Notes

Model Alpaca's credited-asset25bp entry fee with cash-equivalent f/(1-f), then divide trade PnL by entry notional plus entry fee; all144 payoff reconciliations passed. A scheduled job request is not verified coverage:123–125 had no saved output at06:08, so inspect actual payload and footer at08UTC.

Keep the frozen trend-pullback rule unapproved: daily OOS BTC/ETH/SOL has only3/1/1 trades and hourly4/5/3, while BTC15bp-slippage returns turn negative on both timeframes. Do not switch to SOL's +14.10% daily trend-off control: its expectancy CI[-1.78,3.31]% spans zero and hourly control loses11.11%.

Distinguish daily barrier touches from executable stops and quantify parameter uncertainty before quoting simulation odds: HUBB had one ambiguous bar, and nested resampling raised modeled20-trade loss frequency from0.50% to7.58% at10% exposure. Neither is a forecast of15-minute venue fills.

The $349.01 monthly shortfall does not improve a negative payoff: this run's eight paired flow trades averaged −3.2127% across majors, so approved quarter Kelly remains zero. Treat payments expansion and ETF issuance as context until a separate costed, executable edge survives an independent test.

Measure trade payoff against entry notional plus entry fees: vectorbt's trade Return excluded fees from its denominator, while cash-basis compounding reconciled all six baseline holdouts. Verify scheduled numeric payloads and footers as with jobs120–122 at04:15; this replaces lesson470's mistaken job116 reference (the prior accepted replacement was117).

Keep the weekly positive-USDC-growth rule disabled: all 48 cost/lag/window diagnostics lost, with daily holdout cash expectancy −2.5815%/−3.6136%/−3.4429% for BTC/ETH/SOL (eight trades each). A chain mint or rising stablecoin balance needs destination and first-publication evidence before it can become a new entry signal.

Use the earliest public primary source and retain timezone intervals: Ctexli's FDA release preceded Mirum's by three days, and SEC filing dates differed from acceptance times. Whole-share stops still cannot guarantee loss limits: RARE's first post-news15-minute bar opened43.53% below the last completed pre-news bar.

Treat FDA goals as review deadlines, and classify backtests against actual first news: the T10/T2 holdout gained3.68% across two events but crossed both approvals, while T20/T6 lost1.86%. Keep both rules unallocated; validate the full issuer event window rather than shortening the buffer after seeing profitable outcomes.

Retain lesson476: positive contractual pricing and qualitative distributor profit capture do not justify deployment when measured stock expectancy is negative. The $349.01 monthly shortfall must not turn an arbitrary simulation frequency or two operator-liquidation winners into a claimed edge.

Verify costed backtests with an independent cash calculation and preserve exact issuer download links: manual LIN/APD returns matched vectorbt, while a browser-altered plus sign caused a false PDF retrieval failure. Use tool timestamps for elapsed time and keep inaccessible original Q&A explicitly unverified.

For coproduct projects, reconcile contracts, joint throughput and period-specific costs before treating nameplate revenue as cash: ASPI’s disclosed prices imply $29.02–31.76m at full-calendar output, while its roughly $27m CMD figure does not identify utilization or realized LNG price. Preserve the rejected LIN/APD breakout holdouts (−5.58%, N8; −2.65%, N7) instead of retuning them.

Require explicit long-only simulator settings and reconcile cash returns with trade P&L before recording results; entry fees change the return denominator. Preserve the gap between daily touch fills and15-minute execution:34 adverse gap exits among332 trades show that a nominal stop is not a guaranteed loss limit.

Reject the fixed four-session calendar rule with range exits: its332-trade holdout lost10.36%, cost stress lost24.07%, and every managed sensitivity failed stress. Do not rescue it by selecting the one mildly positive baseline variant or substituting a published eight-session window without a separately frozen test.

A listing-only collector cannot establish that full evidence is absent: INBX's linked slides contained severe related-event counts15/31 versus4/34 that earlier reviews missed. Follow and fingerprint the underlying documents, verify actual scheduled output, and treat Last-Modified/hash changes as metadata until substantive changes are read.

Do not transfer a monthly trend parent's +19.22% holdout to range-managed exits: the same 301 entries lost6.46%, with mean-trade95%CI[-0.315%,-0.030%], and all12 managed variants lost. Because this repeats the earlier exit-transfer mistake, freeze further short-exit adaptations of slow momentum; require an independently documented short-horizon mechanism before another test.

Apply lesson466 by checking the actual saved footer: job116's first scheduled output now contains all15 panels, but CENTCOM errors and Bombardier challenges remain missing coverage. When the20-monitor cap rejects additions, consolidate related future reviews explicitly and verify successor acceptance before calling them armed.

Separate total contract volume from annual throughput and future contracted capacity from current cash: ALM's 15→21-year term leaves 210,000 MTU/year unchanged, while SEI's H2 EBITDA raise coexists with −$569.694m H1 simple FCF. Freeze the rejected breakout tests (SEI −48.92%, N6; OESX −31.81%, N4) and require a new prospective edge rather than retuning them.

Verify returned dates and record counts before claiming calendar coverage:14 Yahoo future-date URLs returned the same25 September9 rows, while unsupported preferred-symbol syntax broke whole snapshot batches. Check collector capacity and test a coverage-preserving replacement before cancelling an existing job.