How a volatility-regime model rode ARM both ways and still trailed the stock

A volatility-regime rule met ARM's strong run by trading both directions, banked a real return, and still finished behind the shares. This is a backtest of those fixed rules replayed across ARM's own saved history, not a forecast or a recommendation, with a live sample still too thin to grade.

Volatility, not price, decides whether this model engages. It gauges how active ARM has recently been and places that activity against the stock's longer record of percentile readings; the rule arms only once the level settles into a favored slice of that span, and stays idle whenever the tape turns calmer or more frantic. With the gate open it leaned whichever way the setup pointed, splitting its book into 18 long entries and 15 short, and 60.61% of those 33 positions resolved as winners, adding to +208.92% at a single repeated stake.

How the model is built, end to end.

The reading the model lives by is ARM's recent volatility, placed against its own past. When the tape is in one of its quieter or its most frantic stretches, the reading sits outside the favored band and the model stays out. When volatility settles into the middle slice the rule prefers, the gate opens and a position can form. Because the band is defined by ARM's own percentile history, a level only counts as favorable relative to how this stock normally behaves, never against some fixed outside yardstick.

A favorable volatility regime is permission, not a position. The model still needs a direction read to agree before it commits, which is how a volatility filter ends up trading both long and short rather than always leaning one way. When the regime and the direction line up it opens with a stop and a target set in advance. After that the trade ends in one of three places, and on ARM the settling reading did the most work: more positions closed when volatility drifted out of the favored band than on either the target or the stop.

The illustration above walks the model across ARM in a single line: the volatility reading, the percentile band it must sit inside, and the long or short it produces. None of the pieces is exotic; the discipline is in standing aside until volatility is in the regime the rule was built for. The model's lineage belongs in the frame too. It began as one candidate among many in an automated idea search and earned its place only by clearing backtest and walk-forward checks, the whole route from idea to deployment standing behind every number here.

At a glance

These figures come from a backtest of the model on ARM, scored against fixed acceptance gates, not from a live track record.

Walk-forward verification

A trade walked through

The walked example is a short held about a week. With ARM's volatility inside the favored band and the direction read pointing down, the model sold near 423 dollars; the slide ran its way and it covered near 324 dollars for +23.28%. It is the model working as designed on the bearish side, a regime-permitted entry and a clean exit once the move delivered.

What the full trade record shows

Across its 33 ARM trades the model won 20 and lost 13. The exits leaned on the settling reading: 16 closed when volatility drifted out of the favored band, 10 reached the profit target, and 7 were stopped out.

A split this heavy on signal exits is the mark of a regime model: most positions are closed not by hitting a hard level but by the volatility regime that justified them fading away, with the target and the stop sharing the remainder.

The biggest winner was a long that ran to its target for +29.56% over about four days in April 2026, opened while volatility sat squarely in the favored band. The patient winners tended to close on the settling reading, taking what the regime offered before it lapsed.

The worst trade was a long stopped out for -18.85% over roughly eight days in early April 2025; that single loss is also the equity curve's deepest drawdown, and it is the reason the drawdown gate is the one that bites hardest.

No single trade carries the record, and with ARM up close to threefold over the window, the lesson of the book is the part of that climb the model sat out whenever volatility wandered outside its favored regime.

How does this compare to just holding ARM

Over the same window the model was tested on, simply buying ARM and holding would have returned more. Setting the two side by side shows whether the rule earned its keep or merely came along on a strong tape. The honest read is the latter, a useful but partial share, and the tiles below put numbers on the distance.

How well does the model reproduce its tape?

Walk-forward verification checks whether the saved rule path reproduces the expected signal behavior on held-out timestamps it was not fitted on. It is a consistency and replay-integrity test, not proof the model will earn money live. A clean reproduction means the deployed rules behave like the studied ones; it says nothing about whether ARM's volatility will keep visiting the favored band the way it did across the test window.

In live trading the model has stayed quiet so far, with too few signals to line up against the backtest. Until more live trades accumulate, the backtest is the only evidence on hand, and it should be read as exactly that, a study of how the rules behaved on saved history.

When this approach fails

The model's failures grow from the same soil as its trades, the regime band. Its sharpest one is misreading the side: volatility sits in the favored slice, the model goes long, and ARM keeps falling, which is exactly how the deepest trade lost 18.85% on a long stopped out in April 2025. Because it commits inside an active regime, a single sustained move against the position can run further than the stop allows. Its quieter cost is selectivity, every stretch where volatility fell outside the band was a move the model never touched.

Three things are worth watching if this ever trades at size. The first is drawdown: the worst stretch in the backtest was about 18.89%, deep enough to break the drawdown gate. The second is two-sided exposure, because trading both long and short means a wrong regime call can lose on either side. The third is the gap between live and backtested behavior, the first hint the volatility band is no longer marking the same regimes it did in the study.

Risk and honest limits

On this run the model's automated checks logged a caution rather than a clean pass. It is one more reason to read everything here as backtest evidence about the rules, not a verdict on the stock.

Lifecycle