FST Live attribution (exploratory read of the paper book)
Where did the FST Live paper book make and lose money, by signal family, direction, underlying, days to expiry, entry time, holding time, exit reason, premium band and month?
This is a retrospective, exploratory description, not a pre-registered test. Across 179 closed round trips in the FST Live paper book from 7 August to 10 November 2026 (38 days with closes, 68 weekdays first to last), gross realized P&L was -$3,574.00: mean -$19.97 a trade, with a 95% bootstrap band of -$48.12 to +$12.70 that includes zero, a 49% win rate, an average win of $127.01 against an average loss of $162.10, and 88 wins against 91 losses. The losses were concentrated in single-name trades (66 trades, -$3,473.00, mean -$52.62, band -$90.09 to -$12.77), trades entered 2 to 5 days before expiry (42, -$3,326.00, mean -$79.19), trades entered in the first half hour (49, -$2,572.00, mean -$52.49) and stop exits (83, -$14,377.00). Index-ETF trades made +$1,236.00 on 102 trades (mean +$12.12, band -$24.85 to +$58.11), and trades held under 15 minutes made +$3,003.00 on 64 (mean +$46.92); those holding-time and exit-reason splits are outcomes of the exit rule, not signals. Twenty-five of the 83 stop exits closed more than 5 percentage points beyond the 35% stop. The worst trade was -$585.00 and the worst day -$1,172.00 on 2026-09-02; 24 days carried two or more index-ETF bets, and those bets made +$1,053.00. An assumed $1.30 a contract round trip on 597 contracts would take the mean to -$24.30, but that figure subtracts commission only: the bid/ask spread is not deducted, and it could be measured on only 25 of the 179 rows (median 0.02362 of mid). Eight trades with no entry time, the same eight with no holding time and no month, contributed +$1,000.00 and are included in every total; without them the gross mean would be about -$26.75. The in-sample activity frontier put its knee at 0.37 trades a weekday (25 trades, +$1,775.00, maximum drawdown $595.00); it orders history and does not forecast. The published family count is 1, the publisher's minimum for one exploratory record, not a count of the 85 buckets and 71 setups compared. No bucket is a tested edge and nothing here is promoted.
The study had no pre-registration and judged no trial, and the report's own label is exploratory, so the published verdict is descriptive, not pass or fail. The published family count of 1 is the publisher's minimum for one exploratory record, not a count of the 85 buckets and 71 setups compared on the same 179 trades, and no number is deflated for those comparisons. The overall mean's 95% band includes zero.
Backtested by QS Research
Pre-registration
No pre-registration. The study had none; this record was written on 2026-11-21, after every result had been seen, and is labelled retrospective throughout. The public record date of 2026-11-21 is the date this retrospective record was written, not a date fixed before the data. It fixes no rule, threshold or date in advance, and no result counts as a pre-registered test. It records what the committed code did: 179 closed paper round trips over 7 August to 10 November 2026, gross realized P&L, 2,000-draw bootstrap bands, a commission-only cost sensitivity, risk and sizing counts, an in-sample activity frontier, a league table of other signal sources, and in-sample seed rules. No trial was judged, no holdout was reserved and no pass bar existed. The published family count of 1 is the publisher's minimum, not a count of the 85 buckets and 71 setups compared; nothing is deflated against it.
Registered 2026-11-21Trials · 0 trials · 0 passed · judged at N = 1 trials; 2965 trials run in total
| Trial | Verdict | |
|---|---|---|
| katy+qs_strategist | RESULT | not counted |
| qs_strategist | RESULT | not counted |
| katy+qs_strategist+research | RESULT | not counted |
| research | RESULT | not counted |
| FST Trader+qs_strategist | RESULT | not counted |
| guru picks+katy+research | RESULT | not counted |
| FST Trader+guru picks | RESULT | not counted |
| guru picks+katy+qs_strategist | RESULT | not counted |
| none | RESULT | not counted |
| guru picks+qs_strategist+research | RESULT | not counted |
| FST Trader | RESULT | not counted |
| FST Trader+guru picks+qs_strategist | RESULT | not counted |
| katy | RESULT | not counted |
| guru picks+research | RESULT | not counted |
| guru picks+qs_strategist | RESULT | not counted |
| FST Trader+katy | RESULT | not counted |
| katy+research | RESULT | not counted |
| guru picks+katy+qs_strategist+research | RESULT | not counted |
| qs_strategist+research | RESULT | not counted |
| katy | RESULT | not counted |
| qs_strategist | RESULT | not counted |
| FST Trader | RESULT | not counted |
| guru picks | RESULT | not counted |
| research | RESULT | not counted |
| P | RESULT | not counted |
| C | RESULT | not counted |
