lichess.org
Donate

Is there a table of win probabilities for Lichess games based upon rating difference?

I've seen tables like that for ELO systems. For example, at 100 points rating difference the higher rated player is X% likely to win, and so on.

Just curious if that exists for Lichess. Can't find it with Google or on the FAQ. I realize the rating system here is different so maybe that concept doesn't translate to this system?
Isn't the Lichess system pretty much the Elo system? I was under the impression that it is. If that's the case, I would think the win probabilities would be much the same.

If I'm mistaken, my apologies (and please inform me as to how the Lichess rating system is different, if possible). Thanks.
- - -

Update: So I've done a smidgen of research. Lichess uses the Glicko-2 system, as opposed to the Elo system. The Glicko and Glicko-2 systems are newer and introduce the variable of "rating reliability," designated in the formulas as RD (for Ratings Deviation). From what I'm reading, because the Glicko systems take into account the reliability (RD) of the players' ratings, you can't create a win probability table in the way you can with the Elo system.

Lichess discusses the Glicko-2 system here:
lichess.org/faq#ratings

Wikipedia article on Glicko-2 here:
en.wikipedia.org/wiki/Glicko_rating_system

And here's a table comparing the two rating systems, with an entry on score probabilities:
www.englishchess.org.uk/wp-content/uploads/2012/04/Elo_vs_Glicko.pdf
Elo azned Glicko use same model for win probabilities so the Elo table s serve you. Or maybe Glicko use Logistic and Elo Normal distrinution but they give virtually same probabilities unless difference is very big so you can use the Elo numbers on use the example formulas from wikpedia
Good research @GSP0113 . I think this answers the question:

"The score probability is dynamically calculated by the formula. It depends on both players’ ratings the opponent’s rating deviation"

I think that means the answer to my question is "no". For the Glicko system, it's not enough to know they two opponent's ratings (whereas for Elo it is), to calculate a win probability. You would also have to factor in the rating deviation through a formula. So not so simple that you could easily publish in a basic table.
basically, the elo and glicko equation are pretty similar. So I guess there is a winnin/lose probability table
For any reasonable established rating RD effect is not meaningfull
en.wikipedia.org/wiki/Glicko_rating_system
see the equation fot E() expected score it has the RD in exponent function
the value on function g() bringing the RD into game
RD=60 q = 0.006 (about) gives
g() = 0.98. Otherwise Equation is similar to wikipedia formula for elo (which is actually the logistig one ) apart by this multiplication of 0.98 on which pretty much is not meaningfull. In case of very big rd of opponent it has some meaning.
Glicko-2 is not very good at that point. It reduces the expected score for large RD. it shoudl have impact on step size not expected score.

Well anyway Glicko-whatever does not present state of the art. but is simple enought to implement hence the popularity
@petri999 You wrote: "For any reasonable established rating, RD effect is not meaningful."

That's the way it was designed. RD is supposed to play a significant factor only in instances where there is reason to doubt the accuracy of the rating. What you are saying is akin to, "In the case of the laboratory mice who didn't die, death was not a significant factor."

You wrote: "Glicko-2 is not very good at that point.... Glicko-whatever does not present state of the art. But is simple enough...."

It appears you don't think much of Glicko-2. What do you think is better?

This topic has been archived and can no longer be replied to.