Volatility breakout

Level 6 — strategy families.

The claim

A move far larger than this instrument's normal daily range is the start of something, not the end of it.

The close cousin of trend following, with one important difference: a trend follower waits for a price level to be exceeded, a volatility breakout waits for a thrust — a move large relative to how much this instrument normally moves.

That relativity is the whole idea. A $3 move is enormous in a utility and noise in a high-priced tech name. Measuring the move in ATR — average true range, the typical distance between a bar's high and low, averaged over a lookback — makes "unusually large" mean the same thing on any instrument at any price.

Why anyone believes it

Volatility clusters. This is among the most robust facts in financial data: large moves are followed by large moves, and quiet periods are followed by quiet periods. Whatever caused today's thrust has not finished happening.

A thrust is evidence of forced activity. Ordinary two-sided trading produces ordinary ranges. A day three times the normal range usually means somebody had to transact regardless of price — a fund liquidating, an index rebalancing, a position being covered. Those flows take time to complete.

It fires before a level is exceeded. A channel rule needs price to reach a forty-day high. A thrust rule can fire on the day the move happens, from somewhere in the middle of a range. Earlier entry, and more false ones.

Say the cost

Volatility clustering is reliable. Volatility clustering with a direction is not. The evidence that large moves follow large moves is much stronger than the evidence that they continue the same way, and this family lives entirely on the weaker half of that claim.

What it looks like

A single bar that dwarfs its neighbours, often with a gap. On a chart it is the most visually obvious of all the families, which is part of why it attracts beginners and part of why it is crowded.

volatility_breakout takes two numbers:

Parameter	Means
atr_period	how many bars the average true range is measured over
entry_atr_multiple	how many ATRs the move must be to count as a thrust

flowchart TD
    B["Bar closes"] --> A["Measure the move<br/>in ATRs"]
    A --> Q{"Move ≥<br/>entry_atr_multiple<br/>× ATR?"}
    Q -->|yes| G["Ask the risk gate"]
    Q -->|no| W["Wait"]

    style G fill:#1f6feb,stroke:#1f6feb,color:#fff

The rule cannot act until it has atr_period bars — there is no average true range before then, and Arvo will not manufacture one.

When it fails

The gap is the loss. A thrust often happens overnight, so the move you were waiting for is already in the price when you can act. You are buying after the information, at a worse price, with your stop now further away.

Spreads widen exactly when it fires. Every cost you pay is at its worst on a high-volatility day. This family is the most cost-sensitive of the six, and conservative costs eliminate more breakout findings than any other kind.

Reversal days look identical to continuation days. A capitulation low and a breakdown are the same bar. The rule cannot distinguish them, and the difference in outcome is total.

Too low a multiple turns it into noise. At 1.0 ATR you are firing on ordinary bars several times a week. At 3.0 you may get four signals a year, which is not a sample. The parameter trades signal quality against having any evidence at all, and both ends fail.

Test it

rulesets/thrust.json:

{
  "name": "thrust",
  "label": "Volatility breakout",
  "premise": "A close moving more than two ATRs continues far enough to pay for the reversals.",
  "interval": { "step": 1, "unit": "day" },
  "kind": {
    "kind": "grid",
    "rule": "volatility_breakout",
    "fixed": { "trade_size": 10.0, "atr_period": 14.0 },
    "axes": { "entry_atr_multiple": [1.5, 2.0, 2.5] }
  }
}

Three configurations, with atr_period pinned at 14 rather than searched — pin what you are not asking about, because every axis you add raises the deflation bar the winner must clear.

Research view → pick thrust → pick an instrument → Run. Then:

1. Check the trade count first. At 2.0 ATR on two years of daily bars you may get a dozen trades. Under thirty, the verdict will be Inconclusive and the advice will tell you to widen the window rather than loosen the entry.
2. Read the cost tiers. If the finding is Supported at stated costs and fails at conservative, the advice says Supported only under the stated costs — the edge belonged to the cost assumption, not the rule. For this family, expect it.
3. Check the regime split. Breakouts that only worked in one direction are common, and a rule that only works in trending-up markets is a bet on conditions, honestly labelled.

Watch

• (to be chosen) — ATR explained and computed by hand.
• (to be chosen) — volatility clustering, shown on a real series.

Next

Mean reversion — the opposite claim.