It would be faster to land a new player in the right place, but yeah, I don’t think it will be materially different after that.
If said player plays consistently the same… the system would be even faster… But we are human. The OW system is pretty good as figuring it out, and doing it pretty quickly.
Quick example… All of the Dafran unranked to GM series.
1 Like
Oh hell yeah, players are not… consistent. Groups of players even less so.
1 Like
They are making assumptions which ignore edge cases and in non-ergodic systems where there are feedback loops, edge cases end up being dominant influences over the sort of metrics that take large group averages.
Presuming that the system is putting out accurate estimates is the very source of the problem. It’s always the “given” in a statement where you find fundamental flaws.
I’m saying the “given” shouldn’t be given.
Accurate when tested improperly and the stochastic methodology smoothes out a non-ergodic system and gives an improper impression of the system.
My proof is in his stated methodology which is fundamentally flawed. If he’s doing what he says he’s doing then he’s improperly treating the data, period. …and that I can easily be diamond on one account and low silver on another in Overwatch 2. I proved it on Overwatch 1 by running placements for three different accounts over 2 days using the same characters and having 1000 points of difference. I have the raw footage of the matches online.
The greater point, however, is that he’s smoothing out and fundamentally ignoring feedback-based systems and the effects thereof.
The theory is wrong. The methodology is wrong. I’m telling you exactly how, where, and why. (as well as having evidence that it’s failing)
TrueSkill 2 has individual metrics to make up the difference. If there is regression analysis that tracks a network of players over time (as there should be) they can shore it up even more.
Is TrueSkill ideal? No.
Is it better than this current trash? Undeniably.
At this point I’m beginning to believe you either ARE Menke, or KNOW Menke and don’t like being told you might have seriously screwed the pooch and you aren’t even listening or trying to understand the ergodicity problem.
You’re just defending and telling me to pay you for a course on your greatness. Sorry, no thanks I’ve got better things to do. (like writing a paper on the fundamentals of entropy and information theory which also share the problems of ergodic assumption from the past century)
Ask your sycophants to pay you to worship at your alter. I’m not one. (or perhaps you are one instead of being him) If you’re him, however, and still capable of learning, then try to understand the problems of ergodicity by spending some time in understanding non-ergodic economics.
The last century failed to appropriately treat problems via statistics because of their lack of comprehending chaos and nonlinear systems. It’s all based on oversimplified ideologies that poisoned people’s comprehension of how to deal with systems where outliers can unrealistically skew. (including what you were taught)
If you’d like more information on any of this, I’ll give you sources of information (for free) that you can pursue.
1 Like
I’d like to see a citation for this claim as well…
Citation needed.
It is a simple test, and one which shows it works well.
The entire point of the maths behind a matchmaker is that it can predict the results of a matchup to a reasonable degree, and lo… IT DOES.
You don’t understand that the test shows it isn’t a problem.
You are like a person screaming that a bumblebee can’t fly, and thus the ones flying around you can’t exist.
And yet, they fly.
The matchmaker is putting out good predictions of wins, so that part of it works.
Your matchmaker has basically 2 parts.
The part where you check to see how close a match will be, and the part which is the solver to find the best match.
You can test “The part where you check to see how close a match will be” by getting it to predict known match results and seeing how often it gets it right by rank.
The solver part, well it is a solver.
This biggest problem with the ranking system is grouping. As long as grouping with a wide range of rating/skill level is Ok the matches will always be off.
Boosters would be handicapped hard by tight SR groups. Derank groups (that lead to boosters) would be handicapped. The “I deranked so I can play with my friend” would be mostly gone. The top 500 issue where they group with a low GM smurf would be gone…
Solo queue only or a very strict rating requirement for Comp would lead to massive improvements in matches overnight. But instead of focusing on making Comp serious and boosting the Esport potential of OW we keep the clown show when QP is perfect for all circus attractions.
1 Like
Whilst this is highly likely true in a hypothetical sense, people forget OW is a casual game with a small % of players who either take it seriously or are trying to prove something.
The group ranges are mostly fine, and the only issue is at the highest elos where min/maxing can be abused (see that YSZNA player who should be removed from the game and community (if you don’t know, look into him).
Because you can go like 5-1, 5-0, 5-4, 5-10 and you often have no idea if you will rank up or down or stay the same. And sometimes it’s 2-3 divisions which is weird even if it’s on an old account like mine that plays every season.
1 Like