Ah thanks for the clarification. I think we can safely say there is no handicapping. But I’m trying to point out something different. Let’s say you have climbers, droppers and people who are around their natural rank. Let’s say those are regulars. Climbers are people who are rapidly climbing towards their natural mmr. Droppers are the oppositie.
Team A and team B are pitted against each other. Team A has 4 regulars and two climbers. Team B has 4 regulars and 2 droppers. All have the same-ish mmr though.
I am in team A and I am a regular. For four or five matches I’m in the lucky composition similar to team A. I kind of get carried to above my natural mmr. Now the tides turn. I’m with 5 regulars vs 5 regulars and 1 climber. I drop. This is how droppers come to exist. And without the handicapping this can exist right?
Climbers can get carried down if they are unlucky because they have more droppers or less climbers in their team than the team they’re playing against. This can happen despite there is no avtive handicapping present. Effectively you can be handicapped if luck is not on your side for a few matches in a row. Giving you the perception the system is being meddle with.
I can imagine the above scenario is least present in GM but increases the lower you go. Being the most problematic in diamond rank, where ganesense and mechanical skill start to matter more. The more a season advances the more droppers and climbers get filtered out and the economy gets a bit more balanced.
Do you think this can happen? And going by this theory, would it be safe to assume it’s always best to start mid season when the biggest part of droppers and climbers are where they should be?
“The process of rating players can be compared to the measurement of the position of a cork bobbing up and down on the surface of agitated water with a yard stick tied to a rope and which is swaying in the wind.” - Arpad Elo, the inventor of the Elo rating system in chess.
The effects you speak of will cause the uncertainty of a given player’s rating to be higher. Basically, if a player’s true rating is 2500, his rating at any given moment can be between 2250 and 2750 in normal cases and 2000 and 3000 in extreme cases. Any given player will tend to bounce (over the course of hundreds of games) between where he is nearly guaranteed to win, and where he is nearly guaranteed to lose. I do a mathematical analysis of this kind of movement here: https://us.battle.net/forums/en/overwatch/topic/20759379305
I don’t have any theoretical way of determining the best time in the season to play. Because there is no reset between seasons, time in season tends to be not have as dramatic an effect as you are expecting. My experience is that things are strange during placements, and for the last week or two of the season (in gold). When I was in silver and seasons were three months, the last month or so was bad. When I was in bronze, things tended to be bad all the time. I don’t know about the higher tiers.
Though in chess it’s always a 1 versus 1. Let’s say you’re in a team with 5 players that are natural 2500s but are all bobbles up and you meet them at 2750. When you, are a natural 3000 (or a little below). In the competing team there are all natural 2750s. Everyone has the same mmr, the match is fair according to the system. Those bobbled up players will now drag you down.
I will read your post on the statistics asap though!
So last time you made a post I replied with a data point for you. I have more. I’ve gone up about 500 SR from placements alone. I can’t play often but I do play other modes and read/watch/study. Just anecdotal data for you.
I do have a question about the new Diamond+ system. I’m curious, if the MM is doing its job, is there any functional difference between MMR and SR in that tier?
Also, I have a Reddit post I think you’d like and I would appreciate your thoughts on. The address is https :// redd. it/ 7y87hy
It was implied, but I forgot to say it: In chess, there are only one player on a side, there is only one map, only one choice for your pieces, but it is still a hard problem to rate people accurately. In Overwatch, things are much worse, so the problem is much harder, for reasons such as those you describe.
Could you post it in https://us.forums.blizzard.com/en/overwatch/t/season-9-placement-results/12329 (and include the information in the first post?)
I’m gathering much placement data for analysis.
When a player is decayed (which only happens in Diamond+), SR falls below MMR. In fact, it is the low tiers where the difference between SR and MMR is usually a distinction without a difference. I think you may be under the impression that MMR is a summary of a player’s statistical peformance. Read that section under popular myths in the original post.
I read it. I’ll make some on topic comments. You can ask if there is some point that you want me to expand on, or some point of yours you would rather that I didn’t ignore.
Regarding the “Market Based Browser System” a.k.a. group finder. This is a complex feature to set up. And if it isn’t really well done, it won’t be used. So it has to be done right. Jeff indicated that he would like something like this. https://www.youtube.com/watch?v=RofvPl1zIU8&feature=youtu.be#t=16m but because of the difficulty I expect it to take a while.
However, right now semi-pugs like this would be placed against more experienced groups of similar rating, and would tend to lose, causing the groups to fall apart. Maybe a separate queue for semi-pugs (or people that hit the “stay with group” button)? But that divides the player base and may be gameable. Tricky.
