Evening Star Pattern: The Exact Rule, and Zero Daily Detections
An evening star is the three-bar top: buyers push out a long green session, the market then hangs above it in one small quiet bar, and sellers take the next session with a long red body that closes back through the middle of the original green one.
That is the shape every course teaches. Here is the number none of them publish. Running Vike's batch_evening_star across 479,749 daily candles from 376 Binance pairs produced no detections whatsoever — not one — and the reason is a single clause about where the little bar is allowed to trade. This page gives you that rule, the measurement behind the zero, and real detections found on four-hour candles instead.
What an evening star is actually saying
Demand spends itself, then hesitates in plain sight. The green bar is buyers in control; the small bar above it is a full session where nobody could push price anywhere, held at elevated prices with no follow-through; the red bar is supply taking the level back.
Weight falls on the middle bar. A market that gaps up and then does nothing for a session is a market whose buyers have run out of urgency while still holding the ball, and that is a materially different picture from a violent reversal candle.
Set it against a bearish engulfing, where one red body simply swallows the green one and the handover happens inside a session. An evening star inserts a pause first, and traders read that pause as distribution rather than as a fight.
How to identify an evening star: the exact rules
batch_evening_star measures. The shaded band is the first bar’s body; the upper dashed line is the level the star’s low must stay above, and the lower one is the body midpoint the third close must break. The star is grey because its colour is never read.Six conditions, evaluated in strict order, using only three of the four price columns.
- Bar one is green, and its body is at least the size of the 10-bar average body.
- Bar two is small.
- Its body comes in under 0.3 times that same average, and the crate is entirely indifferent to whether the little bar is green, red or flat.
- Its low has to stay above where bar one closed — the whole range, not just the body.
- Bar three is red, with a body of average size or better.
- That third close must break the midpoint between bar one's open and its own close.
How Vike computes it.batch_evening_starwalks three bars at a time and stamps the last one-100when every condition holds. The 10-bar average it sizes bodies against includes those three candles, so a big rally raises the bar its own reversal must clear.
Notice which midpoint that is. Averaging an open against a close gives the centre of a body, so however long the shadows on bar one run, they cannot shift the level bar three has to break — a detail most descriptions fumble by saying "halfway down the candle".
The clause that produced a zero
Condition three is the whole story, so it is worth stating precisely: the small bar's low — not its body, its low — must be strictly above the first bar's close.
On daily candles the funnel runs like this. Long green first bar: 89,844 cases. Add a small-bodied bar after it and 17,190 remain. Of those, the number keeping the small bar's entire range above that closing price was seven. Two of the seven were followed by a long red bar. Neither closed below the midpoint, so the detector emitted nothing at all.
That is structural rather than statistical. A venue that never closes reopens at the previous close roughly 37% of the time and within a tick of it otherwise, so an unfilled gap is not something crypto produces daily — it is something crypto produces by accident. On four-hour bars the same clause left 136 survivors out of 58,911, and the median qualifying gap was 0.070%. Seven basis points is what "gaps clear above" amounts to in practice.
Why this page uses four-hour candles
I would rather change timeframe in public than quietly present a pattern that the data says does not exist.
Shorter bars give a gap less time to be filled, which is the only thing that makes this rule reachable. Across 1,495,182 four-hour candles from the same 376 pairs, batch_evening_star fired 17 times — roughly one bar in ninety thousand — and every chart on this page comes from that scan. Nothing in the rule was relaxed; the detector is the crate's, unchanged, and the --daily switch on the fetch script reproduces the zero.
Treat that as the honest headline about this pattern rather than as a caveat buried at the bottom. Anyone showing you a daily crypto evening star is either using a looser definition than the engine does, or drawing it.
What the rule does not say
Star colour goes untested. A green star and a red star are equally acceptable, since only its body size and its position enter the arithmetic — which is why a shooting star can sit in that slot and so can an ordinary small green bar.
