Data Analysis · 9 min read · June 2026

The Honest Truth About Crypto Signal Win Rates

Open any crypto Telegram channel and you'll see screenshots of "85% win rate" boasts. Open the actual ledger and… you can't, because there isn't one.

This article does the opposite. It opens the ledger.

The data below comes from the SultraxAI platform, which logs every BUY/SELL signal it fires across crypto pairs and back-checks each one at 1h, 4h, and 24h. Three months of running has accumulated enough signals to talk about with some confidence — particularly on BTC-USD, where we have 672 resolved signals.

The headline number: 54.8% on BTC across 672 signals

Symbol Resolved signals Win rate @ 1h Win rate @ 4h Win rate @ 24h
BTC-USD 672 54.8% 53.1% 51.2%
ETH-USD 418 52.1% 51.5% 50.6%
SOL-USD 247 51.4% 50.8% 49.9%
XRP-USD 164 50.6% 50.0% 48.4%

Is 54.8% actually good?

It depends what you compare it to.

Versus a coin flip: meaningfully better. A 54.8% win rate across 672 trials is statistically distinguishable from random chance. The standard deviation of a 50/50 binomial with 672 trials is sqrt(672 × 0.25) = 12.96 wins, so 54.8% (368 wins) is about 2.5 standard deviations above 50% (336 wins). The p-value is around 1%. In plain English: there's roughly a 1-in-100 chance this is pure luck.

Versus a paid Telegram channel claiming 87%: their number is almost certainly cherry-picked, survivorship-biased, or measured at a horizon they didn't disclose. So "better than them" isn't really a comparison — they're not measuring the same thing.

Versus your own discretionary trading: probably similar. Studies on retail crypto traders consistently put discretionary win rates in the 45–55% range. The systematic signal isn't a magic edge over you; it's a tireless, unemotional version of you.

The catch: directional asymmetry

Win rate alone doesn't tell you whether the system is profitable. You need to also know the size of wins versus losses.

For BTC at the 1-hour horizon, here are the numbers:

Metric Value
Win rate 54.8%
Average win (when correct) +0.89%
Average loss (when wrong) -0.72%
Expectancy per signal (before fees) +0.16%
Standard deviation of directed returns 1.34%

That +0.16% per signal expectancy looks tiny. But across 672 signals over ~3 months, it compounds: roughly +47% in raw return if you traded every signal at constant size and ignored fees.

Fees and slippage matter enormously at this scale. On a typical retail exchange (0.1% maker / taker), each signal costs you 0.2% round-trip. That eats most of the +0.16% edge. The system is roughly breakeven net of fees unless you're trading on a low-fee tier or with a futures account.

The honest framing: a small but real statistical edge that gets erased by retail-tier fees. To make this profitable in practice, you need either lower fees (futures, market-maker rebates) or signal selectivity (only trading the strongest signals — see below).

Signal strength stratification

SultraxAI tags each signal as weak, medium, or strong at the moment it fires, based on the conviction of the underlying conditions (volume, momentum confluence, technical level breaks). Stratifying the BTC dataset by signal strength tells a much more interesting story:

Strength Signals Win rate @ 1h Avg directed return
Weak 361 51.8% +0.04%
Medium 217 55.8% +0.21%
Strong 94 62.8% +0.61%

The "strong" cohort is where the real edge lives. 62.8% win rate with +0.61% average directed return per signal is a meaningful edge even after retail fees.

The practical lesson: in any signal system, you should probably ignore the weak signals and concentrate sizing on the strong ones. The signal-strength tagging exists for exactly this reason.

Why no one publishes numbers like this

Three reasons:

  1. Marketing. "54.8%" doesn't sell newsletters. "Last week's pick returned 11.4%" does, even though it tells you nothing about the system's actual edge.
  2. It's actually hard. Logging every signal and back-checking at fixed horizons requires real infrastructure — a database, a job runner, a sync to live price feeds. Most "signal services" are just a Discord bot with a publish button.
  3. It exposes the methodology to scrutiny. Once the numbers are public, smart traders ask questions: what's the holding period assumption? Were signals filtered after the fact? How is "win" defined? Public numbers invite public testing.

SultraxAI publishes them anyway. The bet is that the audience of traders who specifically value this kind of transparency is small but valuable — the kind of people who'll actually pay for the Pro tier because they've already seen the math and aren't going to churn six months in.

What this means for you

See live signal data on SultraxAI

Related