Double bottom chart pattern: exact rules and how to test them
A double bottom becomes testable only when every part of it has a number. Write the rules before you look at results, include costs, and check whether the pattern beats a simple breakout on several markets and timeframes.
By Soren Huang · Published · updated · Editorial policy
Contents
- What is a double bottom pattern?
- How do you turn a double bottom into rules a computer can test?
- Where should you enter, place the stop and set the target?
- Why do traders disagree about which double bottoms are real?
- How do you test a double bottom across markets and timeframes?
- Frequently asked questions
- References
What is a double bottom pattern?
A double bottom is two similar lows after a decline, with a peak between them called the neckline. It is complete when price closes above the neckline. To test it, turn each part into a number, such as how close the lows are and how far apart. Then backtest it with costs and compare it with a simple breakout.
The pattern has five parts. Each part becomes a rule later, so it helps to name them clearly:
- A prior decline. Price falls first. Without a fall, two lows are only the bottom of a sideways range.
- The first low. Selling stops and price bounces.
- The neckline. The neckline is the highest price between the two lows, usually drawn as a horizontal line. Some traders draw a sloped line, but a horizontal line is easier to define and test.
- The second low. Price falls back to about the same level as the first low and bounces again.
- The breakout. A breakout is a close above the neckline after the second low has formed.
In the traditional reading, the second low shows that sellers could not push price lower. The breakout shows that buyers lifted price above the last bounce. That is a story about the chart, not evidence. Evidence comes from testing a fixed definition. Traders often call the shape a "W" pattern, and the same rules apply. A double bottom is read from the chart rather than calculated, which sets it apart from most technical indicators and is why it needs exact rules first.
How do you turn a double bottom into rules a computer can test?
You turn a double bottom into rules by giving every part of the drawing a number, so that two traders and a computer find exactly the same patterns. "Two lows at about the same level" is an opinion. "Two swing lows within half an average bar range of each other" is a rule.
A swing low is a bar whose low is lower than the lows of a set number of bars before it and after it. It is the usual way to define "a low" without looking at a chart. Average True Range (ATR) is the average size of a bar, including gaps between bars. Use it to measure distances. The ATR indicator page explains how it is calculated. Measuring in ATR instead of points lets one definition work on a slow daily chart and a fast intraday chart. It also lets one definition work on markets with very different prices.
| Part | What you must decide |
|---|---|
| Prior decline | How far price must fall before the first low, and over how many bars |
| Swing low | How many bars on each side must have higher lows |
| Similar lows | The largest allowed difference between the two lows, in ATR |
| Separation | The minimum and maximum number of bars between the lows |
| Neckline height | The minimum rise from the lower low up to the neckline |
| Breakout | A close above the neckline, with or without a small buffer |
| Time limit | How many bars after the second low the breakout may take before the pattern is cancelled |
| Cancellation | A close below the lower low before the breakout ends the pattern |
One timing detail decides whether the test is honest. A swing low with three bars on each side is only known three bars after it happens. Look-ahead bias is the error of using information in a backtest that was not available at the time of the trade. If your code marks the second low on the bar where it occurs, the backtest uses the future.
Hypothetical example
A complete definition on 60-minute bars might read: the market fell at least 4 ATR in the 40 bars before the first low. A swing low has higher lows on the 3 bars before it and the 3 bars after it. The second low is within 0.5 ATR of the first. The lows are 8 to 60 bars apart. The neckline is at least 2 ATR above the lower low. The signal is a close at least 0.1 ATR above the neckline within 30 bars after the second low. A close below the lower low first cancels the pattern. ATR uses 14 bars. These settings are a starting point for testing, not a recommendation.
Where should you enter, place the stop and set the target?
Each entry, stop and target choice trades a higher win rate against a smaller payoff, so test them as separate variants instead of picking one from a chart book. Expectancy is the average result per trade, after wins and losses are combined. Compare variants by their expectancy. The win rate and expectancy calculator shows how win rate and the size of wins and losses combine.
| Choice | Option | What it trades off |
|---|---|---|
| Entry | Next bar after the breakout close | Clear signal; entry is far above the lows |
| Entry | Retest of the neckline after the breakout | Better price; some breakouts never come back, so you miss them |
| Entry | Bounce from the second low, before the breakout | Best price; the pattern is not complete yet, so it fails more often |
| Stop | Below the lower of the two lows | Wide stop; few stop-outs, large loss when hit |
| Stop | Halfway between the lows and the neckline, or a fixed number of ATR | Smaller loss; hit more often by normal noise |
| Target | Measured move from the neckline | Classic target; can be smaller than the risk |
| Target | Fixed multiple of the risk, or a trailing stop | Payoff is set by you; win rate changes with it |
The measured move is the classic target. It projects the height of the pattern upward from the neckline:
Target = Neckline + (Neckline − Lowest low)
Here "Neckline" is the neckline price and "Lowest low" is the lower of the two lows. The difference between them is the height of the pattern, and the target adds that height once more above the neckline.
Hypothetical example
The lows are at 100 and 101, and the neckline is at 110. A bar closes at 111, which is the signal. With next-bar execution, the entry is the next bar's open at 111.5. The measured move target is 110 + (110 − 100) = 120, so the possible gain is 8.5 points. A stop at the lower low, 100, risks 11.5 points. To break even before costs, the trade must win about 58 of every 100 trades (11.5 ÷ 20). A stop at 105 risks only 6.5 points and needs about 43 wins in 100 (6.5 ÷ 15). However, that tighter stop is hit more often.
