ICT Order Blocks: An Empirical Test

Test period: January 2012 – October 2026 · Instruments: EURUSD, GBPUSD, USDJPY, AUDUSD, XAUUSD (M1, bid) · Out-of-sample block: 2012–2018 · Published: November 2026

1. Objective

The order block (OB) is one of the most widely taught ICT / Smart Money Concepts tools. In its standard form, a bullish OB is the last down-close candle before an impulsive move up (bearish: the last up-close candle before a move down). Price is said to react when it returns to that candle, and the 50% level of the candle — the mean threshold — is used to judge whether the block is “still valid”.

This study turns those statements into fixed, mechanical rules and tests each one against a random baseline, across five instruments and nearly 15 years of one-minute data.

1.1 Claims tested

2. Methodology

2.1 Data

• Source: Dukascopy, M1 bid (via the dukascopy-node CLI)
• Range: 2012-01-01 to 2026-10-30, about 5.2–5.5 million one-minute bars per instrument after removing flat (OHLC-equal) bars
• Time handling: all timestamps converted to New York time with full US daylight-saving rules. The 4-hour candles are NY-aligned (opening at 18:00, 22:00, 02:00, 06:00, 10:00 and 14:00 NY), which is one hour off a typical GMT+2/+3 broker’s H4 chart
• Out-of-sample split: all definitions and parameters were fixed while working on 2019–2026. The 2012–2018 block was not looked at until the final run, so it acts as a genuine out-of-sample check. Every headline result is reported for both blocks.

2.2 Order block definition (fixed before testing)

• OB candle: an opposite-colour candle (down-close for bullish) whose low is the lowest of the previous 12 candles (bearish: highest high), with a body of at least 0.1 × ATR(14)
• Validation: within 12 candles, a later candle trades above the OB high (bearish: below its low) before any candle closes below the OB low
• Clean validation: the validating candle must close outside the zone. Without this rule, price can already be inside the zone when validation is confirmed, and a limit order at the near edge would be credited at a price that was no longer available (see Section 4)
• First return: the first touch of the near edge (bullish: the candle open) within 48 candles of validation. The exact touch minute is found on M1 data
• Far edge (stop): the lower of the OB low and the next candle’s low, so that a sweep by the next candle is included
• Fill realism: if the touching M1 bar opens beyond the limit price (a gap into the zone), the fill is that bar’s open, not the limit price. After the clean-validation rule this affects only 1.0% of OB fills and 1.7% of FVG fills

2.3 Random baseline

Every trade is paired with ten random entries in the same direction, at the same minute of the day, on other days within ±60 days, with the same stop distance. This removes the effects of session, volatility regime and long-term drift. The quantity reported throughout is the edge: the real win rate minus the random win rate, in percentage points (pp). An OB with no special property should show an edge of zero.

2.4 Outcomes and costs

• Fixed targets of 1R and 2R; whichever of target or stop is hit first on M1 data, within three days
• If the entry M1 bar itself already touches the stop, the trade is counted as a loss; if target and stop are both hit inside the same later M1 bar, the trade is excluded as unresolved
• Minimum stop distance: 10 × the round-trip cost, so costs never exceed 0.1R
• Net expectancy is reported after an approximate prop-firm cost (spread plus commission, round trip, October 2026):

3. Results

3.1 Claim ① — “A close beyond the mean threshold invalidates the OB”

A popular educational variant defines the OB as the entire run of same-direction candles that starts at the first candle tagging a higher-timeframe level, and advises against the trade once price closes beyond the 50% of that zone. We tested it mechanically on M5:

• HTF level: the previous NY-aligned 4-hour candle’s high for shorts and its low for longs; first tag inside the current 4-hour candle
• OB: the run of up-close (short case) candles starting at the tagging candle; change of state of delivery (CISD): a close beyond the run’s body within two hours
• On the first return to the 50% level of the zone, the M5 close decides the group: A: closed back inside (MT respected); B: closed beyond MT
• Entry at that close; stop beyond the extreme since the tag; fixed 1R / 2R; random baseline as in 2.3

The claim predicts A − B > 0. It was negative at 1R in both blocks and in all five instruments, and essentially zero at 2R. Even the raw win rates go the wrong way: at 1R, trades that closed beyond the MT won 50.7% of the time against 49.8% for trades that respected it. (In most of these trades the stop sits at the 10 × cost floor, so both groups take the same fixed risk.)

In an earlier version of this test that used the zone’s own structure for stop and target, the rule did appear to work, because trades that close beyond the MT start much closer to their stop. That gap matched what a random walk produces at the same distances. Once both groups are compared with random entries at the same stop distance, respecting the MT carries no information.

3.2 Claim ② — Is an OB candle a special level?

We applied exactly the same zone, validation and first-return rules (M5) to every candle, not just OB candles, and grouped the touches by the two properties that define an OB: colour (opposite vs. same direction) and location (at a 12-candle extreme versus elsewhere). From each touch, we recorded whether price moved +1 ATR in the OB direction before −1 ATR. OBs were compared with each control group within matching cells of instrument, direction, displacement multiple and candle size; standard errors use a month-block bootstrap. Sample: 340,935 OB touches against more than 3.3 million control touches.

