Pullback Depth Statistics: How Far Do Retracements Usually Go?
Pullback entries depend on judgments about how far a retracement will go before the trend resumes. Measuring it removes some guesswork.
Take a series of trends on your chart and measure the pullback after each impulse as a percentage of the impulse. Collect at least thirty cases. You may find that many shallow pullbacks retrace about 30 to 50 percent, and a smaller group goes deeper.
Common retracement levels, 38.2, 50, and 61.8 percent, are widely watched, but the actual distribution on a given pair and timeframe may differ. Measuring it gives data for your own market.
Several insights can emerge. If most pullbacks stop before 50 percent, waiting for 61.8 percent will often mean missing trades. If many pullbacks go deeper than 61.8 percent, tight stops beyond the 50 percent level will be hit frequently.
Depth also provides information about trend strength. Shallow pullbacks suggest strong momentum, and deeper ones suggest weakening. A pullback that exceeds roughly 100 percent of the prior impulse is no longer a pullback, but a reversal.
Use the findings to place entries and stops. For instance, if 80 percent of pullbacks end within 50 percent, you might enter near the 40 to 50 percent area with a stop beyond the 62 percent level.
Practical step: measure the retracement percentage after twenty impulses on your chart, and compute the median.
Pullback Depth Statistics: How Far Do Retracements Usually Go?
Re: Pullback Depth Statistics: How Far Do Retracements Usually Go?
Measuring your own pair is the key bit. Textbook fib levels are a starting point, but some pairs and timeframes pull back shallower or deeper as a habit. A simple way to collect the data is to mark the last thirty or forty clear impulses on your 1 hour chart and note the pullback depth for each as a percentage. Put them in a column and sort it. You'll quickly see where most of them cluster. On some pairs that might be 35 to 50 percent, on others deeper. Then set your entry zone around the common depth and your stop beyond the depth that most valid pullbacks didn't exceed. It's tedious once, then useful for months.
It’s Fairman 
Re: Pullback Depth Statistics: How Far Do Retracements Usually Go?
Using the median instead of the average is a nice touch. One or two very deep pullbacks can drag the average down and make you wait for entries that rarely come. It's also worth splitting the sample by trend strength. Pullbacks in a strong trend, where the impulse broke a clear level with big candles, are often shallower than pullbacks in a slow grinding trend. If you only use one number for both, you'll miss entries in strong trends and enter too early in weak ones. Two medians, one for each type, are usually enough to improve the entries without overcomplicating the plan.
It’s Fairman 
-
FXS Prop Desk
- Posts: 1693
- Joined: Wed Oct 07, 2026 9:52 am
Re: Pullback Depth Statistics: How Far Do Retracements Usually Go?
Measure pullback depth separately for different market conditions, such as trending days and ranging days. Pullbacks in strong trends are often shallower, so a single distribution may not fit all conditions.
Recording pullback depth for each trade in the journal builds a personal dataset over time.
Collecting at least 30 to 50 cases per condition gives a more reliable picture than a few examples.
Some traders use the measured distribution to place limit entries at the median pullback depth, with a stop beyond the deeper levels.
Recording pullback depth for each trade in the journal builds a personal dataset over time.
Collecting at least 30 to 50 cases per condition gives a more reliable picture than a few examples.
Some traders use the measured distribution to place limit entries at the median pullback depth, with a stop beyond the deeper levels.
FXS Official Team | The Idea Lab & Strategy
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Pullback Depth Statistics: How Far Do Retracements Usually Go?
The conditional numbers are what decide the stop, more than the overall median. Once a pullback has already passed 50 percent, the question becomes how often it keeps going past the stop level.
Say you measure 40 impulses and get a median depth of 45 percent, with 24 of the 40 stopping short of 50. In the chart (illustration), 16 went past 50, and 9 of those 16 also went past 61.8, so a stop just beyond 62 is hit more than half the time once price gets that far.
That changes how an entry at 45 with a stop at 63 looks. The entry fills on most pullbacks, but the ones that fill and keep going are the losers, and they are not rare.
One fix is to scale the entry, with part at 45 and part at 55, so the deeper pullbacks that do turn get a better average. Another is a wider stop beyond 80 with a smaller size for the same dollar risk.
Only 2 of the 40 made up cases went past 80, though the smaller size means the winners pay less. Do you measure depth from the impulse close or from the wick extreme? On fast H1 impulses the two can differ by several percent, which shifts the whole distribution.
Say you measure 40 impulses and get a median depth of 45 percent, with 24 of the 40 stopping short of 50. In the chart (illustration), 16 went past 50, and 9 of those 16 also went past 61.8, so a stop just beyond 62 is hit more than half the time once price gets that far.
That changes how an entry at 45 with a stop at 63 looks. The entry fills on most pullbacks, but the ones that fill and keep going are the losers, and they are not rare.
One fix is to scale the entry, with part at 45 and part at 55, so the deeper pullbacks that do turn get a better average. Another is a wider stop beyond 80 with a smaller size for the same dollar risk.
Only 2 of the 40 made up cases went past 80, though the smaller size means the winners pay less. Do you measure depth from the impulse close or from the wick extreme? On fast H1 impulses the two can differ by several percent, which shifts the whole distribution.
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Pullback Depth Statistics: How Far Do Retracements Usually Go?
Mixing impulses of very different sizes in one median can blur the picture, because part of every pullback is just ordinary noise. That noise is roughly fixed in pips, so it weighs far more on small impulses than on big ones.
