Free research · Order blocks

Do Order Blocks Work on Gold? We Tested 288 Versions on 9 Years of XAUUSD

Short answer: no version of the order block passed our test, on any chart from 1-minute to weekly. On the hourly chart, buying them made +2.58 basis points a trade, and buying candles that looked identical except that they had no impulse behind them made +1.64. The difference between them, +0.91 bp, is well inside chance (t +0.50). Almost everything the order block appeared to earn was gold going up, and the small remainder was not the block.

Tested onGold (XAUUSD)
DataJanuary 2017 to September 2026 (9 years 8 months)
Rule versions tested288
Win rate we measured (after costs)43.4% over 188,990 trades (M1 buy, the biggest sample)Across every chart in the main table: lowest 40.7% on W1 sell (27 trades), highest 60.0% on W1 buy (25 trades). Winning more often is not the same as making money; the tables below show both.

What is an order block?

An order block is the last candle going one way immediately before price runs hard the other way, the last down candle before a sharp rally. The claim is that a large buyer filled orders there, left unfilled business behind, and will defend the level when price returns to it. So you wait for price to come back to that candle and buy.

Short answer: no version of the order block passed our test, on any chart from 1-minute to weekly. On the hourly chart, buying them made +2.58 basis points a trade, and buying candles that looked identical except that they had no impulse behind them made +1.64. The difference between them, +0.91 bp, is well inside chance (t +0.50). Almost everything the order block appeared to earn was gold going up, and the small remainder was not the block.


Read this first: what this test can and cannot tell you

How to read the numbers (skip if you already know)

Term What it means here
bp (basis point) 0.01% of the price. At $4,300 gold, 1 bp ≈ $0.43 on a 1-ounce position.
The spread you pay Charged once per round trip, at least 15 points, from the broker's own record. Measured across this sample it averages 0.89 bp per trade, and it has been shrinking as gold has risen: 1.19 bp in 2017, 0.79 in 2023, 0.37 so far in 2026, because 15 points is a smaller share of $4,300 than of $1,250.
Win rate Share of trades that ended above the entry after costs. On its own it proves nothing, a rule can win 70% and still lose money.
t How far a result is from zero, measured in its own noise. Roughly: below 2 is ordinary randomness; our bar is far higher because we tried many versions.
The bar (3.75) We tested 288 versions. Test 288 things and some will look good by luck, so the threshold is raised to match (Bonferroni). A version must clear |t| ≥ 3.75.
Alpha The result after subtracting what gold itself did over the same holding time in the same year. It answers: did the rule earn this, or did the market?
The random-series check The identical 288-version grid is run on Volatility 75, a synthetic instrument with no economy, no auctions and no institutions. If a rule "works" there, our machinery is manufacturing results and the whole test is void.

What we tested

The idea, in plain words. An order block is the last candle going one way immediately before price runs hard the other way, the last down candle before a sharp rally. The claim is that a large buyer filled orders there, left unfilled business behind, and will defend the level when price returns to it. So you wait for price to come back to that candle and buy.

The rules, exactly.


Results by chart: the standard version (impulse 1 ATR, body zone, exit after 12 candles)

Chart Direction Trades Win rate Net per trade (bp) t Passed the bar?
M1 buy 188,990 43.4% −0.82 −37.14 no, loses
M1 sell 188,679 43.0% −0.86 −39.62 no, loses
M5 buy 36,745 48.4% −0.58 −5.15 no, loses
M5 sell 36,935 46.2% −1.12 −10.03 no, loses
M15 buy 11,549 49.8% −0.29 −0.82 no
M15 sell 11,683 46.9% −1.45 −4.28 no, loses
M30 buy 5,676 51.3% +0.06 +0.08 no
M30 sell 5,819 47.5% −1.59 −2.24 no
H1 buy 2,865 52.7% +2.58 +1.86 no
H1 sell 2,917 47.3% −2.84 −2.08 no
H4 buy 817 54.6% +12.99 +2.55 no
H4 sell 835 47.1% −11.43 −2.30 no
D1 buy 156 57.1% +30.52 +1.31 too few trades
D1 sell 170 45.9% −37.35 −1.59 too few trades
W1 buy 25 60.0% +120.39 +0.83 too few trades
W1 sell 27 40.7% −343.09 −2.30 too few trades

Across all 288 versions: zero passed. The strongest was nowhere near the bar. On the random control series, also zero passed, with a maximum |t| of 3.13, the machinery is not manufacturing findings.