A strong leader could overcome the problem of pugs falling apart at the first loss, but that would be better served by a solid in-game clan system. This has a different issue. Organizing large groups of people is challenging and is a skill almost entirely unrelated to Overwatch. I have significant experience with it from my Warcraft raiding days. If I had more time, I would probably be doing it now (on a discord channel or something), but it is not something anyone can do on a lark.
And what about a group of people that like to play together (real-life friends or good synergy in-game) who are not (or have stopped being) the same rank? Another desirable feature is that there be a team queue. The team would have a rating, not the individuals. Members can join and leave the team, but with reasonable restrictions. This is probably the lowest cost to benefit idea.
I’ll put the results in the post you mentioned and I appreciate the link to the Youtube video…
Regarding your other thoughts…there was a 9 page paper in the middle of that post that maybe you missed, because I specifically address a Team Queue in it, among other things you mentioned…? The Reddit post itself says almost nothing…
I’m on vacation, and can’t actually play overwatch, so I took the opportunity to do a rewrite. I lengthed the TLDR (it needs its own TLDR now, I fear). I rewrote the “Popular Myths” section to address more specifically some of the stories that run around the forums. I rewrote the “Season Transitions” section as I finally think I understand what causes the high volatility (it is accounts that are not played much). I made minor edits throughout.
I wasn’t going to say anything, but since you’re rewriting it anyway, I’ve always thought it would be easier if you used the “hidden text” functionality as it’s such a long post.
Section 1
Section 1 body
[details="Section 1"]
Section 1 body
[/details]
Edit: This would be great to make the references easier to find.
Theory 1: MMR is determined by statistical performance and games are fixed before playing it. I mean matchmaker arranges win or lose for you. If your statistics good you get fake wins, if its bad you get fake lose streaks until you placed correctly. Generally you cant change the destiny of the game. But these fixed matches not happens all the time.
Theory 2: Matchmaker gives you win streak for you to learn high tier gameplay and then it gives you lose streak. If you learn that high tier play you able to climb again but this time for real not artificially.
I played 1300 hours competitive. I’m sure there is a forced streaks. Just dont know the reason behind it. There is a some sort of pattern. I think devs are lying to us. They dont have to reveal everything.
Biggest flaw of the system is there is no role select option and matchmaker not account this. So you can get 5 mercy one trick in your team but matchmaker still thinks your winning chance is %50. Its obviously not %50.
Haven’t been here for a while. Nice job on this post OP. It helps.
The Match making is quite horrible right now, lately I had waiting times of 3 minutes instead of 1 minute at low plat (when everyone is supposed to play from 9 pm in EU to maybe 11 pm). Either there are less people (players on the PTR training on new hero, end of holidays) or the MM has changed.
Bliz, it is time to make things visible so we don’t spend time on this trying to show you it does not work very well. If you actually fixed things on it (remember OTP mercys at diamond+? MMR did skew rankings at that time, staying skeptical about MMR is reasonable), it is already admitting it was broken. It still is for whatever reasons. Maybe the MMR system is not that faulty, as the toxicity is so high, players commit less in competitive, so many smurfs everywhere. Still it is your job, Bliz, to do something about it.
Hidden matchmaking rules are bad by design, nobody would trust that in any real sport, or any big decision making system. Clarity over cleverness principle must apply. Hiding / denying problems is not progressive at all.
It’s not bad, it’s just a security rule. If you allow a system to be visible, some people can find an hole and exploit it to freely clim. I see their MM as an industrial strategic secret (like google search engine or nutela receipt)
It’s not bad, it’s just a security rule. If you allow a system to be visible, some people can find an hole and exploit it to freely clim. I see their MM as an industrial strategic secret (like google search engine or nutela receipt)
Again, I will explain. What you describe is “security through obscurity”. Check wikipedia. It is indeed bad.
We need “security by design” or “open security”. This is the only way to stay secure when going on a large scale. Hiding does nothing, and when it gets discovered it is a disaster for a long time.
After reading those article, my description is “security by design”. The core project is only visible to a redistricted team (Devs) and us, user of this system know the basic rule of it. After we are customer and i don’t know a company who give their advantage (techno or other) to their customer. They explain but never give details.
I’m going to stop here. I don’t want to pollute this post but i’m free to exchange on private or on a dedicate post on this topics
It depends on your definition of “basics”. Implementation can stay hidden, but in software, API is indeed public. Right now, we don’t even know all the metrics of MMR which are used and in what proportion.
For instance you buy food. There are laws that imply that they put a description of what is inside. Not the recipe, but the ingredients and the proportions. We don’t even get that with Bliz. You can’t say it is “security by design”. Hidden things all along. With bliz making food, you could get intoxicated and you wouldn’t know why (well many players are already toxic lol).
![:wink:](https://d74zh8edaijub.cloudfront.net/en/overwatch/images/emoji/twitter/wink.png?v=5" ":wink:")