Pullback & Retest

Pullback and retest strategies share one premise: instead of entering at the moment a market moves, they wait for it to move *back* before joining in. A trend makes a leg up, stalls, gives part of it back, then resumes. A price level breaks, and the market returns to touch it from the other side. In both cases the entry comes after the initial move has already happened — at a price that is worse in momentum terms but better in location terms.

That trade-off is the whole concept. Entering on the breakout gets you in while the move is alive, but at the extreme of a range and usually with a distant stop. Waiting for the pullback gets a tighter, better-defined invalidation point, at the cost of missing every move that never comes back. Each version catalogued on this page is, in some form, an answer to how to price that trade-off.

## The three components

Almost every pullback system decomposes into three decisions.

**Context** — which direction is being traded, and on what evidence. This can be a moving-average relationship, a sequence of higher highs and higher lows, a higher-timeframe bias, a broken level, or a session assumption. Context is what decides whether the current move counts as a pullback at all; without it, a pullback and a reversal look identical.

**The pullback itself** — how far, for how long, and to where. Some versions measure depth as a retracement ratio; others require price to reach a specific structure — the broken level, a moving average, a prior swing, a supply or demand zone. Some cap the duration of the correction, on the reasoning that a slow, drawn-out pullback signals fading interest in the original direction.

**The trigger** — what turns the zone into a fill. Two families dominate: passive (a resting limit order at the level, accepting that price may run straight through it) and confirmatory (waiting for a candle pattern, a momentum cross, or a break of the pullback's own micro-structure). Passive entries get better prices and more losers; confirmatory entries get fewer, later and more expensive fills.

## Main variants

**Break and retest** is level-based: something breaks — horizontal support/resistance, a trendline, a range edge, a session high — and the entry is the return to that level. **Trend pullback** is direction-based: no break is required, only a correction inside an established trend, often anchored to a moving average. **First pullback** narrows this to the first correction after an impulse or breakout, on the argument that later pullbacks in the same leg are lower quality. **Retracement-depth** versions filter by how much was given back, treating shallow and deep corrections as different setups. **Failed-breakout** variants invert the logic: a break that immediately fails and closes back inside becomes an entry in the opposite direction. And **generated or optimised** versions hand the parameter choices to a search process rather than fixing them by hand.

## What differentiates implementations

Two systems can both be "buy the pullback in an uptrend" and behave nothing alike. The differences that matter most: how strictly the trend is defined, whether the entry is a limit or a confirmation, where the stop sits relative to the pullback extreme, whether the exit is a fixed multiple, a structural target or a volatility-based trail, and which trades are excluded outright by session, volatility or instrument filters. Exits often explain more of the difference between two versions than the entry rule everyone argues about.

## Common mistakes

Defining the trend after seeing the pullback is the most frequent error, and it is invisible on a chart replay — the rule has to be computable at the moment of entry. Second is having no explicit rule for when a pullback has become a reversal; "it stopped being a pullback" needs a price, not a feeling. Third is assuming a limit order at a level fills at that level, ignoring spread, slippage and the possibility that the fill only happened because price kept going. Fourth is tuning retracement depth to a handful of screenshots. Fifth is underestimating how rare a fully qualified setup is on a single instrument and timeframe.

## Evaluating a version

Start by writing every discretionary word out of the rules — "clean level", "strong trend", "healthy pullback" must become numbers, or the test is not reproducible. Be explicit about fill assumptions and intrabar sequencing, since pullback entries are unusually sensitive to whether the entry or the stop was touched first inside a bar; lower-timeframe data or deliberately pessimistic assumptions help here. Test out-of-sample, and test neighbouring parameter values: a rule that only works at exactly the 61.8% level and breaks at 55% or 70% is describing the sample, not the market. Include realistic costs, and check the trade count is large enough to support any conclusion.

The most informative test is comparative. Run the same exits with a simpler entry — buy the breakout directly, or enter in the same direction at an arbitrary bar. If the pullback filter does not improve the result, the edge lives in the exits or the trend filter, not in the pullback. That comparison is worth running against every decoded version linked below.

Strategies in this concept (69)

Frequently asked questions

What is the difference between a pullback and a retest?

A pullback is a correction against the prevailing direction inside a trend that is already underway — no specific level has to be involved. A retest is a return to a level that has just been broken, approached from the new side. Most catalogued strategies use one or the other as their reference point, and some require both at once: a pullback that happens to land on a previously broken level.

How deep should a pullback be before entering?

There is no universal depth, and treating one as universal is a common source of overfitting. Depth is best treated as a parameter to be tested rather than assumed: shallow retracements imply strong continuation but give a wide stop relative to the entry, while deep retracements give a tighter stop but a higher chance the trend has actually ended. What matters is that whichever depth rule is used, it is fixed before the test and checked at neighbouring values.

Is it better to enter with a limit order at the level or wait for confirmation?

They are different strategies with different failure modes, not better and worse versions of one. A limit order gets the best available price and a clear invalidation, but fills on every approach including the ones that continue straight through. A confirmation entry filters out some of those, at the cost of a worse price, later entry, and missing the setups that reverse instantly from the level. Backtesting both with identical exits is usually the fastest way to see which one the instrument rewards.

How do you know when a pullback has turned into a reversal?

Only by defining it in advance. Usable definitions are structural — price closing beyond the origin of the impulse, the swing point that defined the trend being broken, or the retracement exceeding a stated fraction of the prior leg. Any system without such a rule is relying on the stop-loss to answer the question after the fact, which makes its historical results dependent on stop placement rather than on the pattern.

Why do pullback strategies often look better on charts than in a backtest?

Because charts are read backwards. Once the trend and the resumption are both visible, the pullback is obvious; at the moment of entry it was one of several candidates, most of which did not resume. Backtests also surface things eyeballing hides: fills that assumed liquidity at a level, intrabar ordering of entry and stop, spread and slippage costs on frequent small-stop trades, and how few fully qualified setups actually occur per instrument per year.

Do pullback strategies transfer across markets and timeframes?

The logic is generic, but the parameters usually are not. Trend definitions, pullback depth and stop distances all scale with the volatility and session structure of the instrument, so a version tuned on one forex pair or one intraday timeframe rarely transfers unchanged. Testing the same rule set on several instruments and timeframes with the same parameters is a useful robustness check: broad, moderate results tend to survive better than a single excellent fit.

All strategy concepts · Explore all strategies