Which signals actually work
Every situation our scanner marks is recorded, and 24 hours later we check what the price did. The same measurement runs on randomly picked coins at the same moments — that is the control group. Without it a percentage means nothing.
| Situation | Measured | Sample | Fired | Control | Difference | 95% interval | Verdict |
|---|---|---|---|---|---|---|---|
| Price at yesterday's high | implied side | 45 | 13.3% | 42.8% | -29.5 | 6.3–26.2% | below control |
| Steady downtrend | implied side | 32 | 81.3% | 57.2% | +24.1 | 64.7–91.1% | above control |
| Volatility spike | any move | 534 | 62.2% | 41.4% | +20.8 | 58–66.2% | above control |
| OI surge | any move | 631 | 58.8% | 41.4% | +17.4 | 54.9–62.6% | above control |
| Volatility squeeze | any move | 128 | 57% | 41.4% | +15.6 | 48.4–65.3% | above control |
| Retest of a broken zone | implied side | 64 | 65.6% | 51.4% | +14.2 | 53.4–76.1% | above control |
| Overbought | implied side | 119 | 69.7% | 57.2% | +12.5 | 61–77.3% | above control |
| Bands squeezed | any move | 1,157 | 28.9% | 41.4% | -12.5 | 26.3–31.5% | below control |
| Chart pattern: Double bottom (confirmed) | implied side | 247 | 32% | 42.8% | -10.8 | 26.5–38% | below control |
| Active pump | any move | 1,471 | 51.4% | 41.4% | +10 | 48.8–53.9% | above control |
| Divergence by RSI · 4h | implied side | 309 | 37.5% | 47.4% | -9.9 | 32.3–43.1% | below control |
| Overheated · late stage | implied side | 178 | 47.8% | 57.2% | -9.4 | 40.5–55.1% | below control |
| Stretch from the mean | implied side | 854 | 51.8% | 46.8% | +5 | 48.4–55.1% | above control |
| Resistance zone broken | implied side | 681 | 38.5% | 42.8% | -4.3 | 34.9–42.2% | below control |
| Volume spike, no side | any move | 3,393 | 37.8% | 41.4% | -3.6 | 36.1–39.4% | below control |
| False breakout · sweep | implied side | 5,079 | 51.9% | 49.8% | +2.1 | 50.5–53.2% | above control |
| Chart pattern: Symmetrical triangle (confirmed) | implied side | 1 | 0% | 57.2% | -57.2 | 0–79.3% | not enough data |
| Chart pattern: Falling wedge (confirmed) | implied side | 1 | 0% | 42.8% | -42.8 | 0–79.3% | not enough data |
| Steady uptrend | implied side | 10 | 70% | 42.8% | +27.2 | 39.7–89.2% | not enough data |
| Chart pattern: Ascending triangle (confirmed) | implied side | 8 | 62.5% | 42.8% | +19.7 | 30.6–86.3% | not enough data |
| Price at yesterday's low | implied side | 18 | 72.2% | 57.2% | +15 | 49.1–87.5% | not enough data |
| Divergence by RSI · 1h | implied side | 207 | 41.5% | 47.2% | -5.7 | 35–48.4% | indistinguishable |
| Divergence by CVD · 1h | implied side | 312 | 58% | 52.8% | +5.2 | 52.5–63.4% | indistinguishable |
| New range high | implied side | 182 | 37.9% | 42.8% | -4.9 | 31.2–45.1% | indistinguishable |
| Support zone broken | implied side | 741 | 60.6% | 57.2% | +3.4 | 57–64% | indistinguishable |
| Chart pattern: Triple top (confirmed) | implied side | 219 | 60.3% | 57.2% | +3.1 | 53.7–66.5% | indistinguishable |
| Impulse candle | implied side | 282 | 55.3% | 52.3% | +3 | 49.5–61% | indistinguishable |
| Divergence by RSI | implied side | 544 | 46.9% | 49.7% | -2.8 | 42.7–51.1% | indistinguishable |
| Chart pattern: Descending triangle (confirmed) | implied side | 11 | 54.5% | 57.2% | -2.7 | 28–78.7% | not enough data |
| New range low | implied side | 330 | 59.7% | 57.2% | +2.5 | 54.3–64.8% | indistinguishable |
| Range | any move | 317 | 39.1% | 41.4% | -2.3 | 33.9–44.6% | indistinguishable |
| Divergence by CVD | implied side | 1,352 | 55.7% | 53.7% | +2 | 53–58.3% | indistinguishable |
| Price at resistance | implied side | 2,231 | 59.1% | 57.2% | +1.9 | 57.1–61.1% | indistinguishable |
| Chart pattern: Rectangle (confirmed) | any move | 2,543 | 43.3% | 41.4% | +1.9 | 41.3–45.2% | indistinguishable |
| Price crossed EMA50 | implied side | 1,297 | 52.4% | 50.8% | +1.6 | 49.7–55.1% | indistinguishable |
| Funding overheat | implied side | 647 | 50.4% | 48.9% | +1.5 | 46.5–54.2% | indistinguishable |
| Divergence by CVD · 4h | implied side | 352 | 56.8% | 55.3% | +1.5 | 51.6–61.9% | indistinguishable |
| EMA trend | implied side | 2,160 | 54.6% | 53.2% | +1.4 | 52.5–56.7% | indistinguishable |
| Divergence by RSI · 15m | implied side | 188 | 47.9% | 49.3% | -1.4 | 40.8–55% | indistinguishable |
| Chart pattern: Triple bottom (confirmed) | implied side | 181 | 44.2% | 42.8% | +1.4 | 37.2–51.5% | indistinguishable |
