lichess.org
Donate

On auto-pairing pools

This is going to be long, so sorry about that, heads up!

BACKGROUND

I am aware that, ICC-like auto-pairing pools used to exist on lichess some time ago, but they were scrapped for various reasons.

Correct me if I am wrong, but I guess the following were some of the reasons why it was scrapped:

  1. Not enough players frequenting the pools
  2. Competed with arena tournaments for participants
  3. Competed with lobby seeks

I also heard a bunch of complaints that they were not implemented correctly either - for example, one could see exactly who was in the pool at any given time, allowing a player to selectively engineer his opponents (at least to some degree).

The purpose of an "auto-pairing pool", and the reason why it is so popular on sites like ICC, is that it is an instant but fairly accurate pairing system.

  1. It prevents selective pairing and thus improves rating accuracy.
  2. It is stupid simple to get paired. No need to go to lobby, post a seek / find a seek, etc. Enter pool... get auto-paired. DONE.
  3. You get varied opponents, but always someone close to your skill level - always a good match (which is what most players want 99% of the time).

WHY ARENA TOURNAMENTS CANNOT REPLACE POOLS

The dynamics of a tournament are somewhat different. In the arena tournament, you are playing for the bigger prize, and your strategy is geared towards that. You take more risks, play for quick wins, and of course - sometimes you go bezerk!

When you are in the pool, you just want 10 good fair games. You treat each game as an independent game and you don't have the burden of keeping an eye on your tournament standing.

Another thing about arena tournaments is that it is a common pattern to see players often late-join. The problem is that these tournaments happen in discrete intervals, and if you happen to login at an inopportune time, too bad.

Pools don't have this problem. You can get a good game whenever you want.

WHY LOBBY SEEKS CANNOT REPLACE POOLS

Auto-pairing has a sort of carefreeness zing to it. You just hit a button and get a good pairing.

Currently, as implemented. Seeks require explicit effort (on part of the seeker and the joiner). There is also the 'burden of configuration' (the user has to make a bunch of decisions as to what exact seek he wishes to post). This is something which is sometimes underestimated, but I think it matters - maybe not to power users, but at least to casual users (who are the majority). It is good design practice to shift the burden of decision making away from the user and into the system.

=====

PROPOSAL

Introduce auto-pairing for popular time controls (1+0, 3+0, 5+0, 10+0), but don't expose the pools mechanism to the user. If successful, introduce more time/variant presets.

Specifically, one could provide these presets as buttons in the 'create a game' menu. They would act as quick ways to get paired. (The preset buttons can be also outside the 'create a game' menu - probably better - and more easily accessible directly from the homepage)

In the background lichess would treat such players as if they were in the pool and would pair them accordingly, but the player would never know anything about the pool, or who is in it, or anything at all. The pool system is abstracted away from the user. He presses the button and magic happens, and that's it.

P.S. The auto-pairing could be restricted to rated games (for simplicity sake), so casual games / anonymous games would still have to be explicitly set up.

Correct me if I am wrong, but chess.com has had a flavour of this for a while now.

Pros:

  1. Auto-pairing is instantaneous, hassle-free and magical. Casual users get a good experience.
  2. It does not interfere with lobby seeks. If I want a 3+1 game, I can still do it. If I want a 5+0 game with players between 1500-1800, I can still do it. Seeks are still visible in the lobby. The idea is that for the casual user, who just wants a game, it just works. For those who seek more advanced control, they still have it.

Cons:

  1. It doesn't solve the selective pairing problem. It only mitigates it a little. You can always choose to not use the 'get a quick game' feature. The hope is that with somewhat popular adoption, lichess ratings become just a bit more accurate.
  2. It doesn't give you insight into your performance for the sessions (something that tournaments do very well). However this can be easily subverted by implementing some sort of 'your performance this session / your performance today' global metric for lichess. It can be simple or as detailed as necessary. So when a player plays 10 games on lichess one afternoon, he can end his session by clicking on 'daily performance stats' or something of that sort - which gives him a detailed overview of his performance for that session.

CLOSING REMARKS

I think this is a rather elegant solution to the 3 points I mentioned at the start of this article.

  • Not enough players frequenting the pools

I think lichess has grown exponentially since pools were first tried. Perhaps now, there will be more participants and this will not really be a problem.

