4. Elite vs crowd
When the best-record wallets disagree with the broad consensus, who is right?
Definition
Elite wallets are the top decile by lifetime record as known at the entry day among tracked wallets with 20+ settled markets, computed from the dataset's own replay (two records: shrunk P/L share and shrunk win rate). The elite side is the side with more elite holders. Variants: when the elite side differs from the consensus side, buy the elite side or buy the crowd side (control), for each record; at signal, whether the elite agree or are absent.
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 |
|---|---|---|---|---|---|---|---|---|
| 4a elite (P/L share) disagrees: buy the elite side strategy, compared with base-hold | better | −17.5 % [−31.4, −3.6] n = 386, hit 42 %, price 0.43 | −1.6 % [−26.0, +25.4] n = 150, hit 33 %, price 0.31 | −13.1 % [−25.4, −0.5] n = 536, hit 39 %, price 0.39 | IS −17.9 % (374) OOS −14.1 % (112) | −8.9 [−27.8, +9.4] points vs base-hold, p = 0.8357 | no | nothing |
| 4b elite (P/L share) disagrees: buy the crowd side (control) strategy, compared with base-hold | control | −7.0 % [−15.7, +2.2] n = 386, hit 58 %, price 0.57 | −2.7 % [−15.8, +13.0] n = 150, hit 67 %, price 0.69 | −5.8 % [−13.2, +2.0] n = 536, hit 61 %, price 0.61 | IS −7.4 % (374) OOS +4.8 % (112) | −3.8 [−10.6, +2.0] points vs base-hold, p = 0.8836 | BH-IS no, OOS positive | nothing |
| 4c elite (win rate) disagrees: buy the elite side strategy, compared with base-hold | better | −5.2 % [−29.3, +21.6] n = 106, hit 54 %, price 0.53 | +6.0 % [−23.9, +38.5] n = 88, hit 43 %, price 0.40 | −0.1 % [−20.7, +22.4] n = 194, hit 49 %, price 0.47 | IS −2.6 % (133) OOS −15.3 % (48) | +17.5 [−11.0, +45.6] points vs base-hold, p = 0.1084 | no | nothing |
| 4d elite (win rate) disagrees: buy the crowd side (control) strategy, compared with base-hold | control | −11.7 % [−34.4, +15.3] n = 106, hit 46 %, price 0.46 | −14.8 % [−32.2, +2.5] n = 88, hit 57 %, price 0.60 | −13.1 % [−27.4, +3.6] n = 194, hit 51 %, price 0.53 | IS −10.7 % (133) OOS −7.6 % (48) | +7.2 [−2.0, +18.3] points vs base-hold, p = 0.0704 | no | nothing |
| 4e elite (P/L share) agrees with the consensus entry filter, decided at the signal | better | −5.4 % [−10.3, −0.5] n = 1,108, hit 64 %, price 0.64 | −3.5 % [−13.7, +7.9] n = 263, hit 66 %, price 0.67 | −5.0 % [−9.5, −0.6] n = 1,371, hit 64 %, price 0.64 | IS −2.5 % (1,048) OOS −11.4 % (217) | −0.8 [−5.7, +4.7] points vs the dropped trades, p = 0.6269 | no | nothing |
| 4f no elite holder in the market entry filter, decided at the signal | worse | −2.0 % [−7.0, +3.4] n = 1,694, hit 65 %, price 0.63 | −4.7 % [−7.1, −1.8] n = 3,395, hit 65 %, price 0.64 | −3.8 % [−6.2, −1.3] n = 5,089, hit 65 %, price 0.63 | IS −2.0 % (2,274) OOS −5.5 % (2,458) | +2.3 [−2.3, +7.0] points vs the dropped trades, p = 0.1673 | no | nothing |
What this data can and cannot show
Can: Whether a strict, point-in-time, dataset-internal record picks the right side when it opposes the head count; the skill study asked a similar question with the site's records and found the superior minority lost.
Cannot: Records before the data starts (the first 20 settled markets must be in the dataset), and the live site's quality.net, which uses a different record.
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.