Make quick matches the primary way of finding matches

The quick match finder takes ages, and it feels like it serves little purpose, at least for me as a 15kyu. Most games I pick up through the custom games list. However, there are so many variations of time settings, handicaps, komi, rules, etc. possible that it ‘skews’ the rank in a way. For example, correspondence games often feature analysis, score estimation, and other such tools, which helps me a lot because I’m not that good at reading. But it’s ‘unfair’ for those ignoring those options or are less patient than me at testing all possible variations.

Seems to me it would be fruitful to focus on the quick match finder as the primary way to find games, and have preset types of games. The ranking reflects this already with 9x9, 13x13, 19x19, fast, slow, correspondence. Is it desirable to have only quick match presets count for rank? You could select chinese/japanese ruleset (or both), board size (like now), and fast, slow, correspondence (like now). Then analysis and such tools are disabled. A bonus is that you would have little choice if you play someone better or worse than you.

Then, custom games would be for teaching, variations, practice, tournaments, handicaps, and such, but without ranking. Tournaments could also of course use the presets.

Is it desirable to focus on the quick match finder? Why or why not?

2 Likes

I don’t know about the specifics of the implementation, and I don’t want to debate about which preset settings should be the default for such games.

However I have heard/read this a good bit recently, that the community/matchmaking is somewhat split and feels that way between custom game list and automatch, and naturally splitting the pool of players up makes getting a game harder (with specific settings) in one or the other or both.

I think a certain (probably large) amount of people do prefer hosting custom matches, and it was suggested before to just show the automatch queue as games with default settings in the custom games list. That way someone who is looking for a custom game, and doesn’t usually use the automatch can still accept to match with the automatch queue provided they meet the requirements (usually rank say).

post about long waits for correspondence games with the automatch

1 Like

This could be one solution. I read those posts, however, it feels like for ranked matchmaking, there should be certain standard the game needs to fall within to be counted. 1m + 2 second byomi for example would be outside as one little lag and you’re out. Analysis turned off should be another. Of course, people could choose not to play those games, but if you’re new to the site, or new to Go, it’s kinda confusing where to start. Another issue with this solution, IMO, is that with an automatch making approach, it can gradually expand the rank search range, like most computer games do it. Not so with custom match. I don’t know if the match making in OGS does that already.

It would be interesting to check why people use the custom match feature. Perhaps a seperate post on that would be interesting in its own right.

I was just highlighting the posts for others as well.

I think though, while it could gradually expand the search range, you can also set rank limits on the automatch search - so it’s not always going to do so unless you let it.

I don’t actually know the details myself on how the preferences like Handicap “Prefer” and “Require” impact the search. I am actually noticing (just now) for instance that the defaults for these types of setting depend on the mode Blitz/Live/Correspondence already.

When I first learned that automatch might not get me a game even if there are custom game requests waiting that would perfectly match my automatch request, I was really confused.
I was expecting the server to do everything to give me a game quickly. Particularly when I’m using a generic automatch request that is very open to the settings of the game.
My expectation would be that every match request (auto and custom) is kind of a query with certain constraints. Some queries being more and some less restrictive. But as soon as two queries match sufficiently well, a game would be created.
I still can’t wrap my mind around the fact that the server doesn’t do that.

Isn’t it one of the key features of a game server to bring people together to a game as quickly as possible, if the parameters of the game would work for both of them?

1 Like

I also desperately want automatch to work, but am frustrated at the time it takes to get a game generally.

A couple thoughts that might make this option more popular:

  • Reducing time setting options for automatch. Combining the blitz and live automatch buttons could effectively double the likelihood of getting a game
  • Widening the default max handicap (I think it’s 3 right now, I’d gladly play 4 or 5 stones if it meant I didn’t have to wait 5 min)
  • Pulling from the Custom pool if appropriate. Should be ranked only. I find that sandbaggers like to play unranked games :smirk:
  • adjusting any of the above knobs depending on time of day, rank, time waiting, etc.

I don’t know if any of those would help for sure, but I think getting the automatch down below 1 minute for like 95% of games would be a super duper goal for OGS.

