How a disorder-regime model split even on LUNR and still trailed the stock

LUNR was a violently mobile name across the window, and the interest is in what a model that waits for stillness does with that kind of tape: it caught a few strong moves and was wrong just as often, yet its winners ran far enough to carry the book. What follows replays those fixed rules across LUNR's saved history; it is a backtest, not a forecast or a recommendation, and the live sample is still too small to grade.

The model measures restlessness in a monitored LUNR reading, how unsettled and erratic it has become versus its own history, and it moves only once that restlessness compresses to an extreme, a fleeting spell of stillness. It runs long only; all 34 of its positions were buys taken in those rare quiet spells. The rule split its outcomes evenly, winning exactly half of its 34 trades, and still posted a flat-stake total of +270.91% because the winners were the larger moves.

How the model is built, end to end.

The reading the model lives by is how restless or still a monitored LUNR series has become. When the series thrashes and shifts unpredictably, the restlessness measure runs high and the model holds off; when that same series settles into an unusually quiet, even spell, the measure drops toward an extreme. It is scaled against how this stock normally moves, so a level reads as still only against LUNR's own history. An average reading keeps the model flat; only a sharp compression of restlessness opens a long.

A still reading is a permission slip, not a command. The model still waits for its direction check to agree before it acts, so it does not buy every quiet spell that appears. When the two line up it places a long with a stop and a target set in advance. From there the trade resolves in one of a few ways, and on LUNR the target and the stop both pulled real weight: profit targets closed the most positions, with stop-outs close behind, a sign of how sharply this name can turn after a calm stretch.

The illustration above carries the model across LUNR in a single line: the restlessness reading, the stillness threshold it has to reach, and the long that results. No part of it is intricate; the character is patience, waiting for the thrash to compress and the direction check to agree before any capital commits. The model's lineage belongs in the picture as well: it started as one option inside an automated search and lasted only by clearing backtest and walk-forward checks, the path from idea to deployment sitting behind each number on this page.

At a glance

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

Walk-forward verification

A trade walked through

The walked example is a long held about ten days. The restlessness reading compressed into a quiet spell near 10 dollars, the direction check agreed, and the model bought; it rode the recovery and closed near 13 dollars for +30.01%. It is the model at its best, a patient entry on a still extreme followed by a clean exit once the move played out.

What the full trade record shows

Across its 34 LUNR trades the model won 17 and lost 17, an even split. The exits leaned on the target: 14 reached the profit target, 11 were stopped out, and 9 closed when the calm reading wore off.

A target-and-stop split this balanced is the signature of a model that enters on stillness and is right about as often as it is wrong: the still spells that hold run to the target, while the ones that snap get stopped out fast. The book pays off not on hit rate but on the size of the winners relative to the losers.

The biggest winner was a long that ran to its target for +59.13% over about two weeks off a low base in September 2024. The fastest winner reached its target for +31.88% in roughly a day, the kind of sharp recovery this name can stage out of a quiet spell.

The worst trade was a long stopped out for -37.66% in under a day in March 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, but with the wins and losses split evenly, the book makes plain that this model earns its return from the magnitude of its best moves, not from being right more often than not.

How does this compare to just holding LUNR

Over the same window the model was tested on, simply buying LUNR and holding it would have returned more. Placing the two next to each other shows whether the rule earned its keep or merely tagged along on a strong tape. Here it tagged along, capturing a useful but partial share, and the tiles below put numbers on the gap.

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 built 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 LUNR will keep offering the model the same still spells it found across the test window.

In live trading the model has been quiet so far, with too few signals to compare against the backtest. Until more live trades build up, the backtest is the only evidence there is, 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 losses spring from the same place as its trades, the quiet spells. Its sharpest failure is a stillness that snaps: the reading compresses, the model buys, and LUNR collapses anyway, which is how the deepest trade lost 37.66% on a long stopped out in March 2025 in under a day. Because it is long every time, it has no defense against a quiet spell that breaks hard the wrong way, and a name this mobile can fall faster than the reading ever priced in. Its quieter cost is selectivity, every restless stretch it judged too unsettled to trade was a move it sat out.

Three things are worth watching if this ever trades at size. The first is the regime mismatch itself: the model reads calm and goes long, but on a violently mobile name the calm windows are short and fragile, so a stillness that snaps can hand the model a fast, large loss, which is exactly how its deepest trade lost 37.66% in under a day. The second is that waiting for calm keeps it on the sidelines through the biggest directional moves, capping the upside even when the direction call is right. The third is execution on a thin, jumpy name, where the gap between a backtest fill and a live fill is widest precisely during the disorder the model is trying to dodge.

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