Auto Pairing Pools?

The main advantage is the increased reliability and thus increased competitiveness in the ratings. Pool ratings are looked on as more official, meaningful ratings precisely because they cannot be even subconsciously manipulated (and the fact that the time controls are absolutely uniform--nobody is going to be playing 45 45 games and be in the same rating pool as someone that plays 8 0).

Having measures that prevent someone from "boosting" is not the same as having every single game played in that rating pool played under tournament conditions--ie, you can't pick an opponent from a graph/list (deciding, oh I want an easier game this time, or nah, I'm not playing him, he always gets me in the mouse race, etc.). Rating systems are actually ONLY considered accurate if you cannot select opponents. Once every game ever played in that pool is played under tournament conditions, the rating jump up in importance. There is no cloud of suspicion around someone's rating. If they have it, they earned it (discounting engine abuse, of course).

This is the main advantage. The super simple "click the giant button, start playing" thing is just an extra advantage that makes it even more attractive (and yes, it is easier to just push a big 5 button than it is to click create game, blah blah blah).

As far as fragmenting the userbase, I don't think that's a concern at all. ICC doesn't have issues with that and has 1/4 the online userbase that we do here. Not to mention people still play non-pool games all the time.

But really, if you ask any serious ICC player which rating is more important on ICC, their "blitz" rating or their "5-min" rating, 100% of them will tell you their 5-min rating is way more important. It's why when you watch curtains stream, or ChessExplained stream, or kingscrusher stream, they aren't just seeking a game, they are playing in the pools. It just becomes THE place for serious competition.

I've always thought along those same lines, Clarkey.

Nobody could really explain it better than DunnoItAll has so I won't even try. Of course, some users are going to prefer the manual approach to choosing games and others are going to prefer the pairing pools ...which is precisely why I think having both (as seen on ICC) options of starting games is the best case scenario for Lichess.

I found that the pools did not have "increased reliability" compared with the main lobby. Quite the contrary: The pools attracted mostly stronger than average players, so that a person who is normally rated 2000 would only be rated 1700 in the pools. The ratings were deflated in the pools because over time only the strong endured them.

@ #14: Again, this is exactly why there should be both methods of joining games. Pools should be there so folks like DunnoItAll and others who like them can play in them and manual seeks could be for people like you who don't prefer to play in auto pairing pools. Everyone would be happy! :)

#16:

re: reliability of rating - it is not just a matter of choosing opponents you know you can beat. It is just selecting opponents *at all*. It just mathematically makes the rating system less accurate. You can certainly ask Mark Glickman (creator of the glicko rating systems) about this, or anyone mathematically inclined. As far as the rating system accounting for it, it *attempts* to, but there is just no way to fully do it. The only way to deal with these types of problems is to make sure the user cannot choose whom they play.

On a related note, I've seen, on multiple sites, people discussing who is over-rated ("ripe" or a "good deal") and who isn't so that they can gain rating points *simply by playing them*. I mean, their opponents are over-rated, so simply by winning your normal expected portion of games against them (even if it's 3 games out of 10 or something), your rating will increase, as mathematically, you should only win 1 or 2 games against them. Your EV (expected value) in playing an over-rated player is positive, which means, in the long run, you gain rating points *simply by playing them enough times*. This sort of rating harvesting is very common. I'm not saying it's the main problem that pools try to avoid, I'm just using it as an example of an easy to see reason that opponent selection is absolutely detrimental to the integrity of a rating system.

Again, fragmenting the player base is no issue at all. There is easily a large enough base of players on lichess to handle a few pools. People that play pools don't abandon the regular pool altogether. Also, oddly, we don't have this fragmentation debate every time a new type of arena tournament or variant is added. It's just more options. I've been around internet chess since almost the beginning and seen many of these "fragmenting" arguments. It's never been a problem once the features were actually implemented. And the base here is 4x what it has been on FICS or ICC over the years. Non-issue.

As for the final point of it being easy to start a game, yes, obviously that could happen. It isn't a selling point of pools in and of itself, it's just another thing that makes it attractive.

ICC isn't doing as well today as it has in the past because it's stagnant. Same reason FICS is dying. Lichess is adding features (like this one someday, hopefully) all the time. It is exactly this idea that is killing ICC and FICS, not pools. Pools are one of the reasons ICC was the best for years, not the cause of their (at this point, imagined) demise.

#14: rating inflation and rating deflation are irrelevant phenomena. The scope of a rating pool as a mathematical system is to compare one player to another or others active in the pool at the same time. In that sense, it is outside of the scope of a rating system to compare PlayerA in the 5-min pool to PlayerA in the regular pool. Thus, a comparison of Joe having a 2300 rating in the regular pool and a 1700 rating in the 5-min pool is meaningless. What matter is Joe is 2300 and Max is 2200 in the regular pool and Joe is 1700 and Max is 1600 in the regular pool.

Modern discussions of rating inflation in FIDE are silly because they are trying to compare things like Magnus = 2860 today vs. Bobby 2810 in the 60s or similar. The two players are not active in the same pool at the same time, so the comparison means nothing, mathematically.

It boils down to "ratings are relative, not absolute." Therefore, rating inflation or deflation is absolutely irrelevant to the reliability of ratings, which is what we are discussing.

I see what you're saying. I definitely noticed pool players having a lower rating than their standard counterpart, but you're right to say that the two shouldn't be compared, as they are (literally) different pools.

Carry on.