No gap is required on the third bar, though nearly every illustration draws one. Sellers may open exactly at the star's close and the pattern still qualifies.
Neither does anything look backwards. No uptrend is required, no resistance level, no volume confirmation. Judging whether there was a rally worth ending remains entirely your problem.
Finally, the function loads the open, low and close columns and never touches the high, so any wick standing above the star is invisible to it — the exact opposite blind spot to the one its bullish mirror carries.
Evening star vs morning star
These are two functions in the crate, not one signed function, and they mirror each other clause for clause. Green becomes red, the gap clause changes direction, "closes below the midpoint" becomes "closes above" it.
What the mirror also swaps is which price column gets loaded. batch_evening_star reads the low; batch_morning_star reads the high. Each function is deliberately blind on one side of the star, and that fact is only visible in the source.
Symmetry ends at the code, though. On identical daily tape 7 candidates cleared the evening star's gap clause against 30 for the morning star, and on four-hour bars it was 136 against 187. Crypto gaps down far more readily than it gaps up, so the bearish member of this pair is the rarer of two already-rare things.
How to trade an evening star
Given how seldom it appears, the practical question is what to do on the rare occasion it does.
Do not wait for a textbook trigger. A close below the pattern low arrived on the very next bar exactly once in 17 detections. Insisting on it means never taking the trade, so the entry is realistically the third close or nothing.
Stop above the star. The highest print across the three bars is where the idea is wrong, and since the star is a small bar sitting at the top, that distance is usually modest. This is the pattern's one real structural gift.
Weigh the first bar. A very large green bar pushes the midpoint far below, so the third bar has to travel further and the signal arrives later and stronger. I've been happier with the compact versions.
Supply your own trend. The code checks for none, so an evening star sitting inside a range means nothing at all, and the market has reminded me of that within a day or two on every occasion I have decided otherwise.
Where evening stars fail
GIGGLE on 2026-01-22 is the failure shown here. A green first bar 1.25 times the average body, a star with a body of 0.03 — near enough a perfect doji — a gap of 0.019%, and a red third bar closing 0.62% below the midpoint. Clean on every clause, and price rallied 3.6% instead.
Warning signs were all in the margins. Nineteen thousandths of a percent was the air above bar one; the recovery bar three broke into came to six tenths. My first year of trading was spent assuming a named pattern was a strong pattern, and detections like this one are what eventually broke that habit.
Worse, the full record is not kind. Sixteen of the 17 detections were old enough to grade, and only four followed through by 2% or more while twelve went the other way.
Wider context matters more here than in almost any other candlestick shape, precisely because the rule is so mechanical. Thin books manufacture qualifying gaps, low-volume weekends manufacture qualifying stars, and neither of those has anything to do with distribution by real sellers.
Frequently asked questions
Does an evening star need a doji in the middle?
No. Any body under 0.3 times the 10-bar average passes, in either colour. A separate detector, batch_evening_doji_star, tightens that to 0.1 times the average and leaves the rest alone, so every evening doji star is also an evening star.
Why did it never fire on daily candles? Because bar two's low has to hold above the previous close for a full session. Crypto trades continuously, so the next bar normally opens exactly at that close, and a gap which is never filled all day is close to an accident. Seven daily candidates cleared it out of 17,190; none completed the pattern.
Is an evening star better than a bearish engulfing? It is rarer and its stop is usually tighter, which is not the same as better. An engulfing pair appears often enough to study properly; this one, on crypto tape, mostly does not.
Does the third candle have to gap down? Not in this implementation. Many textbooks draw the gap, but the code only asks for a red bar of at least average size that closes past the first body's midpoint.
What should I use instead on daily charts? Something the engine actually finds. Three black crows fired 79 times over the same daily tape, and a bearish engulfing thousands of times, so both give you a real sample to reason about.
This is educational material, not financial advice. Candlestick patterns describe probabilities, not certainties, and every real trade carries real risk — size positions so a wrong read costs little.