Ghost and FargoRate

FargoRate is the established rating system in pool and the one most players mean when they say "what's your rating". Ghost is an independent implementation of the same statistics, built from public tournament results, with the whole method published.

The two agree on far more than they disagree on, starting with the scale itself.

Both fit a Bradley-Terry model to game-level results, so a match score of 9–5 counts as nine games won and five lost rather than one win. Both scale ratings so that 100 points means 2:1 odds in a single game. Both weight recent play more heavily on a three-year half-life, and both treat a rating as provisional until it has roughly 200 games behind it.

That shared convention is deliberate on our side. A number that means something different from the one players already know would be worse than useless.

Vastly more data. FargoRate is fed continuously by league operators and sanctioned events, and holds millions of games. Ghost has 77,370 matches and 547,550 games. That is not a small gap, and it is the gap that matters most: a rating system is mostly its data.

Coverage of ordinary players. If you play APA, BCA or a US league, FargoRate probably has you and Ghost almost certainly does not. Our archive leans heavily on European tournaments because that is what is published openly.

It is the standard. Tournaments use Fargo ratings for seeding and handicapping. Nobody runs an event off a Ghost rating, and saying otherwise would be silly.

The method is public. Every line of the solver, the decay, the calibration and the crawler is on GitHub, and how it works explains the whole thing including its failures. You can check the arithmetic rather than trust it.

It publishes its own uncertainty. Every rating carries how many games are behind it, how many countries the opponent list spans, and what share of play came from serious long races. Where a rating rests on thin evidence the page says so, rather than showing a confident number.

Accuracy is measured out loud. Fitted on matches up to a cutoff and tested on ones played afterwards, Ghost picks the winner about 69% of the time and its stated probabilities land close to the truth. The test is in the repository and you can run it.

Head-to-head records. Every pairing that has met more than once has a page with the full record and a fair race. As far as we know nobody else publishes that for pool.

Free, with no account. Including working out your own rating from matches you have played.

If you play in a league that uses FargoRate, use your Fargo rating. It is built on more of your own results than we will ever see, and it is what your tournament director will ask for.

Ghost is worth a look if you want to see how a rating is actually calculated, if you want head-to-head records between professionals, or if you want to work out a fair race without paying for anything. The distribution here runs from about 323 to 703, with the middle of the field at 519 - so a rating means roughly the same thing on both scales.

Ghost is not affiliated with FargoRate, takes no data from them, and is not endorsed by them. FargoRate is their trademark.