Agree. One could also do something as ‘drastic’ as turning off ranked for any custom game that does not follow one of the ‘acceptable’ formats. For example, my suggestion would be:

  • no analysis
  • expanding ± rank difference (except tournament), increase by one every 1 minute (except correspondence)
  • 5-10 minutes and 10-20s byo yomi (Blitz)
  • 1-5 minuts and 5-10s byo yomi (Bullet)
  • 10-20 minutes and 30-45s byo yomi (Slow)
  • 20-180 minute, and 30-60s byo yomi (Tournament)
  • 3d-7d per move (correspondence)
  • 9x9, 13x13, 19x19 - free choice
  • Japanese or Chinese scoring - free choice
  • No handicap
  • Automatic komi

Anything outside those presets would be unranked. If so, one could easily combine custom and quick match lists without suddenly matching with a game of no komi and canadian absolute time of 3 minutes.

I agree that simplification is needed. I think a page should be taken out of FoxWeiqi’s book. That server has 10s of thousands of people playing every day, and yet they still only offer 4 auto match settings:
19x19 5+30x3
19x19 20+60x3
9x9 1+20x3
13x13 1+20x3

Players can be matched within one rank for no Komi games. Anything else requires a custom game through their ‘invite’ system (not suggesting OGS implements this bit)

I think simplifying like this is even more essential for a low population server like OGS. Right now there’s just way too many options, trying to please everyone. I think it’s better to chose a server default regarding rules, handicap, then give players a lichess style grid with a small selection of sizes and time controls they can click, no setup or tweaking required.

Put the custom, correspondence and bot match options to the side somewhere. Though I like the suggestion that auto match could detect compatible custom games.

3 Likes

I would argue that it is better to have more options than less. Beauty of OGS is the vast variety of settings we can use, it would be real shame if ogs were to stop supporting long live games. One big annoyance with other go servers i have, is super limited game settings.

Here i can play correspondence, i can play quick games, i can play games with EGF A-class -style settings and i can play teaching games where we talk about variations without the fear of timing out.

I personally feel that having large variety of different settings is definitely the strength of ogs compared to other go servers. ^^

But what i would see happening, is that automatch would also consider current open custom challenges if any of those would be fitting for the settings, which the automatch-user is using.

2 Likes

This is just about the quick-match system though. Nobody is suggesting removing any options from custom games.

I personally feel that having large variety of different settings is definitely the strength of ogs compared to other go servers. ^^

I agree. OGS is the best place for tournaments, correspondance, and playing with friends, but I still spend most of my time on Fox and IGS instead, because I can get games quicker. Part of that is population size, but part of it is also that everyone on those servers is hitting the same quickmatch button that they are presented with the moment they start the client. Anything OGS can do to improve the speed someone can get a game will then also increase the population of active players.

My other suggestion would be to put quick match buttons on the currently mostly barren homepage, which I think I remember hearing is already in the works.

3 Likes

I agree in this, but if one turned off ‘ranked’ for custom games, unless they fit within the presets, then that would incentivise people to choose quick matches for the most part, right? One only needs to make the quick match finder either more accessible, e.g. more visible and intuitive, or it needs to something that custom matches don’t such as ranking.

Just use custom games, there’s absolutely no downside to using them.

Automatic pairing is only needed when the server becomes large and people put up and take challenges so quickly, the open challenges list becomes unusable.

1 Like

There are several downsides, depending on your personality and or experience with Go / OGS:

  • fish for ‘easy’ games
  • avoid ‘scary’ games when you’re a beginner (or trying to rank up)
  • enable tools like analysis, score estimation, etc., which favors those patient/quick/skilled enough to utilize them better than your opponent. In other words, you can gain an edge outside of the board itself
  • custom handicap/komi/time settings that might not be obvious to a beginner
  • non-intuitive if you’re a beginner to Go or online Go - what’s kyu? what’s komi? is 1 kyu my level?
  • non-obvious how they count for the different rank categories (fast, slow, correspondence)
  • most game servers use matchmaking (starcraft, warcraft, dota, etc) <- people may be more familiar with it
  • I don’t want to scroll through the games on the list to determine which one I should take. Usually not a problem but it wouldn’t take much more games before it became one <- when I started with Go, I was confused about this
  • Sometimes games dissapear as you try to accept them, small issue, but is just annoying

But let’s turn it around, what’s the issue with focusing on the quick match finder, and have the custom games more for the different variants that are outside of the standard formats?

2 Likes

The issues with quick match finder (that realistically matter) are being unable to set the time control you prefer and lack of transparency. Also, no name for your game.

