Time tag 34:25 : prediction of “average” players the graphic was just funny as all hell when I saw it. It shows extreme climb followed by extreme loss.
This dudes model predicted win loss streak after an average player plays a bunch of games.
I feel good about this in a way we can finally say it’s not anyones fault. Now obviously the “good player” just continues to move up… but I have no shame in being average, but I have issues of being blamed for loss streaks based on observations and patterns that I see.
I think the next chapter is “why”. How come he could predict it with fake players and how come it happens to real players???
Watched this talk years ago. A few issues. Mainly, the template doesn’t correctly mix teams and the estimates for team performance are weak.
Basically everything elo/trueskill tries to do (match players of skill bracket with other players of skill bracket) is what random SR system could do. No reason for mmr except to force close matches, which is a bulk of his talk.
He goes on about player psychology, engagement, feel, etc. All of which is emotion and nothing to do with raw competition. Raw competition needs uncertain elements. You can’t rig and tamper all of the uncertainty away. It might feel bad to see streaks and stomps, but those kinds of upsets are part of the contest.
This leads to rigging lobbies for entertainment and profit, not legitimate design for esports.
It’s not as technical as I would like, and it’s ethically disapointing.
If you were climbing the chess elo system online, the matchmaking algs wouldn’t cherry pick your next opponent based on hidden analytics of your game play. If you’re 1800 going for 1900 it would search and randomly find you something in that range to play vs. It wouldn’t assess that you’ve been accurate with your Bishops and find an opponent who is really good at shutting down Bishop attacks or finding another opponent who is also good at Bishops to mirror you.
That’s where the chess elo and microsoft trueskill systems break down. There are several papers that show how poorly they do in team settings and several alternatives have been proposed for team vs. team. Furthermore again, it doesn’t matter how well you can rig with bigdata, the ethics of competition state you shouldn’t/can’t rig. So that’s still an open problem.
In regards to the players experience after many games I would say that it feels like the opportunity to learn something more & become something more is taken away over the hours spent going through an Up & Down cycle.
I feel something changes that forces the players stuck in these patterns to take on different responsibilities & learn different means for success than their own technical skills - which the MMR might use against them.
I have a lot of thoughts that could explain better how I mean but I understand it might be best to keep it simply at “this pattern is / can be predictable”.
You might be getting at the NFL theorem, something Josh Menke’s talks can’t absorb. It basically says forcing the micro “closeness” will lead to “distance” at the macro level. When you conflate closeness with quality (a non-competitive and unethical interpretation), natural progress and progression on the ladder scale will suffer. Ranks will become less meaningful, as the integrity of those global metrics deteriorates.
mmr rigging deteriorates the overall ecosystem. and based on what i saw they aren’t even calculating team-level metrics properly (not using harmonic weights and mixtures). It’s the kind of rubbish math that works in entertainment, when lives aren’t at stake, but not in real professions.
Another aspect to all this… hate to say it… alt accounts.
All this Math… but its all based on an assumption people want to win and would do any means to win. Take out the means to win and now you got even more skill / mmr dilution.
Where’s the other group? The folks that say Win/Loss streaks are based on the player? Would like to know their thoughts.
Are you sure you are even average though? Not trying to be toxic here but you were literally not average last time I checked. I don’t know your rank now but you’ve been complaining about matchmaking since you were bronze/silver, pretty sure you did not think you were part of the “bad players” back then even when you were literally far from average.
When they make the team, they avg the mmr for the whole team instead of comparing the roles of the team. (When determining if the match should be thrown out) - solo queue is a non factor in THIS regard.
That makes for a more balanced match such that an individual can exceed the performance of their peers/counterparts and have some level of carry potential. If you removed that you’d just be forced into 50/50 matches all the time which would result in either extremely close matches all the time or one team losing all the time because Overwatch players, in general, don’t play competitive, competitively.
Alts corrupt the ladder according to known resampling and dissipation laws. They introduce SR to where it doesn’t belong, which forces corrections to propagate (until dampened/absorbed). It’s a tax on main account progression not to mention pay2win exploits like wintrading or gatekeeping or just disposable gameplay in general.
So yeah, the elo skillcurve stuff doesn’t work in a live ladder that doesn’t reset ever in 5 years, gain alts by the day, and fails to average player strength into the right kind of team strength.
Not only that, but in the talk, they use arithmetic mean which is WRONG.
If I have 2 teams of 6 pumps to drain a swamp, do I “average” the power of the pumps? Of course not. It’s a parallel operation not a series operation. That’s what skill “ratings” means comes from the word “rate”.
Enough with Josh Menke. that guy is working at Riot right now may as well be a plant cause he’s gonna break the matchmaking over there like the clown he is
Because most of the complaints come from people who have deluded themselves into believing they are better than they actually are. In that graph he would be part of the red line.
Your discrediting the issue by discrediting the person.
I think object of the thread was not to complain but to find reason beyond what is told to them by misinformed and other delusional people. Fine, that person was Bronze what about the NEXT person, or the person after that - soon you just go on a discrediting campaign instead of tackling or bringing attention to a real issue.
Also mind each person on here maybe represents X many people who dont even bother posting on the forum. So to establish if ONE person is below average, is almost inflammatory to THAT person and pointless to the issue. It only serves the “high elo ego”.
Here all we’ve been seeing and hearing about loss streaks are, “YOU” are the cause. But back in 2016, the dude that built the comp system projected radical win and loss streaks AND its only predicted when the predicated match quality is low.
Please, look at the video, its plain as day. Tackle the issue, how can we improve the match quality?
Literally the profiles of 99% of the people who come to vent on these forums.
You, Eth, Cuth, etc to name a few.
If they are below the 2350 median they’re objectively below average and honestly if we are talking about the people who actually care and consume OW content / aren’t completely casual the average is a bit higher.
