The One Black Crow is a two-candle pattern taught as a bearish warning: a green day, then a single long red candle that closes at its low and gives the advance back. It fires often, which makes it one of the few candlestick patterns with enough signals to test rather than argue about.
Across 22 US stocks and sector ETFs, over the window the free backtester covers, it fired 619 times. Buying the close of the red candle and holding ten sessions, 313 of those 619 trades finished green: a 50.6% win rate. The median name lost 11.6% over the period, with commission and slippage on, while sitting in the market about 29% of the time.
The number 50.6% has to be read against
Half is not the bar. The window was mostly a bull market, so most ten-day holds finished green whatever day you picked, and that is the figure a bearish signal has to beat.
We measured it on the same 22 names over the same dates. Every ten-session close-to-close hold, every bar:
| Ten-day holds starting from | Windows | Finished green |
|---|---|---|
| any bar | 24,948 | 55.4% |
| any red bar after a green one | 6,034 | 56.0% |
| a One Black Crow | 619 | 50.6% |
The middle row matters as much as the top one. A crow is, crudely, a red day after a green day, and that flip on its own is worth nothing: it lands at 56.0%, marginally above the market's own base rate. So the whole of the pattern's effect comes from what it adds on top of the flip, and what it adds moved the number down by 5.5 points.
On 619 trades that is unlikely to be noise (p is about 0.006). Buying into a black crow was meaningfully worse than buying on a random day.
Which means the pattern is not the coin toss we first called it
An earlier version of this page compared 50.6% to 50% and concluded the crow said nothing. That comparison was wrong, and so was the conclusion. Against the correct baseline the crow is the one pattern in this study that clearly did point somewhere, and it pointed the way its name says.
Three limits on how far to take that:
- It is one window and one universe. Four and a half years, 22 large caps and sector ETFs, in a period that mostly rose.
- Losing to random is not the same as being tradeable. A signal that costs you five points of win rate over ten days is a reason not to buy, which is cheaper and more reliable than a reason to short. Our engine is long only, so we cannot test the short side here at all.
- It is one result among several patterns we ran. Test enough shapes and one will clear a significance bar by luck. This one clears it comfortably, but a single study is a reason to look again, not a reason to stop looking.
Its bullish mirror, and what the pair does not settle
We ran the identical test on One White Soldier, the bullish opposite, on the same names over the same window:
| Signals | Win rate buying it | Its matched baseline | |
|---|---|---|---|
| One White Soldier (bullish) | 595 | 52.6% | 56.0% |
| One Black Crow (bearish) | 619 | 50.6% | 56.0% |
Both sit below their baselines, and the gap between the two patterns is 2.0 points. That gap is not something 600 trades each can resolve: its 95% interval runs from about 3.6 points the wrong way to 7.7 points the right way. So the honest statement is that this test could not separate the pair, not that the pair is identical. It would have missed a real five-point directional edge.
What survives is the column on the right. Neither pattern beat the base rate for simply being invested, and the crow fell furthest below it.
What is a One Black Crow
Two candles:
- The first is long and bullish, extending an advance.
- The second has a long bearish body with little or no wick at either end.
- The second opens inside the first candle's body and closes below the first candle's open.
The story is a session that opened inside yesterday's strength and ended below where that strength began, with sellers in control from the first hour to the last.
Its family is genuinely disputed. Most candlestick writing files it as a bearish reversal; our own signals surface files it as a continuation, a strong directional bar extending the move it is already in. Our data does not choose between them, because both readings expect the same thing over the next ten days: weakness, which is what we found.
Where the eleven percent actually went
The median name lost 11.6% running this rule, and it is worth being exact about the cause, because most of it is not the pattern.
Re-run with commission and slippage set to zero and the same 619 trades lose 3.6%. The rule trades about 28 times per name, and at roughly 0.3% a round trip that friction is around eight of the eleven points. The signal's own contribution is the remaining 3.6%, plus the opportunity cost of the 71% of the time it sat in cash while the same names gained a median 57.3% just by being held.
Which is the practical shape of the problem: even a signal that genuinely points the right way is expensive to trade one candle at a time.
If you use it anyway
None of this is in the backtest, which takes every signal the detector finds. If the pattern earns a second look, it is under these conditions:
- Under a level that has already rejected price, rather than in the middle of a range.
- Against a longer trend that has already turned, not inside an uptrend that is merely pausing.
- On volume above the surrounding days, which is the difference between distribution and a thin session.
- With the market meter unfavourable, because single-stock reversals cluster in risk-off weeks and mean less in them.
Whether that filtering works is a testable question, and testing it beats believing it.
Test it yourself
The backtester carries this pattern. In the free backtester, open Chart setup, add Candlestick pattern, and set it to One black crow. Choose a ticker and a hold period and run it.
With 619 raw signals there is enough sample here to filter and still have something left, which is not true of the rarer three-candle patterns like the evening star. That is the experiment worth running: add one condition, and see whether the win rate moves or only the trade count.
Scanning for them live
Measuring what a pattern did and being alerted when one forms are separate jobs. TC2000 has candlestick formations in its scan library with alerting, no formula required. Finviz has candlestick patterns in its free screener, which is enough to see which names printed one today.
Frequently asked questions
Is the One Black Crow a reliable bearish signal? On our test it was a real one, though a small one. Buying 619 of them across 22 US stocks and holding ten days won 50.6% of the time, against 56.0% for any red day after a green day on the same names and dates. Being 5.5 points worse than a randomly timed hold is the pattern doing what a bearish signal is supposed to do.
Does a 50.6% win rate not mean it is a coin flip? No, and this is the trap. In a rising market the coin is weighted: 55.4% of all ten-day holds in this window finished green. 50.6% is below that, not level with it. Comparing any win rate to 50% without checking the base rate for the period is how a working signal gets written off, or a useless one gets promoted.
What is the difference between One Black Crow and Three Black Crows? Count, and therefore rarity. One Black Crow is a single long red candle after a green one. Three Black Crows is three consecutive long red candles each closing lower, a far more demanding formation that appears much less often.
Is One Black Crow a reversal or a continuation pattern? The sources disagree. Most candlestick writing calls it a bearish reversal; our own signals surface calls it a continuation. Our test does not have to choose, because both readings predict weakness over the next ten days, and weakness is what the numbers show.
What is the opposite of a One Black Crow? One White Soldier: a long green candle after a red one, closing above where the decline began. We tested it identically. It finished at 52.6% against the same 56.0% baseline, which is below the bar but within the noise of a 595-trade sample.
Should I short a One Black Crow? Our engine is long only, so we have not tested that and will not claim it. What the test supports is narrower and more useful: buying into a crow was worse than buying at random by about five points of win rate. That is a reason to wait, and waiting costs nothing.
Before you trade this pattern
The figures here come from one run over 22 large-cap US stocks and sector ETFs, in a window that was mostly a bull market, with the rule invested about 29% of the time. Win rates are measured before costs; the total-return figures include commission and slippage, and are stated both ways above. The script that produced them, including the --baseline flag behind the 55.4% and 56.0% figures, is in the repository (backend/scripts/research/candle_study.py) so the run can be repeated rather than trusted. The tool will show you today's numbers on any ticker you choose, and they will not match these exactly. Past prices are not a forecast, and nothing here is investment advice.