EPTPatch 1.21.3 - Page 6
| Forum Index > BW General |
GrantTheAnt
7 Posts
| ||
|
castleeMg
Canada786 Posts
basically what i am asking is why is there no pre lobby connection test or screening to test latency between two players before they are put into a match? | ||
HaN-
France1920 Posts
So I think it would be a good solution to make it so that the matchmaking is more strict on matching players of similar MMR for these players and less strict for others. Another idea would be to allow players to have the options to choose what what they want to prioritize (MMR, Latency, Queues, Rematch) for their matchmaking. | ||
|
halomonian
Brazil255 Posts
| ||
|
MarcoJ
Germany146 Posts
On February 14 2018 03:52 GrantTheAnt wrote: Hi guys, This is a probably a good point to talk about the changes to the matchmaker. It was quite a large algorithmic change to the MM, which was done after consultation with a range of players - from the everyday players through to Korean pros. Let's back up a step and talk about the constraints of a SC matchmaker: LATENCY: "I want matches that permit TR16 low latency" QUEUE TIME: "I don't want to wait more than 30s for a match" MMR: "I want matches at the same MMR as me" REMATCH: "I don't want to play the same player twice" Now, in a game with a huge concurrency (like Hearthstone for example), we can achieve all 4 of those for the most part. In BW/SCR, outside of Korea, we need to be realistic about what's achievable and make sure we're concentrating on the important constraint. Outside of Korea, we can achieve probably one of these constraints, maybe two, depending on region. The feedback we had from the previous matchmaker was "we'd rather play much better/worse opponents if it means playing on TR16, and we'd rather not wait forever for a match". Although we cannot guarantee a TR16 game, I believe we have now delivered on the spirit of this request. We can of course continue to tweak, but we need to be realistic about the ask. Currently, we try to honor the first 3 constraints in the order listed above. I believe latency should remain first, which means we could tweak the other two. If we reversed the other two, we'd be saying "I want to play as low latency/high TR as possible with a similar MMR opponent. I'm prepared to queue as long as necessary for this." (And in some regions, in high MMR scenarios, this queue time could be 30 minutes or more, since we're essentially waiting for an appropriate player to logon). That being said, as always I welcome community feedback, so feel free to discuss and we can continue to tweak. On other issues - we're investigating the failure to join custom games. There's no repro from QA yet but it's under investigation. And as Matt said, we hotfixed the Korean input issue with the latest Windows version. Since this is the second Windows change that has affected Korean IME in recent months, we've also reached out to Microsoft to better understand the changes they're making in that area that may affect us in future. Thanks again for your continued support and passion for SC! To be honest i think there is a clear hierachy to those four constraints. 1. LATENCY: Doesnt matter how good your match is in terms of MMR, QUEUE TIME or NO REMACHT. If it lags its bad. Which TR is the least acceptable should be determined by veterans. 2. MMR: If we have stable LATENCY, but the MMR difference is to high the game will be frustrating as hell. Totally onesided, which hurts the loser and basically no MMR gain for the winner. 3. REMATCH: You dont want to queue several times against the same guy in a row. thats ok. 4. QUEUE TIME: Well its really just an issue imo, when you are willing to sacrifice a bit of the game quality you get. So lets say after 90s of queuing the other constraints get soften up a bit. (although: THIS NEEDS TO BE DOCUMENTED. So players who dont want to sacrifice in game quality can just re-queue) And last, I dont know how your matchmaking works internally but ofc you need some sort of game_quality_func(). So all potential opponnents who pass the hard constraints get valued. So e.g if the minimum TR is 12 but there is a opponnent available with TR16 and similar MMR as the TR12 guy, the game_quality_func() should return a higher value for the TR16 guy. I put a bit of thought into it and would use something like the following: max_tr = maximum TR that can be achieved tr = the tr that is expected to be played on for a given opponnent (if you have a probabilistic model use the expected_val E(x)) mmr_gain = the mmr_gain I can achieve vs a given opponnent mmr_even = the mmr I would get in an dead even game (opponnent exact same MMR as me) tr_metric = tr/max_tr mmr_metric = 1- | (mmr_gain/mmr_even) - 1 | game_quality_func(): tr_metric * mmr_metric of course you could add some weight parameters for either metric. But it basically comes down to constructing a model for all opponnent candidates that pass the hard requirements. @haloman: It works completely fine for me under wine-staging 2.21 with the latest patch (I played today). | ||
|
Ikirouta
Finland744 Posts
On February 14 2018 07:28 halomonian wrote: Have anyone noticed a severe improvement on lag? Also, how are the Linux bros doing? I heard the last patch made the game unplayable via WINE Works perfectly with WINE. They fixed it the day after. | ||
|
sM.Zik
Canada2553 Posts
| ||
|
Matt Sherman
14 Posts
We have just pushed out server-side changes to reduce the max MMR range. As always we look forward to your continued feedback. This will be something we will continually be tweaking and improving. | ||
|
TOIHOIs
41 Posts
On February 14 2018 08:25 Matt Sherman wrote: Greetings, We have just pushed out server-side changes to reduce the max MMR range. As always we look forward to your continued feedback. This will be something we will continually be tweaking and improving. Hello, i have a complain. I can no longer find games. Can you please revert? | ||
|
Sero
United States692 Posts
Anyway the latest change sounds great. Exactly what was needed. | ||
|
SchAmToo
United States1141 Posts
| ||
Harem
United States11393 Posts
On February 14 2018 08:50 Sero wrote: Why do people avoid rematches? We used to play the same people on iCCup over and over because there were so few higher rated players, but suddenly people have unrealistic expectations of a nonexistent player pool. Used to wait forever to find that one person too, and be happy to re them. Anyway the latest change sounds great. Exactly what was needed. Yeah, who doesnt want to keep playing the same terrans who ask from? and other questions while repeatedly doing the same bad builds and playing 500 games in a month. | ||
