Real observationsThrough 2026-10-21
Signal validation
Does strength
keep working?
We test what happens after an asset receives its score—not what happened before. Every displayed result must come from timestamped real market data.
This page has two parts. The historical study replays one year of point-in-time rankings. The prospective section counts only rankings this site saved on their own day. The model accounts, their holding periods and the rebalance rules are explained on the Portfolio page.
Historical research sampleHistorical forward evidence
What happened after first entry?
Each asset is counted once when it first enters the top or bottom 10. “Entries with gain” is the share whose own price rose over the stated horizon; it is not the long/short portfolio win rate.
| Historical cohort | Horizon | Observations | Mean return | Vs. universe | Entries with gain |
|---|---|---|---|---|---|
| First top-10 entry | 7 days | 121 | -0.11% | +1.69% | 44% |
| First top-10 entry | 14 days | 120 | -3.64% | +1.26% | 30% |
| First top-10 entry | 30 days | 118 | -9.37% | +4.76% | 16% |
| First bottom-10 entry | 7 days | 120 | -4.20% | -1.07% | 37% |
| First bottom-10 entry | 14 days | 120 | -8.48% | -1.59% | 28% |
| First bottom-10 entry | 30 days | 119 | -18.10% | -2.37% | 16% |
Historical walk-forward
Point-in-time,
never hindsight.
The retrospective study reconstructs each weekly universe from historical market-cap listings, then uses Binance UTC closes and trailing 30-day dollar volume. Today’s winners are never projected backward.
Published real-data study328 formation dates298 have a complete 30-day outcome · research sample
1-week 20-futures portfolio+49.39%45 periods · after funding2-week 20-futures portfolio+24.98%22 periods · after funding4-week 20-futures portfolio+10.83%10 periods · after fundingDaily rebalance, for reference+8.91%326 periods · after funding · not run as a live account because fees consume the resultWeekly net funding contribution+2.10 ppUncompounded account contribution · long +7.15 pp · short -5.05 pp
All 20 positions are Binance USD-M perpetual futures in one cross-margin account: ten 5% longs and ten 5% shorts, for 50% notional on each side. Signals form at one UTC close and trades enter at the next. If an exact top- or bottom-10 asset has no contract for the full holding period, the next futures-eligible rank is used so the account still holds 20 positions. Weekly exact-rank coverage averaged 98.0% for longs and 95.3% for shorts. Real dataset: 404,928 funding events across 238 mapped assets. Results include funding on both legs but remain before liquidation and slippage. The 1-, 2- and 4-week holding periods were chosen after the daily and weekly results of this same sample were known, so the comparison between them is in-sample. Only the live model accounts test it forward.
Historical signal test
Did leaders continue
to outperform?
D1 is the highest-scoring 10% and D10 the lowest. The average decile spread is D1 return minus D10 return on each formation date; it is a signal diagnostic, not a compounded portfolio return.
| Forward window | Top decile return | Top vs. universe | Bottom decile return | Avg. decile spread | Spread positive | Rank IC |
|---|---|---|---|---|---|---|
| 7 days | -1.05% | +0.86% | -2.36% | +1.11% | 55% | 0.013 |
| 14 days | -2.69% | +1.33% | -5.06% | +2.17% | 55% | 0.021 |
| 30 days | -5.28% | +2.69% | -9.12% | +3.64% | 51% | 0.020 |
Historical weekly portfolio
Twenty futures,
one balanced account.
All positions are Binance USD-M perpetuals in one cross-margin account: 50% of account equity is allocated equally across ten longs and 50% across ten shorts. Every funding payment is applied with the correct direction on both sides.
Live validation
Prospective evidence
is still collecting.
This separate tracker uses only rankings saved by the live app. It validates the historical study out of sample as new 7, 14 and 30-day outcomes mature.
| Saved cohort | Horizon | Observations | Mean return | Vs. universe | Positive rate |
|---|---|---|---|---|---|
| First top-10 entry | 7 days | 76 | +3.04% | +0.64% | 38% |
| First top-10 entry | 14 days | 71 | +7.82% | +0.23% | 54% |
| First top-10 entry | 30 days | 55 | +20.93% | -4.07% | 80% |
| First bottom-10 entry | 7 days | 81 | +2.74% | -0.14% | 51% |
| First bottom-10 entry | 14 days | 71 | +6.73% | -1.47% | 58% |
| First bottom-10 entry | 30 days | 50 | +22.74% | -3.22% | 78% |
All observations come from immutable daily snapshots; missing days are not interpolated.
Cohorts measure the first saved appearance in the top or bottom 10, then 7/14/30-day absolute and universe-relative returns.
The prospective live portfolio module still uses a spot-return short proxy until its saved observations mature; it is separate from the historical funding-adjusted USD-M futures result above.