Costs widen this gap. Every entry and exit pays commission and slippage, and intraday patterns have small moves per trade. The guide to trading costs in backtests shows how to compare the average trade with the round-trip cost.
Why do traders disagree about which double bottoms are real?
Traders disagree because a pattern drawn by eye is chosen after the fact, and only a fixed rule removes that choice. On a finished chart, the eye picks the two lows that led to a rally. It ignores the similar pairs of lows that led to a further fall. Examples in books and posts are usually successful patterns, because failed ones do not make good pictures.
A false breakout is a close above the neckline that soon reverses back below it. False breakouts are where a double bottom loses most of its money. Traders use several filters against them. Each filter removes some false breakouts and also delays or removes some good ones:
| Filter | Rule | What it costs |
|---|---|---|
| Close, not touch | A bar must close above the neckline; a wick above the line does not count | Entry waits for the bar close instead of the first touch |
| Buffer | The close must be at least a fraction of ATR above the neckline | Later entry, at a higher price |
| Follow-through | A second close above the neckline, or entry only on a retest that holds | Later entry; strong breakouts that never retest are missed |
| Volume condition | Higher volume on the breakout bar | One more parameter to test |
| Time limit | Cancel the pattern if the breakout takes too many bars after the second low | Slow patterns that would have worked are removed |
Decide which filters to test before you see any results. If you add filters one by one until the equity curve looks good, you are fitting the past. The article on overfitting in strategy generation explains why that inflates results.
How do you test a double bottom across markets and timeframes?
You test a double bottom by running one fixed definition on several markets and timeframes, with costs, and comparing it with a simpler entry. A strict definition on daily bars of one market may produce only a few patterns a year. That is too few to judge. Testing several markets also shows whether the result depends on one market's history. One fixed W definition can run on several markets and bar sizes in the backtesting of AlphaPrime (Valynox), which applies the same fees and slippage to each run.
- Write the full definition first. Write down every part from the rules table, plus the entry, stop and target variants.
- Choose markets and timeframes in advance. For example, choose stock index futures, TAIEX futures and gold on 15-minute, 60-minute and daily bars. Keep all distances in ATR so that one definition fits all of them.
- Include costs and next-bar execution. The signal comes from a closed bar, so the order fills on the next bar.
- Count the trades. Treat any market and timeframe with fewer than about 30 trades as a description of the past, not evidence.
- Build a baseline. Use the same exits, but enter on a plain close above the highest high of the last N bars. If the double bottom does not beat this simple breakout, its extra conditions add nothing.
- Check neighbouring settings. For example, try a tolerance of 0.3, 0.5 and 0.7 ATR. The results should change gradually, not collapse.
- Test on unseen data. Keep a later period out of the design, or use walk-forward analysis to re-tune and test forward window by window.
| Check | Encouraging | Discouraging |
|---|---|---|
| Trades per market and timeframe | About 30 or more | A handful |
| Versus the simple breakout baseline | Better after costs on most markets | Similar or worse |
| Neighbouring settings | Similar results | Only one exact setting works |
| Out-of-sample period | Keeps part of the in-sample result | Turns negative |
| Average trade versus round-trip cost | Several times the cost | Close to the cost |
| Across markets | Same direction on most markets | One market carries the total |
Every variant you try increases the chance that the best one is luck. Sullivan, Timmermann and White (1999) studied technical trading rules with the bootstrap method that White (2000) published as the Reality Check. They showed that when many rules are tried on the same data, the best one can look strong by chance alone. A fair test therefore counts every variant you tried, not only the winner. Keep a record of how many definitions, filters and exit combinations you tested.
A double bottom definition is a set of entry and exit conditions with parameter ranges. That is the input strategy generation in AlphaPrime works with. It turns research ideas into testable trading rules with parameter ranges, entry and exit conditions, and search constraints.
What this means if you use AlphaPrime: write your double bottom as exact entry and exit conditions with ATR-based ranges, and let AlphaPrime evaluate every variant with fees, slippage and next-bar execution. Then keep only the versions that beat a simple breakout on several markets and on out-of-sample data.
Frequently asked questions
Is a double bottom a bullish pattern?
Yes. Traders read a double bottom as a bullish reversal signal after a decline. The pattern is only complete when price closes above the neckline, the high between the two lows. Before that close, the two lows can still be part of a further fall.
How far apart should the two lows of a double bottom be?
There is no fixed rule. Set a minimum and a maximum number of bars between the lows, and test a few ranges. Lows that are very close together are often noise, and lows that are very far apart may not belong to the same pattern.
What is the best timeframe for double bottom trading?
No timeframe is best in general, so test the same definition on several. Shorter bars give more patterns and more trades, but costs take a larger share of each small move. Daily bars give larger moves but fewer patterns, so you may need several markets to collect enough trades.
How reliable is the double bottom pattern?
The answer depends on the exact definition, the market, the timeframe and the costs, so no single success rate applies to every chart. Many published examples were chosen after they worked. Write a fixed definition, test it with costs, and compare it with a simple breakout entry to see what the pattern adds.
What is the difference between a double bottom and a W pattern?
They are the same pattern. "W" describes the shape of two lows with a peak between them. The same rules for the lows, the neckline and the breakout apply to both names.
References
- Sullivan, R., Timmermann, A., & White, H. (1999). Data-Snooping, Technical Trading Rule Performance, and the Bootstrap. The Journal of Finance, 54(5), 1647–1691.
- White, H. (2000). A Reality Check for Data Snooping. Econometrica, 68(5), 1097–1126.