Search for all candlestick patterns and you get the same canon every time: about 40 named shapes, each with a win rate attached. Here is the part the lists leave out: every reliability number we could trace back to a stated test was measured on stocks, an equity fund or equity-index futures. Not one was measured on a currency pair, so if you trade forex and borrow those numbers, you are borrowing a measurement from a different market.
This post audits where the published figures came from, and it names the three structural reasons a currency chart breaks them. Then it tells you which patterns survive the move to forex, and how to measure the rest yourself before you trust them.
Where the numbers behind all candlestick patterns actually come from
Our team took the three studies that sit on the first page of results for this topic and read each one for its own stated method. We compared instrument, bar size, date range and sample size. We did not re-run any of the tests, and the table below only quotes what each author says they did.
| Source | Instrument | Bar size | Date range | Sample and headline figure | Currency pair tested? |
|---|---|---|---|---|---|
| Quantified Strategies | One equity exchange-traded fund | Daily | 1 Feb 1993 onward | 75 patterns screened; 515 trades, 74% win rate, 2.40 profit factor, 8.3% annual return, 13% max drawdown | No |
| TradesViz | Equity-index futures, plus one large-cap stock | 15-minute (futures), daily (stock) | May to Oct 2025 | Bearish engulfing 75.76% at 2.73 profit factor; hammer 71.79%; three white soldiers 83.33%; morning star on the stock 51.85% at 0.79 profit factor | No |
| The Ledger Mind | Not disclosed | Not disclosed | Not disclosed | 59% overall success; 69% at the London and New York opens against 54% mid-session | Claims a forex section, but gives no sample, instrument list or dates |
Read across the last column and the pattern is plain: the two studies that disclose their method are pure equity work. The one source that talks about currencies gives no method at all, so its 69% and 54% can't be checked or repeated.
The Quantified Strategies write-up even flags the problem itself, saying stocks mean-revert more reliably than currencies. That's an honest caveat and the reason its 74% win rate shouldn't be pasted onto a EUR/USD chart. A pattern that works because of mean reversion will lose its edge in a market that reverts less.
TradesViz adds a second issue: according to its own write-up, a volume filter improved results substantially. Hold that thought, because volume is the first thing a spot currency chart can't give you.
None of this says the studies are wrong, since on their own instruments they're well specified, and the Quantified Strategies test runs across more than 30 years. The point is narrower: a win rate is a property of a pattern plus a market plus a data feed. Change any one of the three and you have a new question to answer.
Break one: spot forex has no consolidated volume
Stocks and futures trade on exchanges, where every trade prints to a central tape, so "volume" means real contracts or shares that changed hands. Spot currency doesn't work that way, because it trades over the counter across thousands of dealers, banks and platforms, and there is no single tape that adds it all up.
Data from the Bank for International Settlements 2025 Triennial Survey puts global FX turnover at $9.6 trillion per day in April 2025, up 28% from 2022. Spot made up about 31% of it, roughly $3 trillion a day. The survey needed central banks in dozens of countries to collect that figure once every 3 years. Your charting platform has no way to see it bar by bar.
So what does the volume histogram under your EUR/USD chart show? Almost always it's tick volume: a count of how many times your broker's price updated during the bar. Our glossary entry on tick data vs bar data covers the mechanics. In short, 400 ticks in a 15-minute bar tells you the quote moved 400 times on one feed. It says nothing about whether $5 million or $5 billion traded.
Tick count does tend to rise when the market is busy, so it isn't useless, but it isn't the input the equity studies used. Any rule that reads "pattern plus above-average volume" changes meaning the moment you swap shares traded for quote updates. That includes the filter TradesViz credits for much of its improvement.
There's a quieter problem too, because two brokers with different liquidity providers will show different tick counts on the same bar. A volume filter on one platform may pass a signal that the same filter on another platform rejects. For more on what each field in a bar really holds, see our note on OHLCV data. The V in OHLCV is the least portable letter on a forex chart.
Break two: the forex daily candle has no true open
An exchange closes at night, and while it's shut, news lands, orders pile up, and the next morning the first trade can print well away from yesterday's close. That jump is a gap, and several famous patterns are built on it.
A currency pair trades around the clock from Sunday evening to Friday evening, New York time. The Foreign Exchange Committee sponsored by the Federal Reserve Bank of New York describes the FX market as operating 24 hours a day. When one daily candle ends, the next begins at the very next quote, so the open of Tuesday's candle is, to within a pip or two, the close of Monday's.
Now look at which classic definitions require a gap:
- Morning star and evening star. The textbook version needs the small middle candle to gap away from the first body.
- Abandoned baby. Needs gaps on both sides of a doji.
- Island reversal. Defined entirely by two gaps.