The issue with focusing on the quick match finder is that it’s a waste of time when you can make a custom game. I agree that quick matcher is too raw at the moment, so let’s hide it somewhere until better times and use custom challenges instead. I still see very few downsides.

These are positives. The creator makes the game they want, and the acceptor accept the game they want. Zero problems.

Beginners should get used to it fairly quickly. Until then some obvious way to set “default” challenge would be nice, yeah.

Then you can create a challenge, which is what I typically do. And accept one only if a good one happens to pop up in my view.

It’s so flexible, it’s so nice. You can see all challenges, everyone can see your challenge. In real time you can see which challenges get snatched quickly, and which are slow to take off, so you can learn what settings are better and faster. You can see how many people are waiting for games. You can see which settings are popular. A truly majestic system.

2 Likes

Positives if you want a specific challenge, I agree. And for most on OGS, there’s little issue. But, it’s not so easy for beginners, and Go is not so ‘newbie friendly’ as for example chess is (but that’s another discussion).But OGS is not very beginner friendly I would say. If someone is curious about Go and wanna try it out, a quick match finder is way more intuitive place to start. But there’s no games almost.

Further, in terms of the ranking we have, these things do pose a problem. For example, I recently played a handicap game, how does that count? Was it really a fair match? I don’t know. While it’s nice to have the option to play a handicap game, should it count for rank? Then, creating custom games is not straightforward either. It took me a while to understand how the rank restrictions worked.

But I’m not suggesting to get rid of custom games. They are nice, absolutely, but as Go is a competitive game, then there should be some focus on the fairness of challenges. If I want to rank up I would make a game that is only for people worse than me, with analysis on, and such things. It’s fine for those who want to play someone stronger, but do I get stronger by playing those worse than me?

2 Likes

I like custom challenges. Quick matches shouldn’t become the primary. There is way for both to benefit.
Separate pool of players is a problem - it slows both down.
While quick match challenge waits for other quick match challenge, it should be possible to accept it like a custom challenge for players who meet criteria. Quick match will become faster.
If custom game meets criteria, it should be auto-accepted by players from quick mach pool. Custom games will become faster.
Quick match finder should have even more simple preferences - its for people who don’t wish to think about choosing. Those who care about settings should create custom challenge.

2 Likes

Sorry if this had already been discussed but couldn’t the quick match button just make a custom challenge with some default settings which would then appear in the custom list?

3 Likes

The answer is yes, when you want to align ranks and handicap stones, which various servers do. If if two ranks players are about two ranks apart then (iirc) a 2 handicap stone game should only slightly favour the stronger player playing White to win. Similarly with other rank gaps and handicap stones up to 9 stones/9 rank gaps.

That’s the idea anyway.

It was in my wall of text and in the linked post earlier :stuck_out_tongue:

I think it got brushed aside slightly :slight_smile:

But it’s good we’re circling back to it and combining ideas.

2 Likes

Not sure I understood you here. Do you mean that, if I win against someone who is 2 stones stronger than me, without handicap, I get more points on my ranking, than if I win with handicap?

I agree with this. But, I’m skeptical if all combination of settings should count for your rank. Especially those settings that has an outside the board effect. For example, estimate score and analysis can even out skill difference. A 10 kyu is probably better at reading than a 14 kyu, but with analysis mode, one can - depending on time settings - compensate for this difference. One could argue it’s a crutch that eventually will harm more than benefit learning, but it’s also tempting to quickly check if you read a life and death situation correctly. Turning ranked off for such games seems sensible to me because your rank should as closely as possible reflect your skill at go, not your ability to utilize such tools. That’s one of the reasons why I think ranked games and custom games should be separated, with ranked having a set of preset settings. However, whether all settings should be compatible with ranked is perhaps a different discussion.

That’s what I would expect to happen, if the rating algorithm was a bit more ELO like but…

and

Because there’s this sliding window in the Glicko2 system which is looking at your most recent 15 games/window of 90 days it’s not really clear how a particular win/loss changes your rating.

There’s some small amount of explanations (and further links) here in the docs Ranks and rating · online-go/online-go.com Wiki · GitHub, and more attempted explanations on the forums, but very regularly questions come up on the forums asking why rating is going down after a win.

etc

So the most I can confidently say, is that the handicap in the game factors into the calculation. I believe it’s supposed to be that for each handicap stone you have it treats you as being a rank stronger in the calculation (someone correct me if I’m wrong).