  • Competed with arena tournaments for participants

I think fundamentally, the set of players who opt for the 'get a quick game' format are largely independent of set of players who participate in arenas. As explained above, both are very different experiences. I don't think it will have a large impact on arena participation (but lets see).

  • Competed with lobby seeks

I think this proposal plays rather elegantly with the current lobby seek system. From the point of view of the user, it is like a normal seek, but with a preset configuration. The set of players who seek the traditional way will decrease as some move to auto-pairing (mostly casual users). However, seeks with unusual time controls and variants will continue to be prevalent, Power users with very specific opponent requirements will also continue to use the 'create a game' menu. It will just get less common.

=====

What do you think? Please give your comments below. Perhaps we can have a little discussion.

Thanks :)

This is going to be long, so sorry about that, heads up! BACKGROUND I am aware that, ICC-like auto-pairing pools used to exist on lichess some time ago, but they were scrapped for various reasons. Correct me if I am wrong, but I guess the following were some of the reasons why it was scrapped: 1. Not enough players frequenting the pools 2. Competed with arena tournaments for participants 3. Competed with lobby seeks I also heard a bunch of complaints that they were not implemented correctly either - for example, one could see exactly who was in the pool at any given time, allowing a player to selectively engineer his opponents (at least to some degree). The purpose of an "auto-pairing pool", and the reason why it is so popular on sites like ICC, is that it is an instant but fairly accurate pairing system. 1. It prevents selective pairing and thus improves rating accuracy. 2. It is stupid simple to get paired. No need to go to lobby, post a seek / find a seek, etc. Enter pool... get auto-paired. DONE. 3. You get varied opponents, but always someone close to your skill level - always a good match (which is what most players want 99% of the time). WHY ARENA TOURNAMENTS CANNOT REPLACE POOLS The dynamics of a tournament are somewhat different. In the arena tournament, you are playing for the bigger prize, and your strategy is geared towards that. You take more risks, play for quick wins, and of course - sometimes you go bezerk! When you are in the pool, you just want 10 good fair games. You treat each game as an independent game and you don't have the burden of keeping an eye on your tournament standing. Another thing about arena tournaments is that it is a common pattern to see players often late-join. The problem is that these tournaments happen in discrete intervals, and if you happen to login at an inopportune time, too bad. Pools don't have this problem. You can get a good game whenever you want. WHY LOBBY SEEKS CANNOT REPLACE POOLS Auto-pairing has a sort of carefreeness zing to it. You just hit a button and get a good pairing. Currently, as implemented. Seeks require explicit effort (on part of the seeker and the joiner). There is also the 'burden of configuration' (the user has to make a bunch of decisions as to what exact seek he wishes to post). This is something which is sometimes underestimated, but I think it matters - maybe not to power users, but at least to casual users (who are the majority). It is good design practice to shift the burden of decision making away from the user and into the system. ===== PROPOSAL Introduce auto-pairing for popular time controls (1+0, 3+0, 5+0, 10+0), but don't expose the pools mechanism to the user. If successful, introduce more time/variant presets. Specifically, one could provide these presets as buttons in the 'create a game' menu. They would act as quick ways to get paired. (The preset buttons can be also outside the 'create a game' menu - probably better - and more easily accessible directly from the homepage) In the background lichess would treat such players as if they were in the pool and would pair them accordingly, but the player would never know anything about the pool, or who is in it, or anything at all. The pool system is abstracted away from the user. He presses the button and magic happens, and that's it. P.S. The auto-pairing could be restricted to rated games (for simplicity sake), so casual games / anonymous games would still have to be explicitly set up. Correct me if I am wrong, but chess.com has had a flavour of this for a while now. Pros: 1. Auto-pairing is instantaneous, hassle-free and magical. Casual users get a good experience. 2. It does not interfere with lobby seeks. If I want a 3+1 game, I can still do it. If I want a 5+0 game with players between 1500-1800, I can still do it. Seeks are still visible in the lobby. The idea is that for the casual user, who just wants a game, it just works. For those who seek more advanced control, they still have it. Cons: 1. It doesn't solve the selective pairing problem. It only mitigates it a little. You can always choose to not use the 'get a quick game' feature. The hope is that with somewhat popular adoption, lichess ratings become just a bit more accurate. 2. It doesn't give you insight into your performance for the sessions (something that tournaments do very well). However this can be easily subverted by implementing some sort of 'your performance this session / your performance today' global metric for lichess. It can be simple or as detailed as necessary. So when a player plays 10 games on lichess one afternoon, he can end his session by clicking on 'daily performance stats' or something of that sort - which gives him a detailed overview of his performance for that session. CLOSING REMARKS I think this is a rather elegant solution to the 3 points I mentioned at the start of this article. * Not enough players frequenting the pools I think lichess has grown exponentially since pools were first tried. Perhaps now, there will be more participants and this will not really be a problem. * Competed with arena tournaments for participants I think fundamentally, the set of players who opt for the 'get a quick game' format are largely independent of set of players who participate in arenas. As explained above, both are very different experiences. I don't think it will have a large impact on arena participation (but lets see). * Competed with lobby seeks I think this proposal plays rather elegantly with the current lobby seek system. From the point of view of the user, it is like a normal seek, but with a preset configuration. The set of players who seek the traditional way will decrease as some move to auto-pairing (mostly casual users). However, seeks with unusual time controls and variants will continue to be prevalent, Power users with very specific opponent requirements will also continue to use the 'create a game' menu. It will just get less common. ===== What do you think? Please give your comments below. Perhaps we can have a little discussion. Thanks :)

