Part 4: Compressions, spikes and the sheet · Chapter 11 of 14 · 8 min
Compression and expansion
After a compression (ATR below the 20th percentile of the last 500 candles) the range of the next 24 candles is 6.7 ATR on median, against 5.0 in average conditions and 4.5 after an expansion; the ATR after 24 candles is 1.22 times the starting one. But a shuffled series without memory gives 1.31: the expansion after calm is largely mean reversion of a noisy measure. A breakout taken only in compression is worth +0.13 ± 0.55 R over 42 trades, like chance (56% of random series do better).
Calm before the storm?
A widespread idea says that after a period of calm the market «charges up» and moves with force. A period of low volatility is called a compression (squeeze) and one in which volatility rises an expansion. Chapter 3 tells us volatility has memory: after calm, usually, calm. Here we see what happens at the end of the calm.
To put a number on compression you need a reference. A low ATR in dollars is not enough: volatility rises and falls over time (chapter 4). You look at the position of the current ATR relative to the last 500 candles (about 22 market days): 0 means the lowest value, 100 the highest.
| Condition | ATR position | Candles (out of 4,216) |
|---|---|---|
| Compression | below the 20th percentile | 964 |
| Normal | between the 20th and the 80th | 2,369 |
| Expansion | above the 80th | 883 |
Read the chart in words
Daily gold chart with, below, a pink line between 0 and 100 with two horizontal lines at 20 and 80.
What happens next
After each candle, in each of the three conditions, we measured the range (high minus low) of the next 24 candles in units of that moment's ATR, and the ratio between the ATR after 24 candles and the starting one.
| Condition | Range of the next 24 candles / ATR (median, 25th-75th) | ATR after 24 candles / ATR now (median) |
|---|---|---|
| Compression | 6.67 (4.88-9.56) | 1.22 |
| Normal | 4.95 (3.90-6.57) | 0.96 |
| Expansion | 4.52 (3.41-6.04) | 0.84 |
After a compression the price covers 6.7 ATR on median in 24 hours, 35% more than in normal conditions: when the ATR is low, the next movement is large relative to it. After an expansion it is the opposite: the ATR falls (0.84) and the next movement is small relative to the starting ATR.
A practical application is this: a 1.5 ATR stop after a compression sits, relative to the expected movement, closer than the number says. With a low ATR it is better not to trust the multiple alone and to look at the band (chapter 7).
Accumulated energy or mean reversion?
There is an objection. The ATR is a noisy measure: when it happens to be low by pure chance, it is more likely to be higher the next time, simply because extreme values tend to go back towards the centre. This is called mean reversion and it holds even for a series with no memory at all. To know how much of the effect is the market's merit and how much the measure's, we repeated the calculation on 60 series in which the candles are shuffled one by one (with no clustering):
| Condition | ATR after 24 candles / ATR now: real data | Shuffled series |
|---|---|---|
| Compression | 1.22 | 1.31 |
| Normal | 0.96 | 1.00 |
| Expansion | 0.84 | 0.73 |
The expansion after compression is also present by chance, and it is stronger (1.31 against 1.22). In the real data it is slower because volatility has memory (chapter 3): a compression lasts. So:
- Compression does not «charge» anything: the rise in the ATR after calm is what a noisy measure going back towards its level would do.
- The market has more inertia than chance: the ATR stays low longer (1.22 against 1.31) and high longer (0.84 against 0.73).
A rule on compression
Let us build the classic rule: 20-candle breakout only when the ATR is in compression (below the 20th percentile). Same exits as the plan: stop at 1.5 ATR, target at 3 R, maximum duration 48 candles, entry at the close.
| Rule | Trades | Mean R ± error |
|---|---|---|
| 20-candle breakout, always | 178 | +0.15 ± 0.26 |
| Only with compression | 42 | +0.13 ± 0.55 |
| Only with expansion | 41 | +0.34 ± 0.57 |
The three rows sit inside each other's error. To know what chance would give, the same rule on the 100 random series (one-day blocks, without trend):
| Mean R | |
|---|---|
| Breakout in compression, random series | +0.14 (from −0.32 to +0.64 in 90% of cases) |
| Real value | +0.13 |
| Random series doing equal or better | 56% |
The real value (+0.13) is in the middle of the band of chance and more than half of the random series do better. Compression adds nothing to the breakout, neither in this data nor against chance. Not even the +0.34 of the expansion stands out (error ±0.57).
What follows
- After calm the price moves more, relative to the ATR: 6.7 ATR against 5.0 in the next 24 hours. It is a useful measure for stops and targets.
- It is not a reason to open a position: it does not say in which direction.
- The expansion after compression is largely mean reversion: the same thing happens in a series without memory.
- A breakout in compression does not do better than chance in this data (+0.13 ± 0.55 over 42 trades).
What to do
- Measure the position of the ATR relative to the last 500 candles, not just the dollar value.
- After a compression, widen the band for stops and targets: the next movement is on average 35% larger in ATR.
- Do not buy «because there is compression»: compression does not say the direction.
- Compare every compression rule with random series: that 56% says chance often does as well.
- Remember the mean reversion of a measure: every low value tends to rise without the market «charging up».
Procedure in four steps
- Calculate the position of the ATR relative to the last 500 candles.
- Measure the range of the next 24 candles in ATR, for the three conditions.
- Repeat on a shuffled series to separate the mean reversion.
- Test the rule against chance before using it.
Exercises
Chapter exercises
1. After a compression the ATR rises to 1.22 times in 24 candles, in a shuffled series without memory to 1.31. What does it mean?
2. After a compression the range of the next 24 candles is 6.7 ATR, after normal conditions 5.0. What does it imply for a 1.5 ATR stop?
3. A breakout taken only in compression makes +0.13 ± 0.55 R over 42 trades; 56% of random series do as well or better. What do you conclude?
4. Why is compression defined with the ATR's position and not with a dollar value?
Have a question or want to share your exercise?
Post in the community, or join the free signals room on Telegram.
Educational content, not financial advice. Trading involves risk.