Pinball ratings explained: IFPA rank, Match Play rating and the 0-1 machine score

Open a profile on this site and you get hit with four numbers before you have finished scrolling past the header. IFPA rank. IFPA rating. Match Play rating. Then a column of 0.6-ish decimals next to machine names, which is ours and nobody else's. Somebody asks me what the difference is about once a week, usually at a bar, usually right after they have used one of the four to argue they should have won a game they did not win.

So here is the whole set. What each number measures, who works it out, how often it moves, and the situations where it will lie to you.

IFPA rank: your place in line

Your IFPA world rank is a position, not a score. The International Flipper Pinball Association hands out WPPR points (World Pinball Player Rankings, pronounced "whopper" by everyone) for finishing well in sanctioned events, adds up what you have earned inside a rolling window of roughly the last three years, and sorts every player on the planet by the total. Older results fade out of that window, which is why a player who took a season off slides down the list without playing a single bad tournament.

Two things follow from that, and they trip people up constantly.

The first is that rank rewards showing up. Points are earned, never lost, so the player who enters forty events a year will outrank an equally good player who enters eight. That is a feature of the system, not a bug in it, but it means world rank is a poor answer to "who is better" and a decent answer to "who has done more".

The second is that a rank number means different things at different depths. The gap in real playing strength between rank 300 and rank 400 is enormous. The gap between 6,300 and 6,400 is a rounding error and a single good Thursday night. When you see a rank on a player profile, read it as a bracket rather than a precise measurement.

We store the monthly history of that number, so the chart on a profile shows the actual climb rather than just today's snapshot. The axis is inverted, because rank 1 is up.

IFPA rating: strength, separated from attendance

IFPA also publishes a rating, and it is a completely different animal from the rank. The rating is a Glicko-style estimate of how strong you are, built from how you finish against other rated players. Beat people above you and it climbs. Lose to people below you and it drops. Entering more events does not, by itself, push it anywhere.

That makes the rating the better of the two IFPA numbers for the question people actually mean when they ask about rank. It is also the number this site leans on hardest behind the scenes, for reasons that will make sense in a minute.

Both IFPA numbers appear on every profile header and in the player directory, and the Minnesota rankings page sorts the local field by world rank alongside the NACS state championship standings, which are a separate race with their own qualification rules.

Match Play rating: the one that moves tonight

Match Play Events runs most of the tournaments we track, and it keeps its own rating, updated from the games entered in the app. It is Glicko-flavoured too, so it has a confidence interval hiding behind it, and it reacts fast. Play a league night, and it has moved by the time you get to your car.

That responsiveness is the whole point of it and also the reason not to read too much into a week of it. Four bad games against a strong group will knock it down more than those four games deserve. The trend line matters; the daily value does not. Every profile draws that trend as a chart, and the Pulse page keeps a rating ladder of the local field plus the biggest climbers of the last month, which is the fastest way to see who is in form.

The machine score: ours, and the one worth learning

Now the decimals. Every machine row on a profile carries a score between 0 and 1, where 0.5 is dead average. This is not an IFPA number and not a Match Play number. We compute it from the raw game results, and it answers one narrow question:

On this specific machine, how do you do against the people you actually played against?

Here is the calculation, because a stat you cannot audit is a stat you should not trust.

Step one: every game becomes a pile of pairwise results. A four-player group is not one result, it is six little head-to-head matchups. Finish second out of four and you beat two people and lost to one. Those pairings are what get counted, not the finishing position on its own.

Step two: opponents are weighted by IFPA rating. Beating a 1,400-rated player is worth more than beating an unrated newcomer, so each opponent contributes their rating as the weight on that pairing. A win takes the full weight, a tie takes half, a loss takes none. Unrated players still count, at a default weight, because a win is a win.

Step three: the total is squeezed toward 0.5 by a Bayesian prior. This is the part that saves you from yourself. A raw ratio over three games is noise wearing a lab coat. So before dividing, we add a phantom block of average-strength results, sized to the median amount of evidence other ranked players have on that same machine. Two games at a perfect record barely lifts off 0.5, and on a heavily played machine it hardly moves at all. Twenty-five games at seventy percent stays close to 0.7, because by then the real data outweighs the phantom block.

The practical upshot: a high score on a low game count is the machine telling you it does not know yet. The game count sits right next to the score for exactly that reason. Click it and the game log pops open with every result behind the number, opponents and all.

Score versus win rate, and why both are there

Next to the score you will find a win rate, which is the plain unweighted share of pairwise matchups you won. The two disagree more often than you would think, and the disagreement is informative.

The finish spread, the 1st-2nd-3rd-4th tallies on the same row, adds the last piece. Two players can share a score of 0.62 where one is a metronome of seconds and the other alternates wins and fourths. In a format that pays for wins, those are not the same player.

Where each number lives

• Header of any player profile: IFPA rank, IFPA rating, Match Play rating, games on record.
• The All machines tab on a profile: every machine with score, win rate, finish spread and game count.
• The Insights tab: playing-order splits, early versus late rounds, attendance and upsets.
• Machine pages: the same score computed the other way round, as a leaderboard of who runs that game and who it eats.
• Compare: up to four players side by side, ratings and finish quality in one table.
• Head-to-head: the actual record between two people, which beats comparing ratings.
• Records: the all-time boards, where the extremes live.

A word on what none of these measure

None of these numbers know that the outlane post was moved last Tuesday, that you play better after eating, or that the machine you allegedly cannot beat has a broken flipper coil on the left. They are a summary of results, and results are a noisy signal of skill. Treat a 0.72 as "the evidence leans this way" rather than "this is a fact about me".

With that said, they are an enormous improvement on the traditional method, which is remembering your best game on a machine and forgetting the other nine.

If you want to put the numbers to work rather than argue about them, the scouting walkthrough turns them into machine picks, and the WPPR explainer covers where the points that drive your rank come from in the first place. If the number you care about is how you do against one particular person rather than the field, that lives in head-to-head records. The rest of the archive is on the blog index.