Forced 50% Matchmaker Theory

It sure is technically possible.

But you are suggesting Blizzard utilizes some kind of Minority Report style predictive matchmaking where they use complex mathematical models to predict human behavior and factor in millions of things from the weather this morning,

to the medication you took, the websites you visited in the last 3 weeks and the talents you picked in your last 350 games.

When in reality we have no other evidence than simple MMR values being used for matchmaking.

Though it must be flattering for a maintenance-mode crew with minimal budget, that there are people thinking they have funds and resources to create such a system.

2 Likes

Machine learning ? difficult for you, not difficult for them…

you do realize they dont even have funds or interest to put in avatars beyond imperius release
or create real animated hero trailers anymore

let alone anything more complicated than that which costs money, like a gigantic machine learning system with endless variables

they cant even afford to keep the non-aram brawls online
they even shut down the ticket system and now you can only submit reports on the forum

holy molly…of course this is done not by the maintenance crew…you need specialists in the area…why am i even responding to this?

i dont think you understand what “maintenance mode” means, are you new to hots?

this game is working on a skeleton crew of only a few developers since hgc was shut down

they dont have money to create the fancy system you posted

they are doing the bare necessities by keeping the server online and releasing 1-2 heroes a year

Again, there’s really no evidence of the client using some kind of machine learning or counting endless variables to force