Order Block: How to Spot One and Whether They Actually Work

A plain-language guide with live examples and backtest data across BTC, ETH, and SOL.

Jay Sharma · Founder, Botsfolio

Published August 18, 2026

What it is

An order block is a single candle on a chart. It is always the last opposing-color candle immediately before a strong move that breaks the recent range. In a downtrend that suddenly reverses upward, the last down candle before the rally is a bullish order block. In an uptrend that flips down, the last up candle before the drop is a bearish order block. The idea, popularized by Michael J. Huddleston through the Inner Circle Trader framework in the early 2010s, is that this candle marks the price zone where a large participant absorbed the previous side and initiated the new direction. When price returns to that same zone later, unfilled interest tends to still sit there, and the reaction often repeats.

Two things people commonly get wrong about the concept, which this guide will not:

• An order block is a zone, not a level. The whole candle body plus wick is the zone. But the actual reaction inside that zone almost always happens at a much narrower price band, which is where a properly tuned detector adds resolution. More on that below.
• Not every last-opposing candle is a valid order block. The impulse leaving it must actually break structure. Without a real break of the previous high or low, you have a rejection candle, not an order block.

How to spot one on a chart

Order blocks form in three phases. All three should be visible before treating the zone as valid.

1. The previous swing and the break

Start with a clean recent swing. In an uptrend, that means a well-defined higher low. In a downtrend, a well-defined lower high. Then wait for price to break the opposing side. A bullish order block only becomes valid after price breaks above the most recent lower high. If price never breaks structure, you do not have an order block yet, you have a candle that might become one later.

2. The formation candle

The formation candle is the last opposite-color candle before the impulse. For a bullish order block, that means the last red candle before the up-move. Its body plus wick defines the zone. Wide-range formation candles produce weaker, less reliable order blocks. Tight-body formation candles with long wicks produce sharper ones, because the wicks show where absorption actually happened rather than where the body opened and closed.

3. The refined reaction level

Inside the formation candle's zone, most reactions cluster at a specific narrow price band, not evenly across the whole zone. Botsfolio's detector measures this by clustering the wick tips of the surrounding candles on subsequent tests. On BTC 4H, the average order block zone spans roughly 300 to 600 dollars in absolute terms. The actual reaction band inside it typically averages 30 to 60 dollars. That refined level is the one worth watching. The whole zone provides context. The refined level provides the trigger.

4. Validation and invalidation

Once formed, the order block is either valid or invalidated, never both at once. A 4-hour close below the refined level of a bullish order block invalidates it and flips the structural read to bearish. A 4-hour close above the refined level after a return-to-zone reaction confirms the reclaim. Between those two states, price is testing the zone.

How Botsfolio detects and measures it

Every candle close, our analysis engine runs an SMC detector suite over the last several hundred candles per coin. When a valid order block forms, three pieces of data are recorded together, all derivable only after the impulse candle closes:

• The full zone, taken from the formation candle's body plus wick, in dollars and percentage of price.
• The refined level, the tight band inside the zone where surrounding wicks cluster on retests. Typically 5 to 15 percent of the zone width.
• The invalidation threshold, defined as a candle close through the refined level plus a small buffer.

Then the reaction is tracked continuously: reclaimed, rejected, still testing, or broke. Every outcome becomes a row in our backtest, feeding the numbers you see below.

Two things worth naming, because they matter for how you read the numbers:

• We only surface order blocks that meet a minimum sample threshold in the backtest, at least 20 historical instances at the same (coin, timeframe, direction) grain. Below that we call it thin data and treat it separately.
• Our cost model assumes 0.12 percent round-trip fees, on the higher end of retail perp fee schedules. Real fees vary by exchange and fee tier. The numbers below already have this deducted from expectancy R.

Full methodology at our methodology page.

A live example on BTC right now

Here is a live order block on BTC, drawn as it looked when it formed, alongside how the pattern has performed on BTC across timeframes. If nothing is currently active, the widget shows the most recent formed example and its outcome.

Historical performance across coins and timeframes

The table below is aggregate performance of order block setups across the coins we backfill, at timeframes with enough sample to be meaningful. Book-early management means partial off at first target with the remainder trailed. Both directions combined.

