Yes, absolutely. The preferred setting should just help the algorithm decide.
2 Likes
I feel like that’s how’s it’s supposed to be at the moment right? You have prefer or require, and with prefer it sounds like it will choose your preference like Chinese rules or Fischer but it might still give you other settings.
I think though it’s better to have the automatch times standardised. It’s already confusing enough as is to either know or explain to someone what time settings they might get in automatch cf
Nevermind if you can randomly get 11mins+4x7s because it’s within some tolerance of something you want. It’s just a bit chaotic to keep adjusting to all the different time settings (for me say). It’s nice, I think, to internally have a feeling for how fast you need to play with a certain byo-yomi or a certain increment, and it gets harder to do that when the settings could be different every game.
Currently one can prefer Fischer, byo-yomi, etc., but I can’t say I’m okay with 20-30-5 (20 min + 30 s * 5) and 5-20-3 (5 min + 20 s * 3).
Good point, but then you can’t auto-match players with custom challenges.
1 Like
Well it doesn’t have to be a prefect two way thing.
For instance if the automatch challenges were shown, some with a strange time preference might accept one of the standard ones if they can see someone around their rank in the automatch queueing and accept it in the custom slot.
I think initially it sounds like a good idea to have the automatch pick random offers that are close enough, but I think consistency is probably better and worth pursuing for simplicity.
It’s possible the actual matching algorithm for matching in the automatch could be improved or it’s possible that it would work perfectly fine as is once we reach some critical amount of players using it.
3 Likes
Simplicity is still here with its standardized offers. Even more if you get the automatch offer appear on the list of custom games, participation is encouraged (appear only when someone used the automatch button, waiting for a game.)
I find great that players have different taste but it’s even more great if a procedure help to visit each other and test different settings. Much better as constraining everyone to a standard.
So a good automatch should detect similar settings around a standardized one which is not always obvious for a human (or too much time consuming). And give some information related to it (like estimated time per move for example) and a choice to accept or wait for another offer.
1 Like
I’m not sure. I do have a feeling that user experience is important, and knowing what time setting you’re agreeing to should be better for a user experience. It could be that whatever tolerance is used to match up different settings to each other is wide enough that it leads to more matchups being cancelled.
The simple solution that people semi-independently come up with on KGS or Fox or pandanet for automatch is a single time setting and I think it works well, it just needs a good user base and a good matching algorithm.
Currently we have a lot of options already in the automatch that it’s difficult to even tell someone what time setting they’re agreeing to. The best we can do is say (~30s per move) and most like it’s 20m+5x30s but it might not be for normal speed games.
(Not counting this like ruleset or handicap that you can also have preferences for)
1 Like
Sure. That doesn’t contradict my view of automatch. Players can be well informed, even better with that screen asking if the automatch selection is fine. It can give even more info (like the average time per move) instead of only the enuneration of settings.
I don’t think that OGS has to copy other servers formulae, we can do better. OGS at first should elaborate from his own players base which is not exactly the same (many coming from the world if correspondance) and who enjoyed to set their proper parameters in their custom games.
My proposition never debated on this, i mean it is not linked at all with how many default settings we have in the automatch itself.
It’s not about changing these, it’s about how agregating customs games into automatch frame.
If you think there are too many default setting, well no problem, let’s reduce them. I won’t go less as 1 blitz and 1 live myself.
1 Like
OGS does not have standardized modes of play. Every other platform does.
This is very true, there should be agreed upon standard modes of play.
Even huge platforms like Fox or chess.com (more than 11 million daily active users) have standard modes. OGS’s playerbase is much smaller than those platforms player bases but somehow OGS thinks they alone can pull this off. They cannot.
Yes, it definitely is!
EXACTLY. Ah this is so refreshing to read.
Very true.
Yes, exactly my point. It’s not just the ones you named, every Go and Chess platform I’ve seen has standardized modes of play (with the exception of OGS of course).
This is incredibly true. It’s not just a lot of options: there are 324 possible configurations. Literally 324. No standard modes. There is just a huge button saying “Normal” (whatever that is supposed to indicate, bad wording regardless of this conversation) and players click it and it’s a total Kinder Surprise Egg, you could end up with like any kind of game. Maybe it’ll be a 9x9 game with a canadian time control, Japanese rules and a handicap. Maybe it’ll be a 19x19 games with byoyomi time control, New Zealand rules and without handicap. Noone knows. Let’s spin the roulette wheel.
1 Like
I’m sure there are way more than 324 possible configurations technically. Only for different options of starting time, increment and max time of the Fischer clock in blitz there are total 8624 possible configurations.
But that number doesn’t really matter that much, does it? I don’t think there are too many people who don’t play on the site because they are confused by all the options. There might be a few, and maybe over time we lose a fraction, but I somehow doubt that if we only had 9 well defined game modes that we should then see a notable difference in player numbers. Unfortunately, making an app that people like to use is harder than that 
2 Likes
I was talking about the AF. The AF does not have options for starting time, increment or max time. This whole conversation is about the AF. It’s literally in the title of this thread.