- Kicker. Needs the second candle to open beyond the first one's open.
- Piercing line and dark cloud cover. The strict form needs the second candle to open beyond the first one's low or high.
On a weekday daily chart of a major pair, those gap conditions almost never fire. We haven't run our own count of how often, so we won't put a number on it, and the structural argument doesn't need one. If the defining condition can't occur, the pattern either never appears or the chart software quietly relaxes the rule. Both outcomes wreck any comparison with an equity win rate: in the first case your sample is near zero, and in the second you're testing a different pattern under the same name.
Most traders meet the relaxed version without knowing it. A "morning star" scanner on forex usually accepts a small middle body with no gap. That's a reasonable shape to study, but it isn't the shape the morning star statistics were measured on. Note the TradesViz figure here as well: even on a stock, where real gaps happen, the morning star came in at 51.85% with a 0.79 profit factor, which means it lost money in that sample.
Break three: where the daily candle starts is a vendor choice
On an exchange, the day starts at the opening bell and ends at the closing bell, so everyone's daily candle matches. In forex there is no bell, so the data vendor picks the cut-off. The common choices are 17:00 New York time and 00:00 UTC, and some brokers use their own server time.
Take one ordinary week in January, when New York is on standard time (UTC minus 5 hours). The two conventions split the same price stream like this:
| Convention | One daily candle covers | Daily candles per week | What changes |
|---|---|---|---|
| 17:00 New York close | 17:00 to 17:00 New York (22:00 to 22:00 UTC) | 5 | Sunday evening trading folds into Monday's candle |
| 00:00 UTC close | 19:00 to 19:00 New York (00:00 to 00:00 UTC) | 6 | A short Sunday stub candle of about 2 hours appears; every weekday candle is shifted by 2 hours |
Those 2 hours matter more than they look. Say EUR/USD sells off hard between 17:00 and 19:00 New York on a Tuesday, then recovers through the Asian morning. On the New York feed, the sell-off sits at the start of Wednesday's candle and the recovery leaves a long lower wick: a hammer shape. On the UTC feed, the same sell-off lands at the end of Tuesday's candle and Wednesday opens near the low, so there is no hammer. Same pair, same prices, two charts, one pattern that exists on only one of them.
The 00:00 UTC convention also creates that sixth candle on Sunday. It's tiny and thin, yet it still counts as a full bar in any pattern scan. It can complete a two-bar or three-bar pattern on its own. We cover this boundary problem in much more depth in our piece on how your feed's daily close decides which forex patterns exist, so we won't repeat it here. The takeaway for this audit is simple: an equity study never had to decide where the day ends, but a forex test has to, and the answer changes the sample.
Which candlestick patterns survive the move to forex
Strip away anything that needs real volume or a gap, and you're left with patterns defined purely by the open, high, low and close of one or two bars. That's a smaller set than the full canon, but it's an honest one, because these shapes can be drawn on any forex chart, by any feed, without a hidden assumption.
| Pattern | Bars | Needs volume? | Needs a gap? | Forex verdict |
|---|---|---|---|---|
| Hammer / hanging man | 1 | No | No | Definable; still feed-sensitive on daily bars |
| Shooting star / inverted hammer | 1 | No | No | Definable |
| Doji | 1 | No | No | Definable; set a body-size threshold in pips |
| Bullish / bearish engulfing | 2 | No | No | Definable on bodies; the open sits at the prior close, so it reduces to "bigger opposite body" |
| Harami | 2 | No | No | Definable |
| Tweezer top / bottom | 2 | No | No | Definable; set a pip tolerance for "equal" |
| Three white soldiers / three black crows | 3 | No | No | Definable, with a looser open rule |
| Morning / evening star | 3 | No | Yes (classic) | Only as a relaxed, no-gap variant |
| Abandoned baby, island reversal, kicker | 2 to 3 | No | Yes | Does not form on weekday bars |
| Any pattern with a volume confirmation rule | Any | Yes | Varies | Tick count only; not the tested input |
Two notes on the "definable" rows. First, definable is not the same as reliable; it only means you can measure the pattern without breaking its definition. Our candlestick patterns cheat sheet ranked by reliability sorts the canon by how much context each shape needs, and that ranking still applies. Second, the single-bar shapes are the most exposed to the boundary problem above, because one moved cut-off can create or erase a long wick.
The hammer is a good example of both points. Even on stocks it's only half a signal, as our breakdown of what the next two bars do after a hammer shows. On forex, you add a feed question on top of a confirmation question.
The session effect, with the hours written down
Forex liquidity isn't spread evenly across 24 hours; it bunches up when the big centres are open at the same time, and the heaviest stretch is the London and New York overlap. A hammer formed in that window and a hammer formed in a quiet late-evening hour are not the same event, even if they look identical.