|
dark.matter
179 Posts
On February 14 2018 08:25 Matt Sherman wrote: Greetings, We have just pushed out server-side changes to reduce the max MMR range. As always we look forward to your continued feedback. This will be something we will continually be tweaking and improving. Thank you for your continued efforts! This game and community means a whole lot to me  | ||
|
ProMeTheus112
France2027 Posts
this has been exactly my experience with SC:R. Although I can't stand the slowness of all interfaces as well as the crapped SD display, the inability to play games one after the other on matchmaking is what makes me log off and not even log back in. Increased bugs on custom games with multiple players is one of the very worst things that can happen to the game, it needs exactly the opposite, increased ergonomy and efficiency. when will you add a simple draw option to get out of laggy ladder games? it has been 6 months, my expectations are so low, that I'm not sure if I'll ever be playing my favorite game again that much anymore. the interfaces need to be fast. i'm not gonna go like yay i love it thanks guys if you leave the game with such poor and innefficient interfaces as a "REMASTER" of Starcraft which always had really quick and efficient interfaces! Same goes for SD graphics. Make your moves or I'll keep saying blizzard sucks, sry. [so : log in bnet instant, options instant, profiles very fast as before, etc] (well, if you do it, I'll praise whoever worked on it, not the company ^^ i totally hate the company sorry^^) oh wait, my "goals are off alined", yeaaah PS: you still haven't fully given full functionality back to friends list, it won't show game names, I can see how this was the plan from the start, first break it completely and make it seem like it bugs out, then finally remove it and just call it "ranked" or "unranked" this stuff should have been working 99% from day 1 PS2: interface before, when you want to /W someone on a channel you are on, click them and ALT+W gg. now have to click stuff and then get stuck (must click some place else to switch back to chat...). It's way inefficient, not remastered. also you removed ALT+N which writes person name in chat line, always was handy including to /W. Also often if I make a game, type something in the game name, it will mistake my input for whatever such as leave the creation page for no reason, I'm not even hitting the "leave" hotkey when it does this.. did you test how easy and fast it was to make and join games before? the only improvement is when someone on your friend list is in lobby, that button, cool. (still slower than before : ALT+J type game name hit TAB type password, which you can't do anymore and instead have to click through slow interfaces). Even the game creation interface now takes seconds to load, and then would randomly bug out. PS3: the bottom lines of map are still hidden by the awkward and rather ugly new replay interface (option to use original replay interface wanted). Chat still is divided between groups of observers because you can't set the new observer type group to more than 4. It's debatable (?) whether its a good idea to disallow chat completely between observer group and rest I guess, but it would probably be better to have that ability. Or at least make it possible to set more than 4 observers.. PS4: problems with "SC:R" start when you click starcraft.exe. Everything is slower, clunky, often buggy and inefficient and not remastered. Make everything faster (as fast as before or faster), eliminate bugs so that there are fewer or none, more efficient and functional. Do this in august 2017. No game registering bugs, that was never a thing either and shouldn't be. The blizzard that made starcraft was known for putting games out "when they are done", that's why their games were great and loved, on top of amazing game design and etc. 99%+ polished. With so many problems and mistakes done from the start, as dev team in there you may be in a type of development hell, where you can hardly reach an overall polished goal. I dont like writing this stuff, I want to go away.^^ | ||
|
[sc1f]eonzerg
Belgium6965 Posts
On February 14 2018 08:25 Matt Sherman wrote: Greetings, We have just pushed out server-side changes to reduce the max MMR range. As always we look forward to your continued feedback. This will be something we will continually be tweaking and improving. Testing it a bit,it feels now like 2 patches ago,so all good,this is the good direction i believe,thanks for letting me play my favorite game again the way i truly enjoy it ,im also sorry for my hot blood,i asked for this several times about a different way to the autosave replays name,cuz is a bit difficult to find an specific replay ,something like player player map will be much better. | ||
|
Randos
Germany48 Posts
Thank you for all your hard work! Long live Brood War! | ||
|
iopq
United States1083 Posts
On February 14 2018 07:18 HaN- wrote: I think matching people with similar MMR is more important for players on both end of the spectrum: the very high level players and the very low level players, than it is for the other players. So I think it would be a good solution to make it so that the matchmaking is more strict on matching players of similar MMR for these players and less strict for others. Another idea would be to allow players to have the options to choose what what they want to prioritize (MMR, Latency, Queues, Rematch) for their matchmaking. Actually, it's the opposite, the skill difference between 1600 and 2100 is ridiculous, they should never play against each other 2500 vs. 3000 is much closer (Bonyth vs. eonzerg or something like that) | ||
|
MarcoJ
Germany146 Posts
On February 15 2018 23:51 iopq wrote: Actually, it's the opposite, the skill difference between 1600 and 2100 is ridiculous, they should never play against each other 2500 vs. 3000 is much closer (Bonyth vs. eonzerg or something like that) ofc bc you have a normal distribution in mmr /e: see here https://starlog.gg/en/leaderboard/graph | ||
|
Golgotha
Korea (South)8418 Posts
| ||
|
Jae Zedong
407 Posts
On February 16 2018 00:29 MarcoJ wrote: ofc bc you have a normal distribution in mmr With Elo ratings that isn't the case, not sure what formula Blizzard uses though. In Chess which uses Elo, the gap between a 1500 and 2000 rated player is equal to a 2000 vs 2500 player in terms of expected win/loss ratio. | ||
| ||
