로그인
← Win-rate picks

What is EMA death cross? — chart signal explained

This signal watches how two recency-weighted moving averages (EMAs) — one spanning 20 candles, one spanning 50 — sit relative to each other. Fast line above slow line is generally taken as an up-leaning market; fast below slow, a down-leaning one.

When does it fire? — BaroBara criteria

It fires on the first candle where the 20-candle EMA slips below the 50-candle EMA, after sitting at or above it on the previous candle.

How traders usually read it

With the short-term trend dipping under the long-term one, it's widely read as a warning that the broader direction may be turning down. It's well-known enough that 'death cross' headlines regularly make the financial news.

What to watch out for

It shows up late, only after a sizable decline — so it's not rare for the signal to coincide with a short-term bottom. If the drop ends quickly, a golden cross follows soon after, and that round trip can hurt you on both sides.

What the data actually shows (BTC 1h)

The common reading is a drop — but what actually happened matters more. This signal has fired 213 times on BTC 1h; across the most recent 213, price reached the small target (+0.25%) first about 53% of the time. Widen the target to ±1% and it becomes about 51%. A historical probability, not a guaranteed direction — and it shifts with market regime.

Odds and expected value — with symmetric target and stop (±)

Exactly the barobara framing: which side got hit first, +X% or −X%. Target and stop are set to the same %, and the win rate is how often the upside (+X%) was reached first.

Target = stop (±)Win rate (+ first)EV (before fees)
±0.25%53%+0.02%
±0.5%56%+0.06%
±0.75%54%+0.06%
±1%51%+0.02%
±1.5%46%-0.12%
±2%46%-0.16%
📐 How to read this. EV assumes a symmetric ±X% target and stop: EV = target × (win rate − loss rate), so any win rate above 50% gives a positive EV. Fees are NOT included — they differ by exchange and order type (maker/taker). In reality fees come off the top, and the smaller the target, the bigger the bite fees take (at ±0.25%, even modest fees eat much of the edge). Timeouts (neither side hit within the horizon) are classified by the closing side, and an asymmetric target/stop changes all of these numbers — setting-dependent references, not absolutes.

Broken down by market regime

⚠️ This table uses a different basis than the symmetric (±) table above — a small +0.25% target with a wide −2.5% stop (fees included). The small target makes the win rate look high while EV is often negative — exactly what signal groups hide. And the same signal behaves differently across regimes.

RegimeWin rate (+0.25% target)EV (−2.5% stop)Sample
Bear market88%-0.12%N=58
Sideways84%-0.17%N=68
Bull market91%-0.07%N=79

Recent occurrences

How far price actually moved the last few times this signal fired (MFE = maximum favorable excursion).

DateMFEResult
2026-05-131.11%✅ hit
2026-05-153.25%✅ hit
2026-05-240.0%
2026-05-260.92%✅ hit
2026-05-313.86%✅ hit
2026-06-093.89%✅ hit
2026-06-140.21%— 🔴 stopped
2026-06-170.33%✅ hit 🔴 stopped
🦫 See whether this signal is live right now — and verified odds for other signals and combos — on win-rate picks. Cut trading costs with fee cashback.
Data: full history since 2017 · 21985 bars · ~7.0/month · win rate from the most recent 213 occurrences. For reference, not a prediction. A signal is a historical probability, not a guaranteed direction.
barobara.com · not a signal group — honest signal explainers

A free morning briefing: market heat & chart combos — honest odds, not predictions