CHoCH and BOS: direction, pivot confirmation and signal timing

CHoCH names a selected structural break against the preceding direction. To reproduce that label, state which pivot counts, how a breach is measured and when the pivot was known. A historical marker does not establish that the signal was available during the marked candle, that institutions changed position or that price will reverse.

What a change of character (ChoCh) actually is

In this article’s convention, a downward break of a recorded higher low after an upward structure is bearish CHoCH; an upward break of a lower high after a downward structure is bullish CHoCH. The label names the break direction, not the preceding trend. LuxAlgo explicitly notes varying CHoCH/MSS and protected-swing conventions. The model below is our operational definition, not every SMC indicator’s algorithm.

For our exercise, a pivot low must be strictly lower than the lows of two completed candles on each side; a pivot high mirrors this with highs. Equal values fail. The pivot becomes known only when the second right-hand candle closes. Two bars is an invented teaching parameter, not a demonstrated optimum. A later marker drawn on the pivot’s bar must not be treated as knowledge available then.

How ChoCh differs from a break of structure (BOS)

BOS names a breach in the direction already assigned by the chosen structural rule, while CHoCH names an opposing breach. A label does not confirm future continuation or reversal. In our close-based variant, the final Close must be strictly beyond the watched level, not equal to it. A wick-based variant is another valid definition to compare, rather than universally a mistake.

Count each level’s break once. After CHoCH, suspend the old directional state until a consistent confirmed high/low sequence establishes the next state under our rule. A further break in that assigned direction can then be BOS. Do not automatically label every lower close after one CHoCH as another signal, or call the next arbitrary low a confirmed new trend.

All omitted candles are assumed to satisfy the stated pivot comparisons; the table is a selected-event exercise rather than a complete OHLC data set or trade record. Lows 2 and 8 became known after bars 4 and 10, and highs 5 and 11 after bars 7 and 13. Subsequent closes below 1.1040 do not create another break of that same level. A return above it remains possible.

How to spot a ChoCh on the chart, step by step

Keep the detection interval and pivot rule explicit. If another timeframe supplies a filter, record its final-close availability separately. The same market can have different local and broader structures without contradictory data. The broader technical analysis material provides context; higher-timeframe agreement is an extra condition to test rather than a compulsory source of profit.

Change only the pivot window in a separate fictional comparison. Bar 8 Low = 1.1040; bars 6, 7, 9 and 10 each have Low = 1.1050. Bar 8 passes our two-per-side pivot rule and is known after bar 10. Set bar 3 Low = 1.1010: a five-per-side window includes that lower left-hand value and rejects bar 8. Assume an already confirmed broader low at 1.1000, not derived from this short excerpt. Close 1.1030 breaks the local 1.1040 by 10 pips but remains 30 pips above the broader level. Our two/five-bar rules illustrate different structures, not LuxAlgo’s proprietary sensitivity algorithm.

A retest is an additional observation rule. It may not occur, may cross the zone without reacting, or may change the assumed entry and risk. It does not automatically offer better net reward:risk. Define price basis, allowed orders, entry timing and the chosen stop loss separately; an ordinary stop can execute worse. A structure label alone does not supply those trade rules.

Why ChoCh is an interpretation, not a certainty

Discretionary swing choices can differ; fully specified pivot comparisons and thresholds can be reproduced. A close relative to a recorded level is observable even though choosing that level is a modelling decision. Record the scale and confirmation time before comparing conclusions. The separate H1/H4 example tests close timing only, not the complete qualification of CHoCH.

Candles do not reveal whose stops were placed below a level or whether someone deliberately targeted them. An order block (a defined price zone) or a retest can be an added filter, without proving institutional orders or improved performance. The mechanics of Smart Money Concepts explains this difference between observable prices and interpretation. Compare the whole rule, including rejected and failed cases.

The most common mistakes when trading ChoCh

• Changing a stated close condition into a wick condition after seeing the later outcome.
• Labelling a break without an explicit prior direction or pivot-selection rule.
• Using a retrospectively plotted pivot before its actual confirmation time.
• Treating repeated closes beyond one level as separate breaks without a re-arming rule.
• Forgetting that the break alone is only a warning signal, not proof that a new trend has already begun.

ChoCh against classic reversal patterns

Compare CHoCH with the head and shoulders or a double-top rule by writing each pattern’s selection, availability and confirmation conditions. Different names do not establish which produces earlier useful entries or fewer false alarms. A trend-following approach also needs defined exits, costs and later-data evaluation; structural similarity is not a performance comparison.

What to do tomorrow

1. Fix one price series, interval and pivot window. Record the pivot bar and its separate confirmation bar.
2. Define the starting directional state, watched level, strict/equal boundary convention and close-based or wick-based breach.
3. Classify sequential eligible events without looking ahead, including mixed states and no-signal cases. A practice count does not prove readiness for real money.
4. If adding a breaker block or retest, state it as a separate filter and compare the version with and without it under the same cost assumptions.
5. Keep one event identifier per consumed level, actual detection time and order availability. Evaluate unchanged rules on later data rather than infer an edge from twenty selected examples.