5. Live score + consensus
In play, is smart money on the side that is currently behind a buy?
Definition
PROXY: the datasets carry no scores. Behind = the consensus side's price is 10+ (or 5+) points below its pre-match price while the match is in play (match start and finish from Gamma's gameStartTime and the event's finish; Polymarket's minute price history for the prices; 5-minute decisions). Controls: ahead by 10+, within 5, any in-play entry. Entry at the first in-play moment the baseline gates and the condition hold, held to the result.
Baseline for every comparison: the plain consensus on the same markets (5+ tracked holders, 60 %+ on one side, price 0.05–0.95, first signal, held to resolution), 1 ¢ half-spread plus Polymarket's taker fee; $1 a trade.
Mean return per trade, with 95 % intervals
One marker per window: circle = A (curated wallets, Mar–Apr 2026, clean), square = B (all tracked wallets, Oct 2026, provisional, drift ranges dropped), diamond = pooled. Variants with fewer than 30 trades in a window are left out of the chart; the table below has every number.
Every variant
| Variant | Expected | Window A | Window B | Pooled A + B | IS / OOS (pooled) | vs its baseline (pooled) | Guard | Verdict |
|---|---|---|---|---|---|---|---|---|
| 5a in play, consensus side >= 10 points below pre-match (behind) strategy, compared with s5e-any | better | −10.9 % [−19.1, −2.4] n = 1,362, hit 33 %, price 0.34 | +6.8 % [+0.6, +13.6] n = 1,796, hit 50 %, price 0.45 | −0.9 % [−5.8, +4.9] n = 3,158, hit 43 %, price 0.40 | IS −3.6 % (1,740) OOS +2.7 % (1,351) | +19.7 [+16.7, +22.6] points vs s5e-any, p = 0.0005 | BH-IS no, OOS positive | nothing |
| 5b in play, >= 5 points below pre-match (behind) strategy, compared with s5e-any | better | −8.6 % [−16.0, −1.4] n = 1,620, hit 38 %, price 0.39 | +7.5 % [+1.9, +12.7] n = 2,238, hit 54 %, price 0.49 | +0.7 % [−3.7, +5.1] n = 3,858, hit 48 %, price 0.45 | IS −0.6 % (2,103) OOS +2.3 % (1,667) | +15.4 [+13.3, +17.8] points vs s5e-any, p = 0.0005 | BH-IS no, OOS positive | nothing |
| 5c in play, >= 10 points above pre-match (ahead, control) strategy, compared with s5e-any | control | −0.4 % [−3.4, +2.5] n = 2,113, hit 75 %, price 0.74 | −1.3 % [−3.9, +1.5] n = 3,113, hit 74 %, price 0.73 | −0.9 % [−2.8, +1.1] n = 5,226, hit 74 %, price 0.73 | IS +0.4 % (2,823) OOS −2.3 % (2,264) | −15.7 [−17.5, −14.0] points vs s5e-any, p = 0.9995 | no | negative |
| 5d in play, within 5 points of pre-match (control) strategy, compared with s5e-any | control | −3.3 % [−7.7, +0.5] n = 2,255, hit 56 %, price 0.56 | +1.5 % [−2.2, +4.9] n = 3,012, hit 64 %, price 0.61 | −0.6 % [−3.2, +2.2] n = 5,267, hit 60 %, price 0.59 | IS −0.4 % (2,823) OOS −1.0 % (2,316) | −1.5 [−2.3, −0.6] points vs s5e-any, p = 0.9995 | no | negative |
| 5e any in-play entry (control) strategy, compared with base-hold | control | −5.5 % [−9.1, −1.9] n = 3,207, hit 58 %, price 0.58 | +3.6 % [+0.6, +6.6] n = 4,642, hit 65 %, price 0.62 | −0.1 % [−2.4, +2.2] n = 7,849, hit 62 %, price 0.61 | IS −0.4 % (4,207) OOS +0.1 % (3,460) | +2.2 [−0.9, +5.2] points vs base-hold, p = 0.0894 | BH-IS no, OOS positive | nothing |
What this data can and cannot show
Can: Whether a consensus that holds through an in-play price fall is informed or stubborn. The price proxy tracks the score closely in two-outcome matches.
Cannot: The score itself (a 1-0 lead with 5 minutes left and a 1-0 lead at half-time price very differently; the proxy sees the price, not the clock), matches Polymarket's live feed did not cover, and markets without a minute price series. The provisional October times are the biggest caveat: an in-play entry stamped 3 h early is look-ahead.
Caveats that apply to every signal
- Only tracked wallets' fills exist, from when the site started tracking them; the replay is closer to the chain than to what the site displayed (no Polymarket position overrides, merges missing May–Oct 2026).
- Window B's fill times run early in parts of October 2026; the affected ranges are dropped and the rest is not proven clean. Its numbers are provisional until the repaired export is in.
- Prices are the last tracked fill, not the order book (median error 0.5 ¢, p90 up to 16 ¢ on fast markets); entries happen on the hour (5-minute steps in play).
- The windows are 7 and 3 weeks long; effects of a few points cannot pass a guard over 40 variants there. A failed guard is not evidence of no effect — it is the absence of evidence.
- Nothing here is trading advice. Past replays, even clean ones, do not pay future trades.