TL;DR:

Implement 'get a quick game now' buttons on homepage with preset configurations (popular time controls - 1+0, 3+0, 5+0, 10+0), which act as simple seeks but use auto-pairing magic in the backend.

It's like a marriage between auto-pairing pools and lobby seeks.

chess.com already has it.

TL;DR: Implement 'get a quick game now' buttons on homepage with preset configurations (popular time controls - 1+0, 3+0, 5+0, 10+0), which act as simple seeks but use auto-pairing magic in the backend. It's like a marriage between auto-pairing pools and lobby seeks. chess.com already has it.

Excellent explanation @Pyromaniac! I totally agree.
Lichess, please, bring the pools back!

Excellent explanation @Pyromaniac! I totally agree. Lichess, please, bring the pools back!

Been there, watched them die painfully and embarrassingly, hope we don't go there again.

One of the (many) big problems was that lower-ranked players were too self-conscious about joining (fat people at the gym syndrome) and so the standard of player was off-puttingly high.

All sports have 'their own way' of doing things that seem a bit alien to those outside of them. Chess is no different. I think that another failing of the pools was the lack of an absolute winner. Competitive people enjoy competing.

That's why my counter argument is that there should in fact be another sort of competition altogether: http://en.lichess.org/forum/lichess-feedback/lichess-super-league?page=1#1

Been there, watched them die painfully and embarrassingly, hope we don't go there again. One of the (many) big problems was that lower-ranked players were too self-conscious about joining (fat people at the gym syndrome) and so the standard of player was off-puttingly high. All sports have 'their own way' of doing things that seem a bit alien to those outside of them. Chess is no different. I think that another failing of the pools was the lack of an absolute winner. Competitive people enjoy competing. That's why my counter argument is that there should in fact be another sort of competition altogether: http://en.lichess.org/forum/lichess-feedback/lichess-super-league?page=1#1

#4

Thanks for the comment.

Your idea is also neat. I made a comment there as well...

However, I think we can have both - pools and super league. Both of them serve different userbases.

My stance is that the more formats, the merrier (as long as one format does not steal too many users from another format). Lichess with seeks, arena tournaments, super league, pools, maybe even an official 45+45 league... is just a better, richer lichess.

I can understand the hesitation with pools, knowing its history. But, it has been over a year now and the lichess landscape is very different. Which is why I propose to bring it back, but not like the old pools... this time, with a simple twist.

Fingers crossed XX

:)

#4 Thanks for the comment. Your idea is also neat. I made a comment there as well... However, I think we can have both - pools and super league. Both of them serve different userbases. My stance is that the more formats, the merrier (as long as one format does not steal too many users from another format). Lichess with seeks, arena tournaments, super league, pools, maybe even an official 45+45 league... is just a better, richer lichess. I can understand the hesitation with pools, knowing its history. But, it has been over a year now and the lichess landscape is very different. Which is why I propose to bring it back, but not like the old pools... this time, with a simple twist. Fingers crossed XX :)

+1

I often read that arena tourneys are "the better pairing pools", which is plain wrong if you're looking to play one or two quick games rather than a complete tourney.

