Does CRT Work? Candle Range Theory, Measured

The short answer

A CRT setup is three candles. Candle 1 sets a range. Candle 2 trades beyond one edge of it (the sweep) and closes back inside. The theory says price then expands to the other side of candle 1. We found every one of those setups on daily, 4-hour and 1-hour candles across gold, silver, oil, the Nasdaq, the Dow, the S&P 500, the DAX, four forex pairs, Bitcoin and Ethereum: 137,146 setups, 1 Jul 2022 – 18 Sept 2026.

The textbook trade enters at candle 2's close, puts the stop at candle 2's sweep wick and targets the far side of candle 1. It reached the target first 42–43% of the time, at a median reward:risk of about 1.4 : 1. Those two numbers cancel out. Expectancy before costs was +0.01R a trade on the daily, −0.02R on the 4-hour and −0.03R on the 1-hour. A price moving at random would have scored the same.

So the bare pattern is not an edge. It is not a losing pattern either: the market prices it about fairly. That is a useful thing to know, because it tells you where a CRT trader's edge has to come from. It comes from what you add on top of the three candles, and that is what you have to test.

What the next candle actually does

Start with the claim itself: after the sweep, price goes to the other side. On the very next candle it did so on 29–31% of setups. The range's 50% was reached far more often, on 65–68%, and candle 2's sweep wick was taken out on 41%. These overlap, because one candle can touch several levels.

Two things follow. The 50% is the realistic first target: two setups in three touch it on the next candle. And the full expansion is the minority outcome. A plan that needs the far side on the next candle is betting on something that happens about three times in ten.

Why a 42% win rate is break-even

The break-even win rate of any trade is risk ÷ (risk + reward). A trade that risks 1 to make 1.4 breaks even at 1 ÷ 2.4, about 42%. So a CRT trade that wins 42.3% of the time with a median payoff of 1.4 : 1 is not a bad pattern with a low win rate. It is a pattern whose win rate is exactly what its payoff predicts.

That is also what a random price would do. If price moved at random, a trade with a fixed stop and target would win just often enough to break even: a near target wins more often, a far one less often, and the expectancy is zero either way. So the real test of an edge is not a high win rate. It is a win rate above the break-even line for the same trades. On the 4-hour, CRT won 42.3% against a break-even of 42.7%. On the daily, 43.2% against 43.1%. The risk/reward calculator does the same sum for your own numbers.

The filters CRT is usually taught with

We picked the filters before looking at the results: the ones most CRT traders add. Picking filters after seeing which ones worked would be curve-fitting. This is the 4-hour, all 13 markets together, with the target at the far side of the range; the tool above shows the other timeframes and each market on its own. “Previous day” means the setup points the same way as the last New York day's candle, and the sessions are the New York hour the candle after the setup opens in.

Not one of them moved the expectancy off zero. What each filter did was trade win rate for payoff. The clearest case is the 50% (equilibrium) close: setups where candle 2 closed past candle 1's midpoint won 72% of the time, but the entry was already most of the way to the target, so they paid only 0.4 : 1. Setups that closed short of it won 36% and paid 1.8 : 1. Same expectancy.

Taking only setups that agree with the previous day's candle made no difference (42.0% with it, 42.6% against it). Sessions changed how far price travelled, and the stop and target travelled with it. Bullish setups won more often than bearish ones on every timeframe and paid less, which is mostly four years of rising markets. The 1-hour tells the same story on 108,119 setups.

Every market

The same textbook trade, market by market, on the 4-hour. The ± column is the uncertainty (two standard errors). An expectancy inside it cannot be told apart from zero.

None of the 39 market and timeframe combinations showed an expectancy clearly above zero. On the 1-hour, Gold (XAU/USD), S&P 500, GBP/USD, USD/JPY came out slightly below zero. Part of that is our cautious scoring of 15-minute bars that touched both levels, which costs the 1-hour most. Notice how wide the daily uncertainty is: about ±0.13–0.30R per market, from a few hundred setups in four years. If a sample that size cannot pin down a daily CRT, thirty trades in a notebook cannot either. The sample size calculator shows how many you need.

So where would a CRT edge come from?

From the parts of the method this measurement leaves out, because they are the parts each trader does differently:

• Where the sweep happens. We counted every candle that took the previous candle's high or low. Most CRT traders only want the sweeps that also take a higher-timeframe level: the previous day or week, an old swing, the edge of a fair value gap. That is a narrower, different set of trades.
• Which candles. Many traders only use particular 4-hour candles, such as the 1:00, 5:00 and 9:00 New York candles on a chart that closes at 17:00. Ours are drawn on UTC hours, like the CRTLAB chart.
• The entry. Dropping to a lower timeframe and waiting for a market structure shift inside candle 3 puts the stop far closer than the sweep wick. That raises the payoff and lowers the win rate. Whether it nets out ahead is exactly the question.
• What you skip. News days, Fridays, ranges that are too small or too big. Every skip rule is a filter, and every filter can be measured the same way.

Each of those is a rule you can write down, and each one can be judged the way this page judges the bare pattern: win rate against the break-even line, over a few hundred setups. That is what backtesting CRT is for. The CRT strategy backtester marks every setup on the chart as it forms, so you replay them one candle at a time, apply your own rules and log the results.

What this means if you trade CRT

• Judge it on expectancy, not win rate. A 70% CRT with a 0.4:1 payoff and a 35% CRT at 1.8:1 are the same trade.
• Use the 50% as the first target, not as an edge. It is reached on the next candle two times in three, which makes it a good place for a partial. It does not make the trade better.
• Count your costs. A pattern at break-even before costs is below it after them, and the smaller the stop, the bigger the bite. That falls hardest on the 1-hour.
• Test your version, not the textbook one. If your CRT has an edge, it lives in the rules you add. Measure them with and without, on the same sample, and keep only what beats the break-even line.

How we measured it

CRTLAB's own 15-minute price history, 1 Jul 2022 – 18 Sept 2026, for every market except the DAX (from 11 Mar 2025) and crypto (to 21 Sept 2026). The 4-hour and 1-hour candles are built on UTC hours, the same candles the CRTLAB chart draws. The daily candle is the New York trading day, 18:00 to 17:00, as on the average daily range and best time to trade tools.

A setup is a candle that trades beyond exactly one edge of the previous candle and closes back inside it. A candle that takes both edges is an outside bar, and one that closes beyond the edge it took is a breakout; neither is counted. Each trade was walked forward on 15-minute bars for up to 10 candles to see whether the target or the sweep wick was touched first. A 15-minute bar that touched both is scored as a loss, because we cannot see inside it. Fewer than 2% of trades were still open after 10 candles and are left out. No spread or commission is included: this measures the pattern, not a broker.

Frequently asked questions