How a kurtosis model traded DELL to a near dead heat with the stock

DELL's kurtosis rule did something selective rules rarely manage over this window: it finished in a near dead heat with the stock rather than trailing it, though the slim nominal lead is more an artifact of how flat-sized trades are counted than real edge. What follows replays that fixed rule set over saved DELL history as a study of the rules, not a forecast or a recommendation, and the live sample is still too thin to grade.

This model is a tail-risk reader. It tracks how prone DELL's recent moves have been to sudden, outsized jumps, the fat tails that surface when a stock starts trading in bursts, and it commits only when that tendency climbs to an unusual high. A burst that builds upward turns it long, one that builds downward turns it short. Over about two years it took 32 positions, was right on 59 percent of them, and finished at +200.92% on a flat trade size, a total, rare for this kind of rule, lands neck-and-neck with simply owning the stock.

How the model is built, end to end.

What the model actually inspects is the spread of DELL's recent returns, and above all how fat the tails of that spread have grown, a fourth-moment reading that climbs when big outlier moves start crowding in and sinks when the tape is calm. It judges that reading against where it has typically sat for this stock, so a value is extreme only relative to DELL's own normal. A middling reading leaves the model on the sidelines; only a real surge in tail-heaviness opens the door to a position.

A surge on its own is not enough. Before the model picks a side, a direction check has to point the same way, which stops it from trading every burst of noise. When the two agree, it sends an order with a stop-loss and a target already attached. A position then tends to end one of three ways: it reaches the target, it trips the stop, or the tail surge fades and the model closes out because the reason for the trade is gone, which is why targets, stops, and signal exits each take roughly a third of the book.

The illustration above follows the model end to end on DELL: the tail-shape reading on the left, the gate it must clear in the middle, and the position that results on the right. No single step is clever on its own; the discipline is in staying out until the tails stretch far enough and the direction check agrees. The rule set also carries a lineage, from an idea-search candidate through backtest and walk-forward verification to the packaged model traded here, the chain from idea to deployment that every figure on this page rests on.

At a glance

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

Walk-forward verification

A trade walked through

The walked example is a long that ran for about three weeks. The tail shape stretched to an upward extreme near 128 dollars, the gate agreed, and the model bought; it then held through a shallow dip before the move it expected arrived, and closed at its profit target near 158 dollars for +23.66%. It is the design working as intended: enter on a shape extreme, sit through the noise, exit at the target.

What the full trade record shows

Across its 32 DELL trades the model won 19 and lost 13. The exits split almost evenly three ways: 11 closed at the profit target, 11 were stopped out, and 10 closed when the tail signal relaxed and released them.

That near-even three-way split is the fingerprint of a tail-shape system trading both directions: it opens on a heavy-tailed extreme and is about as likely to be carried to a target, knocked out at a stop, or released when the shape calms back to normal.

The biggest winner was a long that ran to its target for +36.73% in about three days off a sharp upward stretch in May 2026. A signal-release short, held about a month, added +3.48% before the tail-shape reading let it go, the opposite end of the same playbook.

The worst trade was a long stopped out for -11.48% over six days in November 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; the edge is thin and spread across many trades, which is why the win rate sits just under sixty percent and the average trade earns about six percent. Holding DELL outright would have returned almost the same, with none of the trading.

How does this compare to just holding DELL

Over the same window the model was tested on, simply buying DELL and holding it was its own kind of strategy. Comparing the two shows whether the rule earned its keep or just rode the tide. In DELL's case the answer is close to a draw, and the tiles below show how narrow 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 not built on. It is a consistency and replay-integrity test, not proof the model will make money live. A clean reproduction means the deployed rules behave like the studied rules; it says nothing about whether DELL will keep handing the model the same tail-shape extremes to trade.

In live trading the model has been quiet so far, with too few signals to line up against the backtest. Until more live trades accumulate, 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 failures are specific to trading tail shape. Its sharpest one is the gap: when DELL jumps clean over the protective stop between sessions, what should have been a small loss becomes a larger one, which is exactly how the deepest trade in the book lost 11.5% on a long stopped out in November 2025. A choppy stretch where the tails never quite reach an extreme leaves the model flat for weeks, missing moves it never voted on. And because winners and losers land in a roughly even split, a short run of stop-outs can erase several ordinary wins at once. None of this is a defect to patch; it is the cost of waiting for a shape extreme before committing.

Three things are worth watching if this ever trades at size. The first is drawdown: the worst stretch in the backtest was about 11.5%, deep enough that a hard pause rule matters. The second is the even balance of targets and stops, because a breakout in the stop rate is the earliest sign the tail signal has stopped marking real moves. The third is the distance between live and backtested behavior, which is the first place a decaying edge shows up.

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