Three-Bar Reversal

What It Does

Three-Bar Reversal runs on a single contract at a single intraday or daily timeframe. On every bar close it inspects the last four bars and asks the shared three-bar predicate one question: did the three bars ending the prior bar all push the same direction, and did the current bar firmly reject that direction?

The predicate has two flavours:

• Bull-to-bear: three bars in a row each printed a higher high and a higher low, then the current bar printed both a lower high and a lower low. Treated as upside exhaustion. The strategy goes short at the close.
• Bear-to-bull: the mirror. Three lower-high lower-low bars, then the current bar snaps back with a higher high and a higher low. Treated as downside exhaustion. The strategy goes long at the close.

When the strategy is flat and a signal fires that the side gate permits, it enters at the close. From there, there are two ways to exit:

• After hold_bars bars have elapsed since entry, flatten unconditionally.
• Optionally, if exit_on_opposite is on, cut early when the opposite reversal fires.

The time-based exit takes priority. If both happen on the same bar, the hold-bars exit wins.

There are no stops, no profit targets, no trend filters, no scaling. The setup-and-hold is the entire strategy.

Timeframe Matters

The pattern is timeframe-sensitive. On daily bars in trending equity indices it tends to lose; the indices keep trending past each reversal signal. On 1-hour bars the pattern fires often enough to produce a meaningful sample and is roughly break-even to slightly positive over the 2020-2024 window across NQ, ES, and GC. On 5m / 15m the noise overwhelms the signal. The canonical sample for this family runs on 1H bars, which is where the pattern earns its reputation. The form lets you pick any timeframe from 15m up to 1D.

Why The Pattern Gets Talked About

Three-bar reversals show up in just about every classical technical-analysis text. The reasoning is intuitive: three same-direction bars suggest a one-way crowd, and a strong counter-bar suggests the crowd ran out of fuel. Whether that intuition translates into a tradeable edge is another question entirely, and one this strategy is designed to answer honestly.

The pattern is rare. On a daily NQ chart the trigger fires only a few dozen times per year, so the sample size in a five-year backtest is modest. The variance of outcomes per trigger is high, and the bar-count exit is blunt. The presets here are starting points for testing the hypothesis, not a finished system.

Presets

Five presets ship out of the box:

• Both Sides 5-Bar Hold: the default. Fades both directions, holds for five sessions, ignores opposite signals while in position.
• Long Only 5-Bar Hold: takes only bear-to-bull reversals. Useful when an upward drift assumption rules out the short side.
• Short Only 5-Bar Hold: the bearish mirror. Often the harder side to make work in equity indices.
• Both Sides 10-Bar Hold: doubles the holding period to two full trading weeks. Tests whether the post-reversal drift extends past the first week.
• Both Sides Opposite Exit: holds up to ten sessions but cuts early on the opposite signal. Lets the pattern itself decide when the trade is no longer working.

Use the form to set your own side, hold length, and opposite-exit behaviour; the presets are starting points, not endpoints.

Best In

• Markets where genuine sentiment exhaustion shows up on daily bars. Liquid equity index futures and large-cap commodities are the most-studied candidates.
• Studies that want a parameter-light baseline before adding filters. With only four inputs and one of them just a contract count, the setup is hard to curve-fit.
• Comparing the same trigger across instruments. Three samples ship for NQ, ES, and GC at identical defaults so the family page reads as an instrument comparison.

Where It Struggles

• Strong, persistent trends. A bull-to-bear signal inside a roaring uptrend often gets run over within the hold window.
• Low-signal regimes. The pattern is strict, and quiet markets can go entire months without firing on either side.
• Single-bar overshoots. The strategy reacts to the reversal bar at its close, so a sharp intraday spike that closes back into range can still trigger an entry that the next session reverses again.

Possible Uses

• A control strategy for any reversal idea that adds extra filters. If a fancier reversal setup tracks the bare three-bar pattern, the filters are not earning their keep.
• An honest stress test of the chart-folklore claim. Run the presets across instruments and years and see how stable the edge is.
• A starting point for layered work. Combine the trigger with a regime filter (trend direction, realized volatility band, time-of-year) and see whether the conditional edge tightens.

What It Does Not Do

• No stops, no targets, no trailing exit. The bar-count hold is the only guaranteed exit; the opposite-signal exit is opt-in.
• No partial fills, no scaling, no pyramiding. One signal, one position, one contract count.
• No intraday detail. The strategy reads daily highs and lows only; intraday wicks that recover before the close are invisible to the predicate.
• No awareness of the prior trade’s outcome. A losing streak does not change the next entry size or the next hold length.