The launch of a perpetual contract on a stock as an event to trade

The idea: carry the working move of shorting a fresh listing over from coins to perpetual contracts on stocks. It did not hold up: a contract like that has nowhere to detach to, because it is tied to the real price of the stock.

Why it sounded right

Shorting a fresh listing on coins is one of the few moves that survives serious testing. The mechanism there is clear: a new coin has little float, the price in the first hours is set by hype, and after that it gets slowly pressed down by everyone who got tokens cheaper. The event is public, the date is known in advance, and the move fits into two lines of code.

So when a big exchange started stamping out perpetual contracts on stocks in batches, carrying it over looked almost mechanical. This is a listing too: the exact date and minute of the start are known, the contract is new, the book is empty, there are few orders, the spread is wide. It is logical to expect the same picture — a hype-driven first impulse, then a pullback. A separate argument: some of these contracts launch in hours when the stock exchange is closed. So there is no real price at that moment at all, and the first trades are made blind — which would seem to be perfect conditions for a distortion.

On top of that comes a third argument: there are lots of these launches. A big exchange rolls out hundreds a year. If the event produces even a small distortion, at that frequency it is a ready stream of trades — a rare case where you can test a move quickly and trade it often.

How we tested it

We took the same instrument at a different time as the bar. For every launch we computed the result of shorting from the start of trading over several horizons, from an hour to a week. And we matched each one with a control: the same contract, a random moment after the first day, the same hour of the day, the same horizon. The point is to separate "the launch event" from "this instrument just behaves like that": if shorting on a random day gives the same amount, then there is no event.

Separately we split the launches into those that happened while the stock exchange was open and those that happened while it was closed — to check the "there is no price, nothing to anchor to" argument. And of course we counted the fees and the spread of a new contract: on a short horizon they decide it.

How it ended

There is no difference from the control on any horizon. Shorting from the start of trading gives about the same as shorting that same instrument on a random day. The reason is simple and structural: a perpetual on a stock is tied to the real price of the stock, so it has nowhere much to detach to — unlike a new coin, which has no anchor at all and whose price is set purely by hype. Whatever made the move work on coins is absent on stocks by design.

One slice came close to meaningful — the launches with the stock exchange closed, the very first hour. But the fees and the wide spread of a fresh contract eat half of it, and it is one cell out of many: pick through eight slices and one of them will look good for no reason. Something else matters more: the control itself came out positive. Shorting a perpetual on a stock at any moment paid about the same. That is suspicious, and most likely the sample is skewed, because only live contracts made it in — delisted ones did not, and they would have been the most awkward cases.

Live experience agrees. Stepped launches like this never produced anything before, and the money in this area was made somewhere else: when the contract drove far away from the real stock price, or when the venue itself broke. Plus a separate risk — on contracts like these, an order to close a large position can simply fail to fill, and it ends in a forced close.

What this teaches

• When you carry a move over to a new class of instruments, check whether its mechanism is alive there. If "no anchor to a price" was what worked on a coin, then a stock has an anchor, and there is no move.
• The control is the same instrument at a different time, the same hour of the day, the same horizon. Without it, it is easy to mistake a property of the instrument for a property of the event.
• If the control itself is positive, the thing to investigate is not the move but the sample. Most likely it contains no cases that ended badly.
• One lucky cell out of many is not a discovery, it is fishing. Especially when the fees and the spread eat half of it.
• How often the event happens is not an argument. Hundreds of launches a year do not make the move work, they only let you find out quickly that it is not there.