Matching bias(es) at chess.com

Sort:
Martin_Stahl

Looking at an arbitrary archive of a paying member may not tell you anything useful. A premium member may be playing a lot of games within clubs that are premium only, or have a higher ratio of premium memberships compared to basic, so that could skew results.

 

There are some pairing processes that attempt to not pair higher lag players with lower lag ones, and will pair players in the fair play pools (lots of disconnect losses, making players wait in losing positions, etc), so there is a selection being made in those instances. Other than that, the random pairings are based on seek parameters, with an attempt to pair compatibly within those constraints.

 

I'm not saying that there isn't some part of the algorithm that might not pair premiums more often. I'm just a member, and have no special insight into the inner workings of the system, other than what is published in the help articles or has been posted about by staff.

https://support.chess.com/article/369-how-does-matching-work-in-live-chess

Martin_Stahl

Basic members outnumber premium by a decent margin. For players playing exclusively in the random pools, i.e. primarily not in club events or tourneys, I would generally expect similar percentages across a large, random sample of players.

 

As to it impacting ratings, any effect is likely to be small, in my opinion. While the premium pool is smaller, any inflationary impact would be mitigated by the fact the playersw would still be getting paired with basic members with similar ratings ranges. If the results of those games were statistically deviated from what would be expected, then there might be something to it.

Martin_Stahl
thinkneo wrote:

Chess.com is without any doubt biasing its matching, by matching non-paying members to other non-paying members.  Again, you'll see this pattern consistently, over and over.  The fact that this isn't mentioned in the article referenced above really bothers me.  It makes me wonder, what are they up to?

What might a non-paying member do, if their ratings went from a level they considered satisfactory to something considerably less than that?  Would they consider paying for membership, to maybe get some lessons?

Again, there's a reason for this obvious and undeniable biasing, and my guess is that it's related to money. 

 

 

There are plenty of basic members across all rating bands. Even if basic members are being paired with a higher percentage of other basic members, that would not create a downward pressure on ratings. If anything, those ratings would likely be more accurate with a wider pool of players.

autobunny
ghost_of_pushwood wrote:

Wow, I guess Martin just felt like delivering a lecture.

and the op, @thinkneo bailed, with another thread

https://www.chess.com/forum/view/community/discrepancy-in-ratings

and now martin owns the thread.

https://www.chess.com/forum/view/suggestions/the-op-shalt-not-be-deletable

Martin_Stahl
ghost_of_pushwood wrote:

Wow, I guess Martin just felt like delivering a lecture.

 

I almost always quote the OP, but the first post was really long and I didn't feel like editing the quote on my phone to make it more readable ... meh.png