 -> 2; sign flip True -> 0
The code
The code is withheld until every gate is cleared; the ones it missed are marked.
- ✗ the strategy passed the exam
- ✗ fewer than 1 in 20 coin flips matched it on 2008 to 2022
- ✗ the money on those years is clearly more than zero
- ✗ it still makes money with at least four ticks of cost a side
- ✗ no single trade is more than a quarter of the money
How it was scored: Scored under Protocol v1.1 on FirstRate 1-min unadjusted via Thinridge engine. Intake testimony: NOT received (strategy tested as locked). Contamination map sworn: no (seen era as declared or by archive date). Human review of reconciliation residue: NO.