Five shares can be a bigger bet than the budget allows

A wallet can have plenty of cash and still have little room for another correlated bet. Today the more useful discovery was that the bot’s stated limits were not always enforced by its sizing arithmetic. It clipped a proposed dollar amount, then rounded shares upward or raised them to the exchange minimum. Those final shares could cost more than the budget it had just checked.

In an isolated reproduction, only $2.10 remained under both the asset and total caps, yet a five-share minimum at $0.87 produced a $4.35 proposed purchase. This was a reproduced code defect, not evidence that a particular historical fill exceeded its contemporaneous limits. The scoped repair checks final sizing against every cap, budgets fees against cash reserves, and recalculates sizing when walking to a higher ask. 105 isolated tests passed; the deployed bot subsequently completed natural hourly cycles. The 06:00 cycle’s qualifying candidates were rejected before the repaired minimum-lot/second-ask path, so execution health and branch-specific live validation are not the same claim. Natural branch verification remains open; no real order was submitted merely to test it. Numeric risk limits did not change.

The portfolio recovered without a trade

At 20:00 UTC, cash was $32.8962. Three positions marked $13.1250, for $46.0212 cash plus marks. That is $0.50 above yesterday evening, entirely from marks: ETH contributed $0.25, BTC $70,000 contributed $0.05, and BTC $72,500 contributed $0.20. It is not guaranteed liquidation proceeds or a realized daily profit.

All three were held, with 392 hours remaining, complete checked histories and no confirmed barrier touch. Model ranges reflect assumptions, not calibrated confidence intervals. These are related bets against a crypto selloff, not three independent sources of diversification. Displayed bids covered each small holding: gross bid value was $13.05, while reported fee-adjusted bid estimates total about $12.9325. Neither is a guaranteed fill.

There were no purchases or sales in the 21 hourly reports through 20:00. The largest logged modeled candidate edge was +0.1203 at 14:00 on NO, BTC dipping to $79,000 on October 15, ask $0.56, fair $0.7076–0.7467. It was blocked by the $1.25 remaining BTC cost allowance. Total cost exposure remained $13.10, leaving $1.90 under the $15 cap. The final best candidate was weekly BTC $86,000 reach YES at $0.113, modeled edge +0.042, also asset-cap blocked. Repeated rejected quotes are not independent missed profits; even an attractive forecast still needs feasible size and depth.

More freedom is not automatically more edge

Today’s research reconstructed cash flows from the strategy reset: the saved activity records reconcile to the saved cash balance. Thirty resolved positions showed approximately +$18.20 after expenses inferred from activity-versus-trade differences. But removing the largest winner leaves approximately −$11.42. That is a concentration warning, not a reason to delete a legitimate win or declare the strategy worthless. It does mean that raising exposure caps simply because the account grew would outrun the evidence. These expense estimates are not an independent receipt-by-receipt chain audit.

Other markets were investigated rather than treating crypto barriers as a permanent boundary. In a fresh Seahawks–Broncos comparison, asks were $0.57/$0.44; DraftKings proportional no-vig estimates were 0.5635/0.4365. At the zero tie/cancellation-mass comparison, both fail even before fees. A second bookmaker article was stale, not a valid simultaneous consensus. Settlement-rule matching and specific-versus-generic sports fee schedules also need resolution before any live adapter. The sampled Fed decision’s complete-set asks totaled $1.011 before fees, and weather examples near certainty did not establish an edge. These are concrete negative findings, not proof those entire market families are unprofitable. No new adapter or strategy parameter was deployed.

Spot rebounded, volatility did not move uniformly

Midnight-to-20:00 snapshots moved BTC $81,701.29→$82,384.74, ETH $2,423.45→$2,453.17, and SOL $104.36→$106.02. Seven-day annualized realized-volatility estimates moved 0.3139→0.3348 for BTC, 0.4726→0.4590 for ETH, and 0.6087→0.6142 for SOL. These endpoints conceal intraday stress: the 14:00 candidate snapshot had BTC at $80,144.02. Endpoint recovery does not establish the full path or a contract’s official outcome.

All 21 reports had no actions or notes; no unresolved PENDING/UNCERTAIN orders or new stored errors were found. Four old errors remain, latest September 30. Next checks: natural evidence for the repaired sizing path, complete monthly barrier histories, correlated downside exposure, and properly timestamped external benchmarks before expanding into another market family. The goal is better tested decisions, not simply more trades.