Read the shape of that table rather than any single row. Buying wins more often than selling on every chart, and the buy side turns positive from the 30-minute chart up. That is not the order block working. That is gold, which roughly tripled over the sample: any rule that buys and holds for twelve candles inherits it.


The test that matters: does the block add anything?

For every order block we also took the same candle, the same zone, the same touch, with no impulse behind it (the following move smaller than 1 ATR). That is a candle that fails the only condition that makes it an order block. If the concept is real, the blocks should beat these.

Chart Direction Order blocks (bp) Same touch, no impulse (bp) Difference t
M15 buy −0.29 −0.28 −0.02 −0.04
M15 sell −1.45 −1.46 +0.01 +0.02
H1 buy +2.58 +1.64 +0.91 +0.50
H1 sell −2.84 −3.20 +0.39 +0.22
H4 buy +12.99 +8.07 +4.86 +0.72
H4 sell −11.43 −11.36 +0.07 +0.01

Not one difference is distinguishable from chance. Whatever is being measured when you trade an order block, it is not the impulse, the displacement or the "unfilled orders", it is the touch of a recent price level, and an ordinary candle provides that just as well.


Year by year: buying on the hourly chart, exit after 12 candles

We report every result year by year, because effects in gold have been growing and a single pooled number hides that.

Year Trades Win rate Net per trade (bp)
2017 273 46.9% −0.13
2018 305 52.8% +1.47
2019 294 52.7% +0.43
2020 306 58.8% +1.10
2021 296 51.4% −0.95
2022 297 47.1% +0.19
2023 280 51.1% +2.33
2024 288 57.3% +6.43
2025 303 58.4% +17.19
2026 223 49.3% −4.32

The whole of the hourly result is 2024 and 2025, the two years gold rose fastest, and it is already negative in 2026. Subtract gold's own move over the same holding periods and +2.58 bp becomes +0.83 bp, t +0.60: nothing.

This is the single most useful thing on the page. A trader who backtested order blocks over 2024 to 25 would have found a strong, convincing, profitable pattern, and would have been measuring the market, not the method.


With a stop and a target

Traders rarely exit on a clock, so we also ran the textbook plan: stop beyond the far side of the block, target twice the risk, up to 200 candles. A market with no pattern at all reaches a 2:1 target before its stop about one time in three, 33.3% is the number to beat, and beating it is the only thing a win rate can prove.

(Descriptive: this plan was added after the main test and was not pre-registered.)

Chart Direction Trades Target before stop Chance gives Mean result
M15 buy 13,177 28.4% 33.3% −0.15 R
M15 sell 14,751 26.7% 33.3% −0.20 R
H1 buy 3,320 32.1% 33.3% −0.04 R
H1 sell 3,617 29.3% 33.3% −0.13 R
H4 buy 1,038 33.9% 33.3% +0.02 R
H4 sell 1,090 27.1% 33.3% −0.19 R

Every version sits at or below the coin-flip line. The gap below it on the fast charts is the spread, charged once per trade and paid whichever way the trade goes.


Things we checked that did not rescue it


What this page does not say


How we tested

Reproduce it: every rule is written out in full above, so anyone with gold price data can rebuild this test and check our numbers. The candles are the broker's and are not ours to redistribute.


Request a backtest

Want a concept, a variation or another market tested the same way? Tell us what to test and where to send the result. It is free.

We only use your contact to send you this result.

Disagree? Bring results.

We publish numbers, not opinions, and we answer with numbers. If you believe a concept works and our test missed it, send us your exact rules, the market and timeframe, your data source and dates, the costs you charged, and your result with its sample size. We will run your version and reply with the actual statistics: every trade, the win rate, the result after costs, year by year. If your test holds up, we correct this page, dated, with credit to you if you want it.

Claims without results will not change a page. A screenshot of a winning trade is one trade; the tables here are thousands.

Send it to support@fxsynapseai.com with the subject “Research correction”, or use the request form above.