| single_name | RESULT | not counted |
| sector_etf | RESULT | not counted |
| index_etf | RESULT | not counted |
| SMH | RESULT | not counted |
| SPCX | RESULT | not counted |
| AMZN | RESULT | not counted |
| MDB | RESULT | not counted |
| CRWV | RESULT | not counted |
| QQQ | RESULT | not counted |
| IWM | RESULT | not counted |
| NET | RESULT | not counted |
| CVX | RESULT | not counted |
| WMB | RESULT | not counted |
| META | RESULT | not counted |
| GOOG | RESULT | not counted |
| AVAV | RESULT | not counted |
| SNDK | RESULT | not counted |
| SLV | RESULT | not counted |
| ASTS | RESULT | not counted |
| MRNA | RESULT | not counted |
| INTC | RESULT | not counted |
| JPM | RESULT | not counted |
| GDX | RESULT | not counted |
| QCOM | RESULT | not counted |
| NVDA | RESULT | not counted |
| AAPL | RESULT | not counted |
| SILJ | RESULT | not counted |
| XOM | RESULT | not counted |
| MRVL | RESULT | not counted |
| MU | RESULT | not counted |
| AMD | RESULT | not counted |
| SPY | RESULT | not counted |
| 2-5 | RESULT | not counted |
| 6-14 | RESULT | not counted |
| 0 | RESULT | not counted |
| 1 | RESULT | not counted |
| 09:30-10:00 | RESULT | not counted |
| 10:00-12:00 | RESULT | not counted |
| 15:00-16:00 | RESULT | not counted |
| 12:00-15:00 | RESULT | not counted |
| unknown | RESULT | not counted |
| 15-60m | RESULT | not counted |
| 1-3h | RESULT | not counted |
| overnight+ | RESULT | not counted |
| 3-7h | RESULT | not counted |
| unknown | RESULT | not counted |
| <15m | RESULT | not counted |
| stop_loss | RESULT | not counted |
| end_of_day | RESULT | not counted |
| external_close | RESULT | not counted |
| take_profit | RESULT | not counted |
| 2-5 | RESULT | not counted |
| 5+ | RESULT | not counted |
| <2 | RESULT | not counted |
| 2026-08 | RESULT | not counted |
| 2026-11 | RESULT | not counted |
| 2026-09 | RESULT | not counted |
| unknown | RESULT | not counted |
| 2026-10 | RESULT | not counted |
Limitations
- No pre-registration: the record was written on 2026-11-21, after the results were seen. It is a retrospective account, not a test fixed in advance, and no result here is a pre-registered result.
- One paper account, 179 closed round trips over three months (7 August to 10 November 2026); fills and marks are the paper broker's, not real executions. This describes that book, not the market.
- Gross P&L: no source records fees. The -$24.30 net mean is a stated $1.30-a-contract round-trip assumption, not a measurement. It subtracts commission only: the bid/ask spread is not deducted, and it could be measured on only 25 of the 179 rows (median 0.02362 of mid). The true net is unknown; this figure does not measure it.
- The inputs are not committed and no content hash or snapshot id is recorded, and the start and end dates come from the study's commit message and the later forward study, not a result field, so the run cannot be reproduced byte for byte. The two input locations recorded in the result are scratch paths that no longer exist.
- 85 buckets over 11 tables and 71 setups were compared on the same 179 trades, with no multiple-testing correction. The published family count of 1 is the publisher's minimum, not a trial count; no number is deflated against it. The 95% bands are descriptive, and the overall mean's band includes zero.
- The book is one in-sample history. The activity frontier ranks history and does not forecast; the seed rules are in-sample selection rules, not pass bars.
- Exit-reason and holding-time buckets are outcomes of the exit rule, not causes: stop exits are all losses and take-profit exits are all wins by construction. Reading them as strategy signals would be circular.
- Eight trades carry no entry time; the same eight carry no holding time and no month. They contributed +$1,000.00 and are included in every total, so they are a material part of the result; without them the overall book mean would be about -$26.75, more negative than -$19.97.
- Only 25 rows record a decision-time quote; the other 154 record a mark. Execution cost against the mid cannot be measured for them.
- The book is described as paper, long and single-leg, but the code does not filter on mode, side or leg count; that description is an assertion about the input, not a rule the code enforces. Defined-risk structures have no history here, and the study does not test whether the paper fills could be repeated.
- The published report carries the 11 attribution tables. The activity frontier, the league table of other sources and the seed values are in the study write-up but not in the published report; the league verdicts are fixed text, not data.
- There is no holdout and no pass/fail verdict. Nothing here should be promoted or quoted as an edge.
The numbers: every trial's checks, metrics, curve and yearly results, and the provenance
1 published version · current 8d0c29c824eec4bd, published 2026-11-21