SMC Rejection Block: An Empirical Test

Test period: January 2012 – October 2026 · Instruments: EURUSD, GBPUSD, USDJPY, AUDUSD, XAUUSD, BTCUSD (from May 2017), US100 (M1, bid) · Discovery block: 2012–2018 · Replication block: 2019–2026 · Timeframes: M15, H1, H4, D1 · Published: November 2026

1. Objective

The rejection block is an ICT price-delivery array that sits further from equilibrium than the order block. In the ICT Mentorship Core Content, a bearish RB is a swing high with long upper wicks: price runs above the candle bodies, takes buy-side liquidity and is rejected. The zone is defined from the highest open or close of the swing to the highest wick, and is then traded like a bearish order block — either by selling at the body level with a stop above the wick, or by waiting for price to trade above the level and selling on a stop below it. The ideal setting is a swing high within an intermediate-term downtrend (mirrored for bullish RBs).

This study turns those statements into fixed, mechanical rules and tests them against random entries across seven instruments, four timeframes and nearly 15 years of one-minute data. It was run in two passes: a minimal “shape only” definition (v1), and a second definition that follows the lecture as closely as possible (v2).

1.1 Claims tested

2. Methodology

2.1 Data

• Source: Dukascopy, M1 bid (via the dukascopy-node CLI). BTCUSD is available from May 2017 only
• Range: 2012-01-01 to 2026-10-31, about 4.2–5.5 million one-minute bars per instrument after removing flat (OHLC-equal) bars
• Candles: M15, H1 and H4 are aligned to UTC (H4 opens at 00:00, 04:00, … UTC). Daily candles open at 22:00 UTC, which equals the New York 17:00 close in winter and is one hour off in summer. This differs from the NY-aligned H4 used in the Order Blocks study
• Two blocks: 2012–2018 is the discovery block and 2019–2026 the replication block. A result is treated as real only if it appears in both

2.2 Swings

A swing high is a candle whose high is strictly above the highs of the two candles on each side (swing low mirrored). It is confirmed only at the close of the second candle after it; nothing is traded before that moment, so there is no look-ahead. If price is already inside the zone at confirmation, the swing is skipped (no departure from the zone).

2.3 Definition v1 — shape only

• Zone: the swing candle’s own body edge (highest of open/close for a swing high) to its wick extreme
• Wick ratio = wick length ÷ candle range. It is not fixed in advance; every swing is kept and grouped into four bins (<0.2, 0.2–0.4, 0.4–0.6, ≥0.6). The ≥0.6 bin is treated as the RB
• Entry: limit at the body edge on the first return, valid for 50 candles. Stop at the wick extreme. Targets 1R and 2R, maximum holding time 50 candles
• Timeframes: M15, H1, H4

2.4 Definition v2 — following the Core Content lecture

• Multi-candle swing: the swing is the extreme candle plus two candles on each side. The zone runs from the highest open or close among those five candles to the highest wick
• Failure on a body basis: the extreme candle’s body must not exceed the highest body of the candles before it — price runs above the bodies only with a wick
• Zones thinner than 0.1 × ATR(14) are dropped
• Trend context: major swings on the same timeframe (ten candles on each side). If the last two major highs and the last two major lows are both lower, the trend is down; both higher, up; otherwise none. A bearish RB in a downtrend (or bullish in an uptrend) is trend-aligned; results are also reported for counter-trend and no-trend RBs
• Entries: E1 — limit at the body level. E2 — price must first trade beyond the body level (into the wick zone) without reaching the wick extreme; the entry is then a stop order at the body level as price falls back. An order is cancelled if the wick extreme is hit first. Orders are valid for 50 candles (20 on D1)
• Targets: the lowest body of the previous swing low (bearish; mirrored for bullish), as suggested in the lecture, plus fixed 1R and 2R for comparison. Maximum holding time 50 candles (20 on D1)
• Timeframes: M15, H1, H4, D1

2.5 Baselines

• Random entry: each RB trade is paired with a market entry at a random time in the same block, in the same direction, with the same stop distance and the same target distance. In v2 the random time is drawn from candles with the same trend state, so the comparison isolates the RB from the trend filter
• Placebo (v2, E2 only): the close of a random candle (same block, same trend state) is used as a fake “body level”, with the same stop distance and exactly the same E2 rules. If the RB adds anything, E2 on real RBs should beat E2 on random levels

Note: this random baseline is simpler than the same-minute-of-day baseline used in the Order Blocks study; it does not control for session effects.

2.6 Outcomes and costs

• All fills, stops and targets are resolved on M1 data. If the fill minute itself reaches the stop, the trade is a loss. If target and stop are both reached inside one later minute, the trade is a loss. If, for E2, the fall-back and the wick extreme occur in the same minute, the trade is a loss
• Trades not resolved within the holding time are closed at market
• Net expectancy is reported after an approximate prop-firm cost (spread plus commission, round trip, October 2026). No minimum stop distance is imposed, so the mean cost in R is high on low timeframes, where many stops are small:

3. Results

3.1 Claim ① — Is a long-wick swing a special level? (v1)

Across all instruments the test produced 567,561 (M15), 144,925 (H1) and 39,079 (H4) swings. About 18% of them had a wick ratio of 0.6 or more: 105,820, 25,841 and 6,464 RBs respectively — on the 4-hour chart, roughly one per instrument per week.

