We got the research wrong on day one

What we searched

We started where anybody would: three strikes pool rules, "3 strikes" pool tournament, three strikes you're out pool variant. All three came back with bar billiards, three-ball, cutthroat and bottle pool. No Wikipedia article. No ruleset on any of the billiards forums. Nothing.

So we wrote the obvious conclusion into the strategy: the rules of this game are not documented anywhere, and nobody has built software for it. A blue ocean. We built a pricing model on top of that conclusion.

What we missed

The game is called Killer.

In British and Irish pubs — and in a good deal of the rest of the world — the same game, with the same one-shot-per-turn rotation and the same three lives, has been played under that name for decades. It has a Wikipedia article. It has rules pages on multiple pool sites. And it has at least five apps built specifically for it, one of which does a good chunk of what we had just finished describing as an unserved need.

We found all of it in a single query, as soon as we stopped searching the name and started searching the mechanic: lives, elimination, last man standing.

Why this is a normal mistake and still a bad one

Folk games have regional names. That is practically the definition of a folk game. Searching one name and concluding the category is empty is the research equivalent of looking in one drawer.

What makes it a bad mistake rather than a forgivable one is that the conclusion was load-bearing. "Nobody does this" went straight into the positioning, and from there into the pricing — we had planned to charge for night history and player statistics. Both of those are given away free by competitors we had not found. Shipping that pricing would have made us, objectively, the worst free option in a category we thought we had invented.

What changed

Three things.

The positioning. It is no longer "nobody runs this format". It is "nobody running this format handles the money, and nobody is building for a screen on the wall" — which is still true after checking all five competitors, and is a much more defensible thing to say.

The pricing. Unlimited history and full player stats moved to the free tier, where the competition already has them. The paid tiers were re-pointed at the venue: branding, a public page, staff accounts, multiple tables. That is a buyer with a budget and genuinely nobody selling to them.

The content. The rules page now covers both names properly, which it should have done anyway, because somebody who learned this in a pub and somebody who learned it in a bar are looking for the same page.

The part worth keeping

The research document still contains the original claim. It is in section 6, under a heading that says we got it wrong, with the search queries that produced it and the query that corrected it.

Deleting it would have been tidier. But a research doc you can trust is one where the mistakes are visible, because a document that has only ever been right is a document nobody has checked.