Historical study
Is the binding constraint spread width or quote staleness?
Consistent with staleness: on seconds data the loss per fill falls from 362 to 70 cents while the spread earned barely moves.
Adverse selection falls from 362 cents per fill to 70, which is 81 percent of the loss gone. Meanwhile the spread earned barely moves, from 148 to 138 cents. This pattern is consistent with stale quotes contributing to losses. Different days, markets and fill composition remain alternative explanations; this comparison does not isolate the cause.
Data · 5,413,998 book snapshots · 468 tokens · 5 days · 2026-07-30 to 2026-08-03
| Adverse selection at 120s | -362 cents per fill |
|---|---|
| Adverse selection on seconds data | -70.3 cents per fill |
| Spread earned at 120s | 148.3 cents per fill |
| Spread earned on seconds data | 138.2 cents per fill |
Next test · planned
How does quote latency change markout on the same events and books when the market universe, fill rules and capital budget stay fixed?
Research renewal planPayload snapshot 2026-10-12. These figures are frozen; the interactive page reads the same file.
Page export dates do not extend the observation period shown above.
Full report (docs/research/mm_pnl_stream-5tage.md) · Source module (src/mm_pnl.py)
Research tool only: no investment advice, no order placement, no venue affiliation. Public Polymarket and Kalshi data, provided as-is.