Signal vs Execution & Slate Concentration
full firehose vs top-5 / top-3 / top-1 daily filters
Published: 2026-06-30 | Project: Bet Bodhi | Discipline: Quantitative Engineering & Microstructure
Author: Nicholas Alexander MacAskill — Founder & CTO, Flocano Labs | Canonical: https://www.nicholasmacaskill.com/dossier/bodhi-signal-concentration
Signal vs execution (2025)
| Layer | Definition | WR |
|---|---|---|
| All picks | valueTeam wins the game | 60.0% |
| PASS | Edge identified but no tradable route | 62.7% |
| POLY execution | Positive-EV Polymarket route available | 59.4% |
Handicapping beats tradable routes by 3.3 points. Crowd share prices on Polymarket partially price the same edge the model sees — motivating selective execution and slate concentration rather than firing every pick.
Concentration ladder — full report vs daily filters (2025)
| Slice | Picks | WR |
|---|---|---|
| Full report (all picks) | 1,634 | 60.0% |
| Top 5 / day | 923 | 63.6% |
| Top 3 / day | 577 | 62.4% |
| Top 1 / day (all routes) | 203 | 62.6% |
| Top 1 / day (tradable only) | 153 | 66.0% |
Tighter daily filters extract edge from a 60% pick firehose: +3.6 points from full report to top-5 across 900+ bets. Top-1 tradable adds another +2.4 points over top-1 all-routes — the lift comes from days where Polymarket can actually execute, not from re-ranking the model.
Cross-season stability
| Slice | 2024 WR | 2025 WR |
|---|---|---|
| Full report | 59.5% | 60.0% |
| Top 5 / day | 64.0% | 63.6% |
| Top 3 / day | 64.3% | 62.4% |
| Top 1 / day | 57.1% | 62.6% |
| Top 1 tradable | — | 66.0% |
2024 had no Polymarket historical moneylines (--skip-poly). Concentration lift on top-5/top-3 held both seasons; top-1 alone was weaker in 2024 before tradable-route filtering.
Research thesis
This is decomposition research, not a new ML stack: prove signal at scale, separate execution friction, then show concentration improves realized accuracy. Ranking weights and EV thresholds are production IP — outcomes above are reproducible from the public harness without them.
Limitations
- Grading is binary W/L, not closing-line value
- 2024 season lacks Polymarket execution fidelity
- Historical entry prices are reconstructed where kickoff CLOB was not cached
- Win-rate lifts are robust; ROI figures depend on stake model and are not cited here
