Part 4: Compressions, spikes and the sheet · Chapter 13 of 14 · 9 min
The volatility sheet
A single page gathers the whole book: ATR and its position over the last 500 candles, regime, expected range of the day with the band, stop in ATR, size, alert threshold for spikes. On the last candle of the data (6 October 2026, 10:00 UTC) the ATR is $13.5 (16th percentile: compression), the regime is low (10-day mean range $78.5), the expected range of the day sits between $41 and $124, the 1.5 ATR stop is $20.2 and with $10,000 and 1% risk the size is 4.94 ounces. The «high ATR» filter on the plan has 3 trades in the second half: no conclusion.
What it must contain
A book on volatility is worth it if a procedure you actually use remains at the end. The volatility sheet is a page you fill in at each session (or each morning) with eight items. Each has a number, the rule that follows from it and the chapter it comes from.
| # | Item | How it is calculated | Written rule | Ch. |
|---|---|---|---|---|
| 1 | ATR | mean of the 14 hourly true ranges, in $ and in % of the price | Written in dollars and in percent at each session | 2 |
| 2 | ATR position | rank of the current ATR among the last 500 candles (0-100) | Below 20 compression, above 80 expansion | 11 |
| 3 | Regime | mean range of the last 10 days against the thresholds (terciles of the past) | Low, medium or high; thresholds rewritten every month | 4 |
| 4 | Expected range of the day | mean of the previous 14 days, with the band from 0.5 to 1.5 times | Stops and targets stay inside the band; outside the band (17% of days) another choice is made | 5, 7 |
| 5 | Stop | k × ATR, with k chosen from the holding time and the noise accepted | Never below 1 ATR without a written proof | 8 |
| 6 | Size | risk in dollars ÷ stop in dollars, recalculated at every entry | The risk is written as a share of the account, not the quantity | 9 |
| 7 | Spike alert | true range of a candle above 3 × ATR | For three hours, wider stop or no entry | 12 |
| 8 | After a day beyond ±2% | next day's range expected +40% | ATR, band and size are recalculated before the open | 12 |
An example: the last candle of the data
The sheet filled in at 10:00 UTC on 6 October 2026 (last candle of our data), with a $10,000 account and 1% risk:
| # | Item | Value |
|---|---|---|
| 1 | 14-candle ATR | $13.49 (0.324% of the price) |
| 2 | ATR position | 16th percentile of the last 500 candles: compression |
| 3 | Regime | 10-day mean range $78.5: low (threshold $95.4) |
| 4 | Expected range of the day | 14-day mean $82.9, band from $41 to $124 |
| 5 | Stop at 1.5 ATR | $20.2 |
| 6 | Size | risk $100 ÷ $20.2 = 4.94 ounces |
| 7 | Spike alert | true range above 3 × 13.49 = $40.5 |
| 8 | Yesterday beyond ±2%? | (datum to check at each session) |
Read the chart in words
Hourly gold chart with, below, true range bars and the ATR line near the 20th percentile threshold, and two dashed stop lines above and below the last price.
| When | What | Price |
|---|---|---|
| 5 Oct 2026 05:00 UTC | stop of a buy: 1.5 ATR | 4,141.74 |
| 5 Oct 2026 05:00 UTC | stop of a sell: 1.5 ATR | 4,182.21 |
Times in UTC.
What this sheet says. The ATR is low, in compression: so the movement over the next 24 hours is 6.7 ATR on median, that is about $90 ($66 to $129 in the central half), in line with the band of the day's range. A 1.5 ATR stop ($20.2) is, relative to this movement, tighter than usual: chapter 11 says that in compression it pays to look at the band and not at the multiple. With the same 1% risk, with the ATR at the 20th percentile ($14.7) you open 4.53 ounces, with the ATR at the 80th ($26.1) 2.56: the size adapts.
There is no indication of direction. The sheet says what is likely for the size of the movement, not for the direction.
The filter we would have liked: «high ATR» on the plan
Chapter 10 left a question open: in our example plan, do the entries with high ATR do better? Here we make the only test that book 12's rule allows: write the rule, and look at it separately in the first part (65% of the candles, up to 24 June) and in the second (35%, after).
| First part | Second part | |
|---|---|---|
| Plan, all trades | 55 trades, +0.84 ± 0.52 R, +46.5 R | 24 trades, +0.48 ± 0.76 R, +11.6 R |
| ATR above the 80th percentile | 15 trades, +1.39 ± 1.03 R, +20.9 R | 3 trades, +1.66 ± 2.61 R, +5.0 R |
| The rest | 40 trades, +0.64 ± 0.60 R, +25.6 R | 21 trades, +0.32 ± 0.78 R, +6.7 R |
In the first part high ATR does better, in the second too, but with three trades: the error (±2.61 R) is three times the mean. The effect has the same sign in both halves, which is what we would expect if it were true; it is also what we would expect if the second part were little more than noise. No conclusion is possible. The three size rules we tried on the plan end like this:
| Rule (on the 79 trades) | Final account | Drawdown |
|---|---|---|
| Fixed 1% risk | 176% | 5.9% |
| Halved with high ATR | 155% | 5.9% |
| Skipping entries with high ATR | 136% | 5.9% |
Every reduced row took away gain, because in this sample the trades with high ATR were the best. The sheet therefore contains no «high ATR» filter: no proof, in the data or against chance (chapter 10), justifies it. If you want to test it, the only honest way is to write it now, wait for new data and judge it on those (book 12).
What the book did NOT find
- No volatility filter that improves the rules (480 rules: −0.024 against −0.019 for chance).
- No direction signal in the spike (+0.09 ± 0.31 ATR at one hour).
- No accumulated energy in compression: the expansion after calm is mostly mean reversion (1.22 against 1.31 for chance).
- No very precise range estimate: typical error 27%, correlation 0.33.
And what it found that is solid: volatility clusters (chapter 3), grows with the square root of time (7), is measured well with the range (1), has fat tails (6) and must be taken into account in the size (9).
What to do
- Fill in the sheet at every session: eight lines, five minutes.
- Rewrite the thresholds every month on past data.
- Choose the stop in ATR and calculate the size from the risk at every entry, not once.
- Do not put filters in the sheet without proof: none of our filters is proven.
- Mark the spike and the day beyond ±2% in the journal (book 13).
Procedure in four steps
- Calculate the session's ATR, position, regime and expected range.
- Choose the stop in ATR from the holding time and calculate the size.
- Check the alerts (spike, day beyond ±2%).
- Update the thresholds every month and count how often the real range left the band.
Exercises
Chapter exercises
1. The ATR is $13.49. How wide is a 1.5 ATR stop?
2. The current ATR is at the 16th percentile of the last 500 candles. What condition are we in?
3. With ATR $13.49, from which true range of a candle does the spike alert (3 ATR) trigger?
4. The «high ATR» filter has, in the second half of the plan, 3 trades with +1.66 ± 2.61 R. What do you conclude?
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.