The result is the same in both blocks. Being at a local extreme accounts for the whole difference (about +1.1 to +1.5 pp). The candle’s colour — the property that makes it an “order block” rather than just a swing point — adds nothing. An OB behaves like any other candle at a local high or low.

3.3 Claim ③ — Mean threshold on higher-timeframe OBs

For 1-hour and 4-hour OBs, we followed price on M1 after the first touch of the near edge and asked two questions: how often price reaches the MT at all, and whether judging the MT on the close of an M1, M5, M15, H1 or H4 candle helps to separate winning and losing trades.

• Price reaches the MT in 78% (1H) and 77% (4H) of first returns. The MT is not a boundary that price usually respects; it is reached most of the time.
• Waiting for the higher-timeframe close costs information: by the H1/H4 close, about 22% of these OBs have already been broken through the far edge.

A − B: edge of trades whose candle closed inside the MT, minus those that closed beyond it (1R). Entry at that close, stop at the far edge.

No judging timeframe shows a consistent positive A − B. Signs flip between the two blocks and every pooled value is within about ±1.4 standard errors of zero (18 comparisons, none surviving a multiple-comparison correction).

3.4 Claim ④ — Trading the first return to a 4-hour OB or FVG

Limit entry at the near edge of a 4-hour zone on its first return, stop at the far edge, fixed targets. Variants test the common fixes for being stopped out just beyond the zone: wider stops, and waiting for the sweep and an M5 close back inside before entering.

Net expectancy is per trade, in R, after the cost table in Section 2.4.

• A small, real edge. The 4-hour OB first return beats random by about +2.4 pp at both 1R and 2R (z ≈ 3.6–3.8), with almost identical values in the out-of-sample 2012–2018 block (+2.5 pp) and in 2019–2026 (+2.2 pp). At 1R, four of five instruments are positive; GBPUSD is the exception.
• FVG confluence does not help. OBs whose displacement left a 4-hour FVG (confirmed before the touch) show +2.5 pp, against +2.3 pp for OBs without one.
• Common “fixes” make it worse. Getting stopped out just beyond the OB and then watching price reverse happens often, but moving the stop further away dilutes the edge, and waiting for the sweep before entering eliminates it.
• The FVG first return shows a smaller edge (+1.3 pp at 1R) that is also present in both blocks at 1R, though weaker at 2R in 2012–2018.
• Costs decide. Before costs, the baseline OB entry earns +0.034R per trade at 1R. The assumed prop-firm cost is 0.064R per trade, so the net result is -0.030R (1R) and -0.056R (2R). Costs would have to be roughly 53% of the assumed level for the setup to break even.

4. Corrections made during this study

5. Conclusion

• ① The rule “a close beyond the mean threshold invalidates the OB” is not supported. Its apparent success comes from stop-distance geometry, not from information in the MT.
• ② OB candles react no differently from any other candle at a local extreme. The colour condition that defines an OB adds nothing.
• ③ Price reaches the MT on about 77–78% of first returns, and no lower-timeframe close relative to the MT reliably separates winners from losers.
• ④ The first return to a 4-hour OB carries a small edge (about +2.4 pp over random) that held in the untouched 2012–2018 data. It is too small to survive realistic prop-firm costs, and FVG confluence, wider stops or sweep confirmation do not improve it.

In short: the measurable part of the order block concept is “price tends to react a little at prior local extremes”. The parts specific to ICT — the candle colour, the mean threshold, the FVG requirement — did not add anything that this data could detect.

6. Limitations

• Each concept was tested with one mechanical definition. Discretionary traders apply OBs with context (higher-timeframe bias, liquidity, time of day) that is not captured here.
• Events overlap in time and are not fully independent; the event-level z-scores are somewhat optimistic. The block-bootstrap errors in Section 3.2 are the more conservative figures.
• Costs are representative estimates, not a specific account’s live fills; slippage on stops is not modelled. Results are on bid prices only.
• 4-hour candles are NY-aligned. Broker H4 candles (GMT+2/+3) are shifted by one hour and produce somewhat different OBs and FVGs.
• An earlier, shorter pass (2019–2026, also covering BTCUSD and US100) pointed in the same direction on every claim; those figures are not reported here because they predate the corrections in Section 4.

7. Next Steps

• Re-test the 4-hour OB first return in lower-cost venues (e.g. index futures), where the measured edge might exceed costs
• Test whether the “local extreme” effect in Section 3.2 is simply a liquidity-pool effect (equal highs/lows near the zone)
• Check robustness to the 4-hour candle alignment (NY 17:00 vs 18:00 open)

Appendix — Technical notes: Zone detection and outcome loops are JIT-compiled with Numba and run on M1 data for exact touch and stop/target ordering. Random baselines use 10 same-direction, same-minute-of-day entries within ±60 days per trade. Instruments and costs: EURUSD 1.4 pips, GBPUSD 2.2 pips, USDJPY 2.5 pips, AUDUSD 1.6 pips, XAUUSD $0.58 round trip. Full methodology and scripts available on request.