| Trapped traders | implied side | 3,923 | 51.7% | 50.4% | +1.3 | 50.2–53.3% | indistinguishable |
| OI change | implied side | 1,021 | 54.6% | 53.4% | +1.2 | 51.5–57.6% | indistinguishable |
| OI flush | any move | 148 | 42.6% | 41.4% | +1.2 | 34.9–50.6% | indistinguishable |
| Oversold | implied side | 115 | 41.7% | 42.8% | -1.1 | 33.1–50.9% | indistinguishable |
| Chart pattern: Double top (confirmed) | implied side | 235 | 56.2% | 57.2% | -1 | 49.8–62.4% | indistinguishable |
| Equal highs — liquidity above | implied side | 2,496 | 43.6% | 42.8% | +0.8 | 41.7–45.6% | indistinguishable |
| Equal lows — liquidity below | implied side | 2,785 | 56.6% | 57.2% | -0.6 | 54.8–58.5% | indistinguishable |
| Price at support | implied side | 2,083 | 42.2% | 42.8% | -0.6 | 40–44.3% | indistinguishable |
| Structure break | implied side | 6,542 | 51.4% | 51.1% | +0.3 | 50.2–52.6% | indistinguishable |
| Divergence by CVD · 15m | implied side | 247 | 50.6% | 50.4% | +0.2 | 44.4–56.8% | indistinguishable |
| Leverage buildup | implied side | 33 | 51.5% | 51.5% | 0 | 35.2–67.5% | indistinguishable |
When signals coincide
A single observation about a coin is weak — and that is true of every indicator, not just ours. The question worth asking is what happens when several different observations land on the same coin at once.
"Different" is checked, not assumed: an RSI divergence, a CVD divergence and an OI divergence are one disagreement seen by three instruments, so they count once. The same goes for a funding spike and leverage build-up — one open-interest number read through two thresholds.
| Signals coinciding | Sample | Price moved | Direction correct |
|---|---|---|---|
| 0 | 488 | 44.5% | 53.4% |
| 1 | 2,419 | 41.5% | 50.7% |
| 2 | 4,180 | 39.1% | 51.8% |
| 3 | 2,336 | 43.5% | 51.8% |
| 4 | 1,030 | 52.2% | 52.1% |
| 5 | 390 | 54.9% | 56.8% |
| 6 | 109 | 59.6% | 53% |
| 7 | 53 | 56.6% | 43.5% |
| 8 | 12 | 50% | 50% |
What the numbers say
Movement becomes more likely the more independent observations agree: from 44.5% on a lone signal to 59.6% when 6 coincide. Direction does not: it stays near a coin flip no matter how many agree. That is why Sonarcast shows you where something is happening and never tells you which way to trade it.
Both columns have their own denominators and are not two views of the same thing: "price moved" is measured on situations that name no side, "direction correct" only on those that do.
Method
A situation counts as resolved when the price moves past a threshold normalised to the coin own volatility (ATR), within 24 hours. The "Measured" column says which of two questions the row answers, because they are not the same number. "Any move" rows have no implied side: the rate is the share of all resolved observations where price crossed the threshold in either direction, against a control of how often any coin moves at all. "Implied side" rows do have one: the rate is the share of DECIDED cases where price crossed the threshold on the side the situation implies BEFORE crossing it against — observations where neither threshold was hit are dropped from both parts of the fraction, and the control is how often the market itself goes that way first. Do not rank a row of one kind against a row of the other. The control group is the same measurement on randomly picked coins at the same moments, so the comparison is against the market, not against zero. The interval is a 95% Wilson interval: when it covers the control rate, the difference is not established — that is what "indistinguishable" means. Types with fewer than 30 resolved observations get no verdict at all.
What this does not say
This is one window in one market regime. A type can behave differently in a trend and in a range, and a single month cannot tell those apart. "Indistinguishable" means no edge was measurable at this sample size, not that the type is useless. Nothing here is a trade recommendation: we publish what happened, not what to do.
Use it
The data on this page is published under CC BY 4.0. Use it, quote it, build on it — the only condition is a credit with a link back to this page. CC BY 4.0
Suggested citation:
Sonarcast, "Which signals actually work", 2026-11-08, https://www.sonarcast.app/en/study
Machine-readable: /api/v1/research/situations-study?locale=en
Updated 2026-11-08