Yeah, why not give auto-pairing pools a second chance, I don't get it? Given the massively grown user base I'd be very surprised if it wouldn't work out.

+1 I often read that arena tourneys are "the better pairing pools", which is plain wrong if you're looking to play one or two quick games rather than a complete tourney. Yeah, why not give auto-pairing pools a second chance, I don't get it? Given the massively grown user base I'd be very surprised if it wouldn't work out.

#5 I'm glad that you like the idea also. I'm not convinced by your 45+45 - all it takes is one or two people to rage quit, and the whole of lichess is in uproar!

#5 I'm glad that you like the idea also. I'm not convinced by your 45+45 - all it takes is one or two people to rage quit, and the whole of lichess is in uproar!

Maybe it's an idea to do statistics based automatic pools: when there are at least say 3 games in the same time/game category, let a button appear for it to start a quick game (auto-pairing automatically on) while you can turn auto-pairing off explicitly when creating a lobby search. This is especially useful to present on the end of a game alongside the rematch button, to keep people playing. You could also slightly favor those buttons to variant/time controls used by the current game, and of course add a button to start a new lobby game for the same game type you just played.

Maybe it's an idea to do statistics based automatic pools: when there are at least say 3 games in the same time/game category, let a button appear for it to start a quick game (auto-pairing automatically on) while you can turn auto-pairing off explicitly when creating a lobby search. This is especially useful to present on the end of a game alongside the rematch button, to keep people playing. You could also slightly favor those buttons to variant/time controls used by the current game, and of course add a button to start a new lobby game for the same game type you just played.

#7

Actually I have a couple of rough ideas on how to implement a 45+45 league.

In short, my idea is to programatically implement the architecture of the current lichess4545 league. Players mark their free slots on the lichess weekly calendar. The pairing algorithm then automatically matches players with intersecting schedules and players are then notified when paired. To ensure more robustness, players can cancel/change their schedules at any time up to 12hrs before the game. If that happens, both players get re-paired.

For this to work, there are somethings that need to be considered, and I haven't worked out all the kinks yet.

Most likely, this system would need some redundant players - a waitlist, to ensure that if players change/cancel their schedule, there are backup pairings possible.

Scoring also has to be worked out. I like how lichess4545 have implemented it, and that might be the best way to do it- TEAMS. Scores are team based, so teams can have 4 primary players + 2 backup players - and this takes care of redundancy as well.

These are just some ideas, and it may not work at all... But judging by the success of the lichess4545 league, I think it might be worth considering.

#7 Actually I have a couple of rough ideas on how to implement a 45+45 league. In short, my idea is to programatically implement the architecture of the current lichess4545 league. Players mark their free slots on the lichess weekly calendar. The pairing algorithm then automatically matches players with intersecting schedules and players are then notified when paired. To ensure more robustness, players can cancel/change their schedules at any time up to 12hrs before the game. If that happens, both players get re-paired. For this to work, there are somethings that need to be considered, and I haven't worked out all the kinks yet. Most likely, this system would need some redundant players - a waitlist, to ensure that if players change/cancel their schedule, there are backup pairings possible. Scoring also has to be worked out. I like how lichess4545 have implemented it, and that might be the best way to do it- TEAMS. Scores are team based, so teams can have 4 primary players + 2 backup players - and this takes care of redundancy as well. These are just some ideas, and it may not work at all... But judging by the success of the lichess4545 league, I think it might be worth considering.

I find it disappointing that I've seen people giving this feedback many times without response. I prefer the Lichess interface and ethos to other sites, but use them for my games, simply because I can go to the site and click 'play'. I don't have to manually choose a game or join a tournament where I'm going to get beaten every game by way higher rated players.

I find it disappointing that I've seen people giving this feedback many times without response. I prefer the Lichess interface and ethos to other sites, but use them for my games, simply because I can go to the site and click 'play'. I don't have to manually choose a game or join a tournament where I'm going to get beaten every game by way higher rated players.

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