RB = wick ratio ≥ 0.6, limit entry at the body edge, 1R target. Edge is the win rate minus the random win rate; t compares the expectancy (in R) of each RB trade with its paired random trade.

The win rate is slightly above random in every row, but the difference in expectancy is not significant on H1 or H4, and the only significant result in the discovery block (M15, t = +3.6) reversed sign in the replication block (t = −2.2). After costs, every row is negative.

3.2 Claim ② — Full definition, aggressive limit entry (v2, E1)

Trend-aligned RBs only. Counter-trend and no-trend RBs give the same or a worse picture.

With the full lecture definition and the trend filter, the limit entry at the body level does no better than random on any timeframe, in either block; several M15 and H4 rows are significantly worse than random. Net expectancy is negative everywhere except one daily row (+0.03R, n = 214).

3.3 Claim ③ — Confirmation entry (v2, E2) and the placebo test

At first sight, E2 looked like a breakthrough: on M15 and H1 its gross expectancy beat random by t = 7 to 18 in both blocks. But the effect was the same size for counter-trend and no-trend RBs, which is not what the concept predicts. The placebo test explains it.

Trend-aligned RBs, 1R target.

• The E2 profit belongs to the entry mechanics, not to the RB. A stop order that is filled exactly at a level that price has just re-crossed produces a positive gross result on random levels too. On M1 bid data, this is most likely a fill-model effect (no slippage on stop orders), possibly with some very short-term momentum. It is largest where stops are smallest (M15), and costs remove it completely.
• One small RB-specific effect. Trend-aligned RBs on H1 beat the placebo in both blocks (t = +2.6 and +2.3, about +0.05R). Out of roughly 70 comparisons, this is close to what chance would produce, and the net expectancy is still −0.28R to −0.32R.

3.4 Claim ④ — The daily chart

The daily chart, where the lecture’s examples are drawn, gives only 152–214 trend-aligned trades per block. The best variant — E2 with the prior swing-low body as the target — earned +0.064R (2012–2018) and +0.141R (2019–2026) per trade after costs. However, E2 on random levels scored almost the same (t vs placebo = +0.05 and +1.0). With this sample size, a real edge of this size cannot be told apart from noise.

4. Corrections made during this study

5. Conclusion

• ① Long-wick swings beat random by about 0.5–2.6 pp in win rate, but the expectancy difference is not significant on H1/H4, and the only significant M15 result did not replicate. Net of costs, negative everywhere.
• ② The full lecture definition with a trend filter does not help the limit entry; it is no better than random.
• ③ The confirmation entry looks profitable before costs, but random levels traded with the same rules look just as good. The rejection block itself adds about +0.05R at most (H1, trend-aligned), which costs remove.
• ④ The daily chart is the only place with a positive net result, but the sample is too small and the placebo performs similarly.

The 2012–2018 block covers the years in which the ICT Mentorship Core Content was taught, and it gives the same answer as 2019–2026. So the result is not “it worked then and stopped working”: within what can be defined mechanically, the rejection block shows no tradable edge in either period.

6. Limitations

• Each element was tested with one mechanical definition. The lecture describes context (long-term resistance, a bearish order block or old orders nearby) that is not modelled here.
• The trend definition (major swings of ten candles on each side) is one of many possible choices.
• Candles are UTC-aligned; broker H4 and daily candles are shifted and produce somewhat different swings.
• The random baseline does not control for time of day. Events overlap in time and are not independent, so the t-values are somewhat optimistic.
• Stop orders are filled exactly at the level with no slippage, which flatters E2 (Section 3.3). Results are on bid prices only; costs are representative estimates.
• BTCUSD covers May 2017 onwards only; in 2012–2018 it adds about 1.5 years.

7. Next Steps

• Restrict RBs to major swings and to prior-day / prior-week highs and lows, where the lecture places them
• Add a higher-timeframe PD array filter (the RB near a bearish order block or long-term resistance)
• Re-run the daily test when more data, or more instruments, are available

Appendix — Technical notes: Swing detection and trade resolution are JIT-compiled with Numba and run on M1 data for exact fill, stop and target ordering. Swings: fractal of two candles per side, confirmed at the close of the second candle after the swing. Trend: major swings of ten candles per side on the same timeframe. Random baselines: one random entry per trade, same block, same direction, same stop and target distance (v2: same trend state). Placebo: one random-level E2 trade per RB. Costs: EURUSD 1.4 pips, GBPUSD 2.2 pips, USDJPY 2.5 pips, AUDUSD 1.6 pips, XAUUSD $0.58, BTCUSD $30, US100 1.5 points round trip. Full methodology and scripts available on request.