Weekly note: what random entry costs — the paper controls
Week of Nov 2, 2026
Random entry at the recorded best ask: −0.9¢ per paper trade over 10,193 fills, 95% interval −1.7¢ to −0.2¢ — below zero; the week before, −1.3¢ (below zero).
Favourite 30 seconds before the close: −1.2¢ per paper trade over 1,852 fills, winning 93% of them at 94¢ on average — below zero; the week before, −1.3¢ (below zero).
These two rules run next to the published ones every night on purpose: they measure what trading at all costs, so that no rule gets credit for riding the market.
| control | paper fills | per trade | 95% CI | win rate | avg price |
|---|---|---|---|---|---|
| random entry | 10,193 | −0.9¢ | [−1.7¢ ; −0.2¢] | 49% | 50¢ |
| favorite at T−30s | 1,852 | −1.2¢ | [−2.2¢ ; −0.2¢] | 93% | 94¢ |
| control | paper fills | per trade | 95% CI | win rate | avg price |
|---|---|---|---|---|---|
| random entry | 9,995 | −1.3¢ | [−2.0¢ ; −0.5¢] | 49% | 51¢ |
| favorite at T−30s | 1,755 | −1.3¢ | [−2.4¢ ; −0.2¢] | 93% | 95¢ |
Weeks published so far
| week | random entry | favorite at T−30s |
|---|---|---|
| Nov 2–Nov 8 | −0.9¢ [−1.7¢ ; −0.2¢] | −1.2¢ [−2.2¢ ; −0.2¢] |
How it was measured
Paper fills simulated at the recorded best ask, $1 stake, at most one per rule and window, never in the last 20 seconds; weekly 95% interval = mean ± 1.96·σ/√n over the week's fills.
Two disjoint weeks, Monday to Sunday UTC: Oct 26–Nov 1 and Nov 2–Nov 8. Outcomes are Polymarket's official resolutions.
Limits
A week is a small sample: a difference between two weeks smaller than the swing from one week to the next is noise. That is why the previous week is always shown beside the current one.
Windows overlap in time and the four assets move together; the weekly figures are not independent draws.
Data analytics on Polymarket crypto markets · not financial advice · 18+. Figures from our own records of every window; outcomes are Polymarket's official resolutions.