Here are the hours, stated in one timezone so you can line them up with your own chart:
- London and New York overlap: about 08:00 to 12:00 New York time. That's 13:00 to 17:00 UTC in northern winter and 12:00 to 16:00 UTC in northern summer.
- The awkward weeks: the US and the UK change their clocks on different dates. For about 2 to 3 weeks in March, and again for about 1 week around late October and early November, the gap between New York and London is 4 hours instead of 5. Any session filter keyed to a fixed UTC hour drifts during those weeks.
- The thin stretch: roughly 17:00 to 19:00 New York, right after the daily rollover, when spreads widen and quotes thin out.
The Ledger Mind's 69% against 54% split points the right way: patterns at busy hours should behave better than patterns at quiet ones. But without a sample size, a 15-point gap between the two figures could just be noise. If you want to use a session filter, write the hours down in New York time, adjust for the clock-change weeks, and log the session for every signal you take. That turns a vague "trade the opens" rule into something you can check.
This applies on intraday bars too. A 15-minute engulfing candle at 10:00 New York sits inside the deepest liquidity of the day. The same candle at 18:15 New York forms while the market is resetting. When we review traders' journals, the second kind is where "the pattern failed" shows up most often, usually with a wide spread eating the first few pips.
The one gap forex does produce
There is one place a currency chart gaps reliably, and that is the weekend. The market shuts on Friday at 17:00 New York and reopens on Sunday evening, around 17:00 New York, when trading starts again in Wellington and Sydney. Anything that happens in the 48 hours between can move the Sunday open away from Friday's close.
The irony is that no candlestick pattern was built around that gap, since the equity patterns assume a gap every morning, and forex delivers one a week. Feed conventions make it messier. On a 17:00 New York feed the weekend gap shows between Friday's candle and Monday's. On a 00:00 UTC feed it shows between Friday's candle and the Sunday stub. If you study gap behaviour on forex at all, study the weekly open on its own terms, not through a morning star template.
What to do instead of importing an equity win rate
If a number hasn't been measured on your market, the fix is to measure it. You don't need a research team for that, just a written procedure and enough patience to wait for the sample to fill. Here is the methodology we use when we test a candle shape on a currency pair.
- Fix the feed and the cut-off. Write down the broker or data vendor, the daily rollover time and the timezone. Never mix feeds in one test.
- Define the pattern in numbers. "Long lower wick" becomes "lower wick at least 2 times the body, upper wick no more than 0.3 times the body." Set doji and tweezer tolerances in pips.
- Drop volume and gap rules unless you can say exactly what replaces them. A tick-count filter is a new rule, so test it separately.
- Pick the outcome before you look. For example: close 5 bars later is higher than the signal bar's close, for a bullish pattern.
- Measure the baseline. Run the same outcome on every bar, not just pattern bars. If EUR/USD closes higher 5 bars later 50.4% of the time anyway, a 52% pattern hit rate is close to nothing.
- Log the session for each signal, using the New York hours above.
- Subtract costs. Use your real spread at the hour the signal fired, not the advertised minimum.
Then wait for the count, because a win rate from 30 trades is a guess. Here's why, using the standard 95% interval for a proportion:
| Signals in sample | Observed hit rate | Approximate 95% range |
|---|---|---|
| 30 | 55% | 37% to 73% |
| 100 | 55% | 45% to 65% |
| 400 | 55% | 50% to 60% |
| 1,000 | 55% | 52% to 58% |
At 100 signals, a 55% result can't be told apart from a coin flip. You need roughly 400 before the lower bound clears 50%, and that's before costs. On a daily chart, a pattern that fires twice a month takes more than 16 years to reach 400. That's the real reason most traders never measure: the honest sample takes a long time. Dropping to 1-hour or 15-minute bars fills it faster, at the price of more noise and more spread per trade.
In our experience the most useful output of this exercise isn't the win rate at all. It's the list of patterns you stop trading because they never clear the baseline on your feed. That list tends to be longer than people expect.
Where candlestick patterns fit next to an indicator
Candles describe what one or two bars did, but they don't describe the trend those bars sit in, or whether the market is ranging. That's why the patterns that hold up best tend to be the ones read in context: a hammer at a level in an uptrend, not a hammer anywhere.
Quantzee builds analytical indicators for TradingView that give that context on any market, forex included, with non-repainting signals so a historical test reflects what you would have seen live. They're tools for your own analysis, not trade advice, and they don't promise any result. If you trade currencies, our guide to forex indicators for 2026 covers which kinds of tools travel well across pairs and sessions. Pair any candle rule with a trend read, and test the pair together on your own feed before you size up.
