An eight-cent ticket and a forty-cent model
At 19:00 UTC, YES on Bitcoin reaching $86,000 during September 21–27 had an eight-cent ask. The bot’s model assigned a 39.31–42.80% probability. After estimated fees and the extra safety deduction, the calculated edge was 29.8 cents per share—the highest in today’s retained snapshots.
That gap is striking. It is not evidence of arbitrage, and it is not a realized return. A model can disagree with a market because the market is wrong, because the model is wrong, or because the inputs and contract interpretation do not line up. The stored numbers alone do not settle which explanation applies.
The order that was never sent
There were no trades through 20:00 UTC and no September 25 order-journal entries. The strongest signal was skipped with cap:asset=0.80: Bitcoin purchase-cost exposure stood at $9.20 against its $10 cap. Total open purchase cost was $13.50, leaving $1.50 under the $15 portfolio ceiling. Both remaining budgets were below the configured $2 minimum sizing budget.
The 19:00 book showed 800 shares at the ask, but the candidate stopped at the sizing cap before the later execution checks. It would be misleading to call this a verified executable profit forgone.
The same $86,000 weekly contract led the hourly reports from 01:00 onward. Its computed edge rose from roughly 9.7 cents to 29.8 cents as the ask fell. Raising a limit would change whether the bot could buy it; it would not establish whether the probability estimate deserved trust.
Existing tickets tell the uncomfortable half
The cash balance remained 36.8233 USDC, but unchanged cash did not mean unchanged wealth. The portfolio’s displayed value fell to $5.7625. Cash plus marks was $42.5858, down $3.42 from yesterday’s 20:00 snapshot. At current bids, cash plus positions was $42.3853 before exit fees, down $3.4480 over the same interval.
These are unrealized valuation changes, not today’s realized trading P&L. Against $13.50 of open purchase consideration, the displayed unrealized loss is approximately $7.74, excluding additional fees.
| Position | Shares | Entry average | Model fair range | Bid | Displayed value |
|---|---|---|---|---|---|
| BTC reaches $88k, Sep 21–27 — YES | 13 | 0.1899 | 0.0939–0.1186 | 0.004 | $0.0975 |
| BTC dips to $82k, Sep 21–27 — YES | 7 | 0.3399 | 0.3815–0.4168 | 0.070 | $0.5950 |
| BTC reaches $90k, September — NO | 5 | 0.8700 | 0.8749–0.9015 | 0.912 | $4.5725 |
| ETH dips to $2,600, Sep 21–27 — YES | 5 | 0.3900 | 0.2880–0.3488 | 0.070 | $0.3750 |
| ETH dips to $2,300, September — YES | 5 | 0.4700 | 0.0018–0.0057 | 0.022 | $0.1225 |
All five positions remain hold under the configured exit rule. The BTC $90k NO ticket is the only one showing an unrealized gain. The three weekly tickets have 56 hours remaining; the two monthly contracts have 128 hours.
The weekly BTC $88k YES ticket illustrates the disagreement especially clearly: its displayed mark is 0.75 cents, versus a model range of 9.39–11.86 cents. The model is still more optimistic than the market even after substantially reducing its estimate. That is a reason to scrutinize calibration—not to count the difference as money already earned.
Market inputs, not a news narrative
From 00:00 to 20:00 UTC on September 25, the stored inputs changed as follows:
- BTCUSDT: spot $84,410.25 → $84,020.01; seven-day annualized realized volatility 41.97% → 37.09%.
- ETHUSDT: spot $2,688.08 → $2,694.07; seven-day annualized realized volatility 45.13% → 41.83%.
- SOLUSDT: spot $117.05 → $121.98; seven-day annualized realized volatility 60.90% → 54.58%.
Bitcoin moved lower, Ethereum ended slightly higher, and Solana rose more substantially across those endpoints. Seven-day realized volatility fell for all three. For untouched barriers, less time and changing volatility alter the remaining chance of a hit; spot direction alone does not explain a ticket’s price.
Next check
All 21 hourly snapshots from 00:00 through 20:00 are present. No new recorded errors, PENDING orders or UNCERTAIN submissions were found. The two errors shown by status are retained historical failures from September 19 and September 23.
The next scheduled scan is September 25 at 21:00 UTC. The reported weekly deadline is September 28 at 04:00 UTC, with monthly contracts at October 1 at 04:00 UTC. No parameters changed in this review. Before treating today’s large gap as grounds for more exposure, the unresolved question is whether the probability model and the exact contract horizon describe the same event. This review documents the discrepancy; it does not claim to have validated the apparent edge.