feat(rules): the promotion gate prices fills per product, and the restatement that fires it (#335)

[eaitbrahim]

Closes #335. It waits for "a thin-asset candidate to reach the gate" — measuring the universe answers a sharper question than the trigger asked.

Every document in this repo prices fills at a floor nobody reaches

slippage_for_quote_volume reaches its 5bp floor only at the model's $500M/day anchor. Over the 24-asset universe's own cached candles:

assets at the floor	0 of 24
cheapest	BTC-USD, 5.5bp — 1.1×
dearest	TON-USD, 183.8bp — 36.8× (the cap)
median	≈ 10×

#335 names "the STX/CRO-class 1.15–1.30× floor entries" as the live example of a thin-asset candidate. The live example is the entire universe.

What it costs

240 trials — five rules at shipped defaults × 24 assets × both regimes, run in one driver so the A/B is internally consistent rather than compared across documents.

rule	flat	per-product	delta
turtle_breakout	0.336	0.267	−0.063
rsi_meanrev	0.261	0.175	−0.082
pullback_continuation	0.042	0.012	−0.021
cusum_event	0.343	0.243	−0.080
triple_barrier	0.338	0.237	−0.092

Median across 120 cells: 0.309 → 0.219. Every one of the 120 deltas is ≤ 0.

The comparison that matters

That 0.090 median overstatement sits next to the other number measured this week: the triple barrier's better exit bought +0.033 of gross profit factor.

The error in the cost model was 2.7× larger than the best genuine improvement any rule change produced. Every strategy comparison in this repository has been made through a lens that mis-priced execution by more than the differences being compared.

The corpus's only positive cell dies

turtle_breakout on WLD-USD: 1.061 → 0.626, at 24.2× the rate it was priced at, already below the trade floor at n=58. Zero of 120 clear PF 1.0 per-product.

Why the deferral was safe, and why it isn't now

#259 deferred on the reasoning that the correction is conservative-only — real cost is higher, a corrected profit factor can only fall, and per-product pricing can never manufacture an edge. Confirmed: all 120 deltas are ≤ 0.

What it doesn't survive is the magnitude. A correction assumed to be a rounding adjustment is worth 0.090 of median PF and kills the only positive cell in the corpus. A gate pricing promotion decisions at the best rate the model can produce is not being conservative.

The change

rules backtest / rules promote now price per product, from the product's cached ONE_DAY bars by the same one definition simulate.slippage_assumptions uses — deliberately not the rule's own trading granularity, because median_daily_quote_volume is a per-bar median despite its name, and an hourly figure handed to a daily-anchored model reports every asset as maximally thin. (triple_barrier.per_product_round_trip can't read daily bars — a pure rule has only what it's handed — and scales explicitly instead. Both say so where they do it.)

No daily bars falls back to the flat floor and is flagged as a fallback, never presented as a measured verdict. ResolvedBacktest carries both, so a front-end can print the assumption beside the number it produced.

Verification

7 mutants, 7 killed — two were the call sites. Removing slippage_pct= from _backtest_rule or from backtest_resolved left every helper test green, because the helper being right proves nothing about anything using it.

Full suite 5247 passed / 3 skipped; ruff and mypy clean. Ledger verifies at 93 rows.

Deliberately not done

The older documents keep their flat-priced figures. Their verdicts don't move — the correction only pushes them further from 1.0 — and re-running them isn't free. A pointer note is worth more than the compute.

The null is unchanged and firmer: 0 of 138, now at honest cost.