Measured statistics · weight0
Crypto Surge Signal Base Rates
We publish, with real measurements, the share of coins holding each pre-surge signal that reach +30% within 168 hours (7 days), compared against the base rate of Upbit's entire liquid KRW-market universe — not a pre-filtered candidate pool. Signal predictive power is still under verification (weight0); past statistics do not guarantee future results and this is not investment advice.
Full-universe base rate 12.2%
(based on 6,336 matured episodes)
How to read this table — what each column means
- Signal — a pre-surge condition or pattern.
- Sample — how many times this signal has occurred historically (episodes). More is more reliable · under 100 is grayed out (reference only).
- Hit Rate When Signal Present — of the cases where this signal fired, the share that reached +30% within 168 hours. Higher means surges followed this signal more often.
- Base Rate — the same +30% threshold reached by the full universe regardless of signal = the comparison baseline. Must be beaten to matter.
- Lift — hit rate − base rate. +green = observed above the baseline in this sample, −red = observed below the baseline in this sample. A difference observed in this sample — not a claim about future predictive power or profitability.
- 95% CI Lower Bound — the lower end of an interval estimate under the stated model and sample conditions, not a guaranteed minimum or proof of predictive power. This must exceed the base rate to stand apart from chance.
- Significance — whether the difference observed in this sample is unlikely to be chance. ✓Wilson = confidence-interval lower bound exceeds the base rate · ✓BH = still significant after multiple-testing correction · neither (—) means indistinguishable from chance. Not a claim that the same difference will recur, or that it is profitable.
Each signal's sample, hit rate, and base rate appear in a mobile card.
| Signal | Sample | Hit Rate w/ Signal | Base Rate | Lift | 95% CI Lower | Significance |
|---|---|---|---|---|---|---|
| Open-Interest Surge | 92 low sample | 28.3% | 12.2% | +16.1%p | 20.1% | low sample · unresolved |
| 20-Bar High Breakout | 276 | 16.3% | 12.2% | +4.1%p | 12.4% | ✓Wilson ✓BH |
| Relative Strength vs. BTC | 1,747 | 15.6% | 12.2% | +3.5%p | 14.0% | ✓Wilson ✓BH |
| Volume Surge | 523 | 15.5% | 12.2% | +3.3%p | 12.6% | ✓Wilson ✓BH |
| Extreme Funding Rate | 194 | 15.5% | 12.2% | +3.3%p | 11.1% | — |
| Bollinger Band Squeeze | 880 | 11.7% | 12.2% | -0.5%p | 9.7% | — |
low sample = under 100 observations (reference only, grayed out). ✓Wilson = 95% CI lower bound exceeds the base rate · ✓BH = significant after multiple-testing correction.
We do not mark a ✓ below the sample minimum — a confidence interval clearing the base rate on a handful of observations isn't meaningful "significance."
This table has not yet controlled for volatility.
The US-stock base-rate page only marks a signal "significant"
after it clears an ATR-quartile stratification gate, but crypto's sample hasn't grown
large enough to stratify (even the largest current sample is below the minimum), so that
gate isn't applied yet. These figures are therefore pre-stratification
and, as repeatedly confirmed on the stock side, may be inflated by volatility confounding.
We'll republish under the same gate once sample size allows.
Aggregation Methodology
- Denominator = Upbit's entire liquid KRW-market universe at or above a 24h-turnover floor (not just pre-filtered candidates)
- Hit definition = highest price within 168 hours (7 days) reaching +30% versus the entry-point close (based on hourly-bar highs, which can differ from actual fill prices)
- Consecutive signals on the same coin are counted once per episode (72-hour window — removes duplicate observations)
- Coins with missing data (delisting- or halt-like gaps) remain in the sample · Wilson 95% confidence interval + BH (FDR) multiple-comparison correction
- Crypto trades 24/7 · sample sizes differ by signal since collection start dates differ · past statistics do not guarantee future results · not investment advice (weight0 measurement stage)
Get the morning briefing + data alerts —
Follow Telegram @dawnbrief
for US-market close briefings and base-rate/fear-greed data updates.