CRT Strategy BacktesterTest Candle Range Theory Free

Candle Range Theory reads price through the range of a single higher-timeframe candle. The core idea: a candle's range sets a zone, price sweeps one side of it to grab liquidity, then reverses and expands toward the other side. Build a range, sweep one side, expand to the other — that rhythm is the whole model. We break the concept down fully in the CRT reference entry and at article length in What Is CRT (Candle Range Theory)?

Because those levels come from a candle's range, CRT is rules-based — and anything you can write as an if-this-then-that, you can backtest.

This is the step almost everyone skips, and it is the one that decides whether the results mean anything. "Wait for the sweep and enter on the reversal" is not a rule — it is a description. Two people running it on the same chart will produce different trades, and one person running it twice will too, which means the win rate that comes out belongs to nobody.

A rule is testable when a stranger could apply it to your chart and mark the same trades you did. That takes five decisions, made in advance and written down:

1. Which candle forms the range. Pick one and fix it. The prior 4-hour candle, the prior day, the prior weekly — not "whichever range looks cleanest", which is hindsight wearing a rule's clothes.

2. What counts as a sweep. A wick through the edge? A close beyond it and back inside within one candle? A minimum distance past the level, so a one-tick brush does not count? Any of these is defensible; leaving it undefined is not.

3. What confirms the reversal. The entry trigger on your execution timeframe — a shift in structure, a return into the range, a displacement candle. State the timeframe too, because "a shift in structure" means something different on the 1-minute and the 15.

4. Where the stop goes. Beyond the sweep's extreme, and by how much. This number decides your position size and therefore most of your result, so it cannot be improvised trade by trade.

5. Where you are out. The opposite edge of the range is the model's own target and the honest default. If you use partials or a runner, define them now — deciding at the time is how a 1:3 test quietly becomes a 1:1 one.

Write those five down before the first replay candle. If a setup arrives and you find yourself reasoning about whether it qualifies, the rule was not specific enough — fix the rule, restart the sample, and do not fix it mid-test.

The sweep-then-expand pattern looks obvious in hindsight and vanishes in real time. Backtesting is what closes that gap: it tells you how often a clean CRT setup actually appears on your pair, what your real win rate is across a few hundred of them, and which sessions and instruments it works best on. You can't feel your way to those numbers — you have to count them.

Record the reward:risk alongside the win rate, because neither means anything alone. A 40% win rate is excellent at 1:3 and fatal at 1:1 — the risk/reward calculator gives the break-even win rate any ratio demands, which is the number your CRT results have to clear before they count as an edge.

Testing also builds the conviction to hold a CRT trade through the manipulation leg. The traders who don't panic on the sweep are the ones who've already watched it resolve hundreds of times in replay.

Press the button on this page and the chart opens with the CRT overlay already on. It watches the model timeframe you choose — 4-hour by default, adjustable from 15-minute to monthly — and the moment a candle sweeps the previous candle's high or low and closes back inside its range, it marks the setup: bullish or bearish, the sweep itself, and the opposite edge as the target. It only ever sees candles up to your replay cursor, so a CRT is marked when it would have been visible live and not a candle earlier.

What the overlay does not do is decide whether you would have taken it. That part stays yours: your entry trigger on your execution timeframe, your stop, your exit. The overlay removes the drawing; it does not remove the judgement, and a backtest where the tool picks the trades would be measuring the tool.

The order matters. Set the range candle first, before the session starts — on a fresh replay the range for the day is already known and the sweep is not, which is exactly the information you would have had live. Then step forward. When the sweep happens, mark it before you know whether it holds. That single discipline is what separates a backtest from a scroll-back, and it is the reason bar-by-bar replay produces win rates ten to twenty points lower — and truer — than reviewing a chart you have already seen.

Log every setup, including the ones that failed and the ones you were unsure about. A journal that only records the trades you liked measures your memory, not your edge; what to log in a trading journal covers the fields that turn a log into an answer.

For the full write-up with worked examples: How to Backtest the CRT Strategy.

The four things to mark on every CRT setup, and where each one gets fudged.

Thirty trades constrains almost nothing. That is not a rule of thumb, it is arithmetic: at 30 samples the confidence interval around a win rate is roughly ±18 points, which is wide enough to contain both "excellent" and "unprofitable" at once. Traders abandon working strategies and marry losing ones inside that band all the time.

The sample you actually need scales with the size of the edge you are trying to detect. A strong CRT ruleset producing +0.5R a trade clears the bar in well under 50 occurrences. A thin one at +0.05R needs a sample nobody will ever trade. The backtest sample size calculator gives the number for your figures, and tells you the odds a strategy with no edge at all would have scored what yours just did.

Know the baseline before you judge a number. We measured the bare three-candle pattern on 13 markets over four years, about 137,000 setups: with the stop at the sweep wick and the target at the far side, it won about 42% of the time at a median 1.4 : 1, which is break-even before costs on the daily, 4-hour and 1-hour alike. So a CRT ruleset that wins 42% has shown nothing yet. It has an edge only when it clearly beats its own break-even rate, and the CRT statistics tool gives that baseline for the market and timeframe you trade.

Two practical consequences. First, count occurrences, not months — a year of a rule that fires twice a week is a bigger sample than a year of one that fires twice a month, and the calendar tells you nothing. Second, resist testing five variants and reporting the best; that is a search across rules, and the winner of a search is high partly because it won a search. If you do vary the rules, keep a fresh slice of history the variants have never touched and confirm on that — out-of-sample testing explains how to hold that data back properly.

Picking the range candle in hindsight. The single biggest one. With the outcome visible, there is always a candle whose range explains the move, and choosing it feels like analysis. Fix the range rule to a specific candle before the session and the problem disappears.

Marking the sweep after it resolved. A sweep that reversed is obvious; a sweep that kept going is called a breakout in hindsight and never enters the log. Mark it when it happens or your win rate is measuring your memory.

Moving the stop. A stop that gets "a bit more room" on the trades that would have been hit converts losses into wins on paper and cannot be repeated live. Same for a target taken early on winners but held on losers.

Ignoring costs. Spread, commission and swap come straight off expectancy, and a CRT ruleset at +0.05R gross is frequently negative net. Subtract your round-trip cost from every winner before you score the sample.

Testing one instrument in one regime. A rule tuned to a trending quarter on the Nasdaq is a rule about that quarter. Run it on a second market and a different stretch of history before believing it — why backtest results differ from live covers the rest of the gap, and the mistakes that make backtest results worthless is the full list.

A win rate without a position size attached is trivia. Once the sample has given you a real win rate and an average R, three numbers turn it into a plan.

The stop distance sets the position, not habit — a sweep that ran deep needs a smaller size for the same risk, and the position size calculator does that conversion. The risk/reward calculator gives the win rate your ratio has to beat to break even. And because CRT setups cluster — quiet weeks produce none, expansion weeks produce several — the losing runs are longer than the win rate feels; the risk of ruin calculator tells you what size survives them, and the drawdown recovery calculator what it costs if it does not.

CRTLAB is built around this whole workflow. The replay itself is not gated in any way — all 38 markets, the full history on each, every overlay the method uses — so you can mark CRT setups by hand on your weekly candle allowance and never touch a plan.