A short squeeze is the same cascade, running upward

A cascade downward and a squeeze upward are usually told as different stories, one about fear and one about pain. In the arithmetic they are the same story. The price at which a venue closes a position, ignoring fees, is the entry price adjusted by the initial margin less the maintenance margin — and the only thing that distinguishes the two cases is which way that adjustment points.

For a long the closing price is the entry times one minus one over the leverage plus the maintenance margin rate. For a short it is the entry times one plus one over the leverage minus that same rate. At twenty times leverage with a half-percent maintenance margin, a long is closed by a 4.5 percent fall and a short by a 4.5 percent rise. Same distance, opposite direction.

This is not an idealisation. Hyperliquid's published liquidation rules carry a side variable that is one for a long and minus one for a short, and are otherwise identical for both; Binance states the trigger as collateral — initial collateral plus realised and unrealised profit and loss — falling below the maintenance margin, with the same immediate-or-cancel order sent to offload the position either way. Neither venue has a separate rulebook for the upside.

What changes with the sign is which side of the book absorbs the damage. A falling price closes longs, and a long is closed by selling: those forced orders walk down through the resting bids. A rising price closes shorts, and a short is closed by buying it back, so those orders walk up through the resting asks.

The rest of the loop is unchanged and needs nobody's intent. The forced buy consumes ask levels, the price it leaves behind is higher than the price that triggered it, and that higher price is the liquidation threshold of some other short held at a lower entry or a higher multiple. Each forced purchase is simply an input to the condition that produces the next one.

The reason it bites is the same reason it bites downward. Resting depth on a large venue mostly belongs to market makers, and a market maker facing one-directional flow widens or steps back. Forced orders are the most one-directional flow there is, so the order that cannot be delayed arrives exactly when the side it needs has thinned. A squeeze does not require anyone to be buying for a reason. It requires only that somebody has to.

Here the symmetry ends, and it ends in the one place that matters. A falling price has a bound: it can reach zero and no further, so a downward cascade is arithmetically guaranteed to run out of room. A rising price has no such bound. There is no price at which the arithmetic says a rise must stop.

The loss behaves differently too. A long that goes to zero has lost exactly its stake, and that is the end of it. A short that doubles has lost twice its stake, and at four times the entry it has lost four times over — the exposure keeps growing with the price. Liquidation is meant to close the position long before any of that, which is precisely why it is sent as a market order rather than a patient one, and why the venue does not wait for a better fill.

The simulator on this site treats that as a real case rather than an edge case. When a forced order is larger than everything resting on the side it meets, the walk through the book leaves no finite price behind, and the run reports that depth was exhausted instead of printing a number. That is one of five distinct reasons the loop can end, kept separate on purpose: a shock that cleared the book on its own is not a forced order that outgrew what was left, and a side that was empty from the start is not a side that was emptied.

The cascade simulator here has run this direction since it was built, and no page said so. State a book, position buckets with their entries and leverage, a maintenance margin rate, and a shock: a negative shock closes the longs among them and a positive one closes the shorts. Nothing else in the model changes. It observes no venue, reads no open interest and predicts nothing — it answers what the arithmetic does to a scenario you supplied.

Three real effects are absent, and all three make a genuine squeeze smaller than the simulated one. Insurance funds absorb part of it: Binance's futures insurance fund bears the loss from a bankrupt position, and when that is not enough, auto-deleveraging closes bankrupt positions along with opposing ones held by traders who were never in trouble. Hyperliquid sends only twenty percent of a position over 100,000 USDC at first and then waits thirty seconds. And resting orders arrive precisely because the price moved — a refill the model does not invent, because nobody gave it a number for one.