Yes, it does.
That’s not the point. Not even close to the point. Actually, that is so far from the point that you couldn’t even see the point from there using the hubble space telescope.
Making apps is easy. User experience design is easy. Innovation is hard but doing something wrong that every other competitor already figured out is just being sloppy.
I mean you can repeat it really confidently, but you’re still wrong.
4 Likes
3 Likes
4 Likes
The correct way would be to finally do some analysis on the behaviour of players, and get the data on what people do, what new players do, how long people wait in queues and so on.
After reading this thread I realized the major challenge at the heart of this discussion. The problem is that OGS targets two demographics at once – live players and correspondence players, with them having completely different behaviour. Much of the problems highlighted in this discussion stem from the motivation to appease both crowds with absolutely different customer journey. Therefore, in order to improve automatching, we first need to rethink whether or not we want to split live and correspondence games. By “split” I don’t mean creating a separate website. I mean somehow highlighting those two different playstyles and funnelling people into the right one instead of merging them all together. Take a look at lichess.org They have “quick pairing” (=automatch), “lobby” (custom games for live games), “correspondence” (list of correspondce games). They don’t have an automatch for correspondence game, which seems like a minor feature, but in reality it clearly separates the attention of players, allowing them to clearly comprehend what to expect from a game. “Quick paiting” / “lobby” => live games. “Correspondence” => long games. We also have “Rengo” which doesn’t help in improving the clarity of the interface. I don’t propose anything specific yet, but I think we should consider this as a crucial direction for our brainstorming if we want to improve automatch, make new players stay on OGS and make old players, well, stay on OGS as well instead of switching to Fox.
1 Like
Just got done reading a bunch of posts, and thank you to @Groin and @Conrad_Melville for being consistently staunch defenders of user choice
As everyone else seems to be, I’m all for AF creating a CG when someone joins the queue, so that people looking for CGs can accept it, and also all for AF detecting a CG offer within parameters and automatically accepting it. This seems to be a nice QoL thing that wouldn’t break anyone’s experience, and might make some people’s experience better
What I’m against is trying to sideline uncommon but far from unreasonable settings. One form of this would be to unrank CGs which aren’t within AF parameters: I am against this. A more subtle form would be to mark CGs which aren’t compatible with AF: an indication in the game creation dialog that the settings are or aren’t compatible with AF might be a nice QoL feature, but a mark visible to others on the CG itself is something I’m very much against, as it serves no purpose other than to single out people who use more personalized settings
I always use CGs and never have trouble finding matches despite using NZD Rules, a rather niche choice. @Regenwasser gave an example of a 13x13 CG offer taking 20 minutes to fill, which in my experience is an absurd strawman: such a game would take a few minutes tops to be accepted in my experience; certainly not long enough that I’ve ever bothered to time it
3 Likes
Correct.
One of the problems, yes.
Not sure how much. There is a very loud crowd that wants unlimited match customizability in all scenarios.
The issue is that there are quite a few long time users that are opposed to the necessary changes. I personally will stay on Fox until we make progress on this matter. Which is a shame because I really like OGS and this forum.
Yes, but I’ve had a lot of trouble finding matches using the AF. So I stopped using OGS for playing Go.
Not sure what you mean by this. Could you quote what I said? Has been a while but does not sound like me because I don’t think I have ever used CG for matchmaking.
Here is what I was referencing
2 Likes
I have absolutely no clue how you could read that paragraph and then arrive at your conclusion. This paragaph is not talking about CG waiting times. That’s not the point of that scenario I outlined. The point I tried to make is that two people, that want to play the same game mode on the same server and are “queueing” at the same time, are not matched against each other.
1 Like
My conclusion is that it is an absurd strawman. I explained why I thought so: because I never have to wait that long to get my custom games accepted, and I’m pouring water on my offering by using an obscure ruleset many people have never heard of. Are you saying that you did significantly overstate the amounts of time involved, but that it was not a strawman? or that it was a strawman, but it’s not particularly relevant because said strawman does not affect the thrust of your argument? or that the times you gave were accurate and it’s I who have oddly short waiting times?
I’m not really sure how else to phrase this.
I’m saying it’s not a strawman because it’s not a point I’ve even made in the first place. You are just claiming I made a strawman argument and then started argueing against that. I guess that makes this kind of a double-strawman argument or maybe a strawmanception?
But let me try again to clear this misunderstanding up. The time frame referenced in that scenario is simply supposed to outline that no matter how many minutes both players are waiting they will never be matched against each other. That is the point of that scenario, that the players are not matched against each other. This scenario in no way or form tries to address average wait times of any game mode.