Reading the table honestly, three observations:

Higher timeframes materially outperform lower ones. On BTC, the 6H and 1D bands show noticeably higher win rate and expectancy than 1H. The same pattern holds on ETH and SOL. Order blocks are a structural read, and structure is more meaningful when derived from more information per bar.

Sample size drops fast as you go up in timeframe. The 1D rows have far fewer instances than the 1H rows. A 65 percent win rate on 34 instances is directionally informative but not the same certainty as a 47 percent win rate on 428 instances. Below 20 instances we grey the row and label it thin. Read those rows with caution, or wait for more data.

The best-performing coin, timeframe, and management combination in a given period is often not the best combination in the next period. This is the honest thing to say about backtested edges: they are historical facts, not forecasts. What the numbers argue for is watching only the setups that have sample-supported edge, and treating each one as a probabilistic event rather than a confident call.

What tends to invalidate an order block

Every backtested setup carries a reversal rate: the percentage of instances that reached +1R at some point, then finished at or below breakeven. This is one of the most useful failure metrics in trading, because it isolates setups that worked and then did not.

Three failure patterns account for the vast majority of order block invalidations in our data:

High volatility regime. When the trailing 14-period ATR exceeds roughly twice its trailing 50-period average, order block reliability degrades sharply. Wicks pierce the refined level frequently, and price often continues through the whole zone rather than reacting inside it. Our detector still fires in these regimes, but the historical stats degrade materially. A setup that would be a fine read in normal volatility is a much lower-confidence read during a volatility expansion.

Concurrent high-impact macro event. Order blocks that form or get tested within 12 hours of a FOMC decision, CPI print, or NFP release show reduced follow-through. When a setup coincides with a scheduled event, our live setup card surfaces the event in the upcoming events panel so the reader has that context.

Structure break during the test. If price closes through the refined level on a lower timeframe than the setup formed on, aggregate outcome from that point is worse than the base rate. This is why we track a per-setup status of reclaimed, rejected, testing, or broke, and update it every candle close.

How order blocks interact with other patterns

Order blocks are strongest when they stack with other Smart Money Concepts patterns in the same impulse leg. The compounding is real. A single-signal setup and a triple-confluence setup are structurally different opportunities even though both technically qualify as an order block.

• Order block plus fair value gap on the same impulse leg. The FVG acts as a magnet pulling price back to the OB. Historically the highest-confluence long-side setup in our data. See Fair Value Gap.
• Order block plus liquidity sweep. When the impulse that formed the OB also swept a prior swing high or low, the engineered-liquidity thesis is strongest. See Liquidity Sweep.
• Order block within a discount or premium context. A bullish OB in the discount half of the recent range is more reliable than one in the premium half, and vice versa. This is trend-alignment by another name.

Related: an order block that fails becomes a Breaker Block. An order block that price returns to in order to fill leftover interest before continuing is closely related to the Mitigation Block.

How reads of this concept commonly go wrong

Five patterns show up repeatedly when we look at how order blocks get misread. Each is framed as an observation about the data, not a directive.

1. Treating the whole zone as a single price. The zone is context. The refined level is the trigger. Reactions in our data cluster at the tight band inside the zone, not evenly across it. Instances traded off the far edge of the zone show a materially different outcome distribution than instances traded off the refined level.
2. Skipping the structural break requirement. A last-opposing candle before a move that fails to break the prior swing is not an order block. It is a rejection candle. The break of structure is a prerequisite. Instances tagged as order blocks without a confirmed break under-perform base rate by a wide margin.
3. Reading order blocks on very low timeframes. Sub-15-minute order blocks fire so frequently and invalidate so quickly that base-rate outcomes barely differ from noise. We do not publish them for this reason. This is not a claim they cannot work, only that our data does not support a stable edge there.
4. Ignoring higher-timeframe direction. An order block formed against the prevailing higher-timeframe trend shows lower base rates than one formed with the trend. Multi-timeframe alignment is a real factor in the outcome distribution.
5. Sizing off the whole zone width instead of the refined level. If a stop is placed just outside the refined level while position size was calculated assuming a stop at the far edge of the zone, the actual position ends up much larger than intended. Our position sizing tool uses the refined level by default for exactly this reason.

Frequently asked questions

Related concepts

Educational analysis, not financial advice. Past performance does not predict future results.