How table tennis Elo ratings work
Every prediction on DeuceLab comes from one number per player: an Elo rating. Here is exactly how it works, with no black box.
What an Elo rating is
Elo is a self-correcting strength estimate. Every player starts at a baseline of 1500. After each match the winner takes rating points from the loser; beat a much stronger player and you gain a lot, beat a weaker one and you gain little.
Because the system only ever reacts to real results, it needs no human judgement. Over hundreds of matches a player's rating settles around their true level and keeps tracking their form as it changes.
From rating gap to win probability
The rating difference between two players maps directly to a win probability with the standard Elo formula: a 100-point edge is roughly 64%, a 200-point edge roughly 76%. The bigger the gap, the more one-sided the prediction.
We use a K-factor of 24, which controls how fast ratings move after each result — responsive enough to catch form, stable enough not to overreact to a single upset.
Why we calibrate
Raw Elo probabilities tend to be over-confident at the extremes. So we apply temperature calibration: a single number, fitted on thousands of past results per league, that gently pulls probabilities toward reality.
The result is honest numbers. When DeuceLab says 60%, players in that bucket really do win about 60% of the time — you can check it yourself on every league's calibration chart.
What it can't do
Elo knows only results. It has no injury, fatigue, travel or day-form data, and it cannot beat an efficient betting market. We have tested that directly and found no reliable edge — which is why you will never see a value-bet badge here. What you get is a transparent, well-measured read on who is likely to win.