Say routine M5 wiggle on a pair is about 5 pips. A 12 pip impulse that gives back only that wiggle shows a retracement of about 42 percent, while a 40 pip impulse with the same 5 pip wobble shows 12.5 percent.
Neither number says much about trend strength, yet both go into the same column. A median built mostly from small impulses will look deeper than one built from large ones, even on the same pair in the same week.
Expressing each impulse as a multiple of the average bar range and splitting the sample into two or three size groups fixes most of this. The FXS Prop Desk point about trending and ranging days still applies inside each group.
A minimum impulse size is the simpler fix. Ignoring anything under about three average bars keeps the measured depth about structure rather than spread and wiggle.
Where do people set that cutoff? Too high and the sample shrinks below the 30 cases suggested in the opening post very quickly.
Say routine M5 wiggle on a pair is about 5 pips. A 12 pip impulse that gives back only that wiggle shows a retracement of about 42 percent, while a 40 pip impulse with the same 5 pip wobble shows 12.5 percent.
Neither number says much about trend strength, yet both go into the same column. A median built mostly from small impulses will look deeper than one built from large ones, even on the same pair in the same week.
Expressing each impulse as a multiple of the average bar range and splitting the sample into two or three size groups fixes most of this. The FXS Prop Desk point about trending and ranging days still applies inside each group.
A minimum impulse size is the simpler fix. Ignoring anything under about three average bars keeps the measured depth about structure rather than spread and wiggle.
Where do people set that cutoff? Too high and the sample shrinks below the 30 cases suggested in the opening post very quickly.
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: Pullback Depth Statistics: How Far Do Retracements Usually Go?
When price breaks a level where many stops are clustered, the triggered orders add fuel to the move. Stop-loss orders become market orders, pushing price further, which triggers more stops. The result is a cascade: a sharp, fast move that overshoots.
Some market participants may deliberately push price toward stop clusters to trigger the cascade, called momentum ignition. Regulators treat deliberate manipulation as illegal in regulated markets, though proving intent is hard, and spot forex has weaker oversight.
For a scalper, the cascade is a feature of price behaviour regardless of its cause. Obvious levels, such as round numbers and recent extremes, attract stops, and breaks of them can produce fast moves.
Trading cascades is a challenge. Entering early requires a guess about whether the level will break. Entering late means chasing an overextended move that may reverse sharply as soon as the stops are exhausted.
A cautious approach is to study what happens after cascades: how far they typically run, and how often they snap back. Measuring this on recorded data gives a realistic view.
A deeper examination looks at the minute-by-minute path after a break: how far price travels before the first retracement, how deep the retracement is, and whether the original break level holds as support. The distributions give a sense of what "normal" looks like for a cascade, and what an abnormal one means.
Stops placed just beyond obvious levels are the most exposed. Moving them a few pips further away, or placing them at less obvious prices, may help, but wider stops require smaller size, so the trade-off needs to be calculated.
One structured way to study cascades is building an event list. Define a trigger: price breaks the high of the previous hour by at least one pip, with the one-minute range at least three times the average. For each event, record the maximum extension in the first five minutes, the size of the first pullback, whether the pullback holds above the broken level, and the position 30 minutes later.
Grouping the events by session shows differences. Cascades in the London open tend to be larger but also more likely to retrace partially. Cascades in late New York often fizzle quickly because liquidity is thin. Cascades following data releases behave differently from those without a catalyst.
The aim isn't to find a magic rule. It's to replace folklore with distributions. If 60% of events retrace more than half within ten minutes, a fade strategy has something to work with, subject to costs. If only 30% do, chasing may be the better approach. The data decides, and the answer may differ by pair.
What's been observed after stop cascades, and is it better to trade them or avoid them?
Some market participants may deliberately push price toward stop clusters to trigger the cascade, called momentum ignition. Regulators treat deliberate manipulation as illegal in regulated markets, though proving intent is hard, and spot forex has weaker oversight.
For a scalper, the cascade is a feature of price behaviour regardless of its cause. Obvious levels, such as round numbers and recent extremes, attract stops, and breaks of them can produce fast moves.
Trading cascades is a challenge. Entering early requires a guess about whether the level will break. Entering late means chasing an overextended move that may reverse sharply as soon as the stops are exhausted.
A cautious approach is to study what happens after cascades: how far they typically run, and how often they snap back. Measuring this on recorded data gives a realistic view.
A deeper examination looks at the minute-by-minute path after a break: how far price travels before the first retracement, how deep the retracement is, and whether the original break level holds as support. The distributions give a sense of what "normal" looks like for a cascade, and what an abnormal one means.
Stops placed just beyond obvious levels are the most exposed. Moving them a few pips further away, or placing them at less obvious prices, may help, but wider stops require smaller size, so the trade-off needs to be calculated.
One structured way to study cascades is building an event list. Define a trigger: price breaks the high of the previous hour by at least one pip, with the one-minute range at least three times the average. For each event, record the maximum extension in the first five minutes, the size of the first pullback, whether the pullback holds above the broken level, and the position 30 minutes later.
Grouping the events by session shows differences. Cascades in the London open tend to be larger but also more likely to retrace partially. Cascades in late New York often fizzle quickly because liquidity is thin. Cascades following data releases behave differently from those without a catalyst.
The aim isn't to find a magic rule. It's to replace folklore with distributions. If 60% of events retrace more than half within ten minutes, a fade strategy has something to work with, subject to costs. If only 30% do, chasing may be the better approach. The data decides, and the answer may differ by pair.
What's been observed after stop cascades, and is it better to trade them or avoid them?