How a tail-weight rule won three of four trades yet trailed SNDK badly

SNDK was the standout mover of the set, a run of more than fifty-six fold over the window, and the backtest asks the natural question: even at a 75% hit rate, how much of a move that violent could a selective rule keep. What follows replays SNDK's stored history through the model's fixed rules as a study of those rules, not a forecast or a recommendation, and with only one year and twenty trades the live sample is far too small to grade.

The condition this model waits for is SNDK's recent returns departing from a bell-shaped spread, the excess kurtosis that builds when a distribution sheds its ordinary rounded form and develops heavy edges. It stands aside while the spread stays close to a bell and engages only after that departure mounts to an unusual peak, going long where the deviation resolves upward and short where it resolves downward. The up read held nineteen of its twenty times. Winners came in at 75% across the 20 trades, the highest hit rate of this group, and the running total reached +366.94% on an even-sized book.

How the model is built, end to end.

The reading the model is built on is how far SNDK's recent returns have strayed from a bell-shaped spread, a fourth-moment measure of excess kurtosis that rises as the distribution grows heavy-edged and peaked and falls when it settles back toward an ordinary rounded form. The model sizes that stray against the level SNDK itself usually shows, marking a reading extreme only by the stock's own benchmark. A near-bell reading keeps the model out; nothing short of a real stray from the rounded form will clear an entry.

A pronounced departure earns the model's attention, nothing more. It will not commit unless a companion directional check confirms the same way, a safeguard against trading on one freak session. Confirmation in hand, it stages a bracketed entry, protective floor and objective both pinned ahead of time. The exit then arrives by one of several routes, and for SNDK the route taken by a clear margin was the objective being reached, the mark of a model that, once its read on the distribution was right, was usually right in a hurry.

Picture the model laid out across SNDK from end to end: first the bell-departure reading, then the gate standing in its way, then the trade it yields. Nothing in that chain is ornate; its governing instinct is to hold back until the spread breaks from its rounded form and the companion direction check confirms before a dollar is put at risk. The rule set's lineage rounds out the picture, for it was drawn from a wide field of machine-proposed candidates and held on through nothing but its backtest and walk-forward results, the road from idea to deployment beneath every figure here.

At a glance

These numbers come from a backtest of the model on SNDK, 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 five days. The tail-weight reading climbed to an upward extreme near 206 dollars, the direction read agreed, the model bought, and the position reached its profit target near 267 dollars for +30.06%. It is the model at its best, a patient entry on a heavy reading and a quick exit at the target as the move ran in its favor.

What the full trade record shows

Across its 20 SNDK trades the model won 15 and lost 5. The exits leaned on the target: 12 reached the profit target, 5 were stopped out, and 3 closed once the reading settled back to ordinary.

A target-heavy split like this is the mark of a model that, when its read was right, tended to hit its objective fast rather than wait the trade out. Stops still took a quarter of the book, the early entries that did not work, while only a handful drifted to a signal-based close.

The biggest winner was a long that ran to its target for +38.17% over about nine days in January 2026. The book's winners mostly closed at their targets, the pattern of a model catching heavy-tailed stretches and booking them quickly while the move was hot.

The worst trade was a long stopped out for -10.11% over roughly four days across late November into December 2025; that single loss is also the equity curve's deepest drawdown, and it is one reason the drawdown gate is a bar the backtest missed.

No single trade carries the record, and the book itself is only twenty trades over one year. With SNDK up more than fifty-six fold, the lesson is the scale of what a selective rule leaves behind, profitable on its own terms yet a fraction of an extraordinary move.

How does this compare to just holding SNDK

Over the same window the model was tested on, simply buying SNDK and holding it would have done vastly better. Setting the two side by side is how you judge whether the rule earned its keep or merely tagged along, and here it tagged along on an exceptional tape, banking a profitable slice while leaving the bulk of the run alone. The tiles below put numbers on how large the gap was.

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 never fit on. It is a consistency and replay-integrity test, not proof the model will make money live. A clean reproduction means the deployed rules act like the studied ones; it says nothing about whether SNDK will keep concentrating weight in its tails the way it did across the test window.

In live trading the model has been quiet so far, with too few signals to set against the backtest, and the backtest itself spans only a single year of twenty trades. Until more trades accumulate, that thin record 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 come from the same place as its trades, the tails. Its sharpest one is being early: weight piles into the tails, the model buys, and SNDK pulls back, which is how the deepest trade lost 10.11% on a long stopped out over late November into December 2025. Because it is long almost always, it has little defense against a sharp dip, and that single stop-out set the equity curve's worst stretch. Its quieter cost is selectivity itself, the long stretches of an extraordinary run it sat out entirely.

Three things are worth watching if this ever trades at size. The first is the sample, since a one-year book of twenty trades is thin and missed the size bar the gates set, so the win rate may not hold up. The second is the long-heavy exposure, because the model has little way to profit from or hedge a sustained decline. The third is the gap between live and backtested behavior, the first sign the tail-weight reading is no longer marking the same moves.

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