The measured-cost table
In plain words: what trading has actually cost, measured from each fill’s own quote receipt — never estimated, never modeled after the fact. Costs are the quiet killer of most trading ideas, which is why this table is a standing exhibit rather than a footnote.
| book | fills | notional (USD) | paid (USD) | avg bps | cap bypasses |
|---|---|---|---|---|---|
| prob-fartcoin | 1 | 80.00 | -0.01 | -1.6 | 0 |
| prob-jlp | 1 | 80.00 | -0.01 | -1 | 0 |
| prob-jup | 3 | 244.57 | 0.10 | 4.1 | 0 |
| prob-pengu | 1 | 80.00 | -0.33 | -41 | 0 |
| prob-ray | 3 | 233.66 | 0.16 | 6.7 | 0 |
| rails | 1 | 87.59 | -0.02 | -2.2 | 0 |
| s0-rebal | 5 | 387.33 | -0.01 | -0.3 | 0 |
| s3-lite | 4 | 247.10 | 0.00 | 0.1 | 0 |
| total | 19 | 1,440.25 | -0.12 | -0.8 | 0 |
Measured from each FILL’s own quote receipt: both legs valued in USD at the fill’s mark, the gap being the quoted LP fee plus price impact actually paid. A negative total means the quoted route beat the mark used to value the legs — price improvement, recorded as-is. Paper mode charges no router or priority fee, and says so on every receipt (§5). Costs are facts, not verdicts — the referee already nets them.
Compiled from the ledger’s fill receipts by the press; re-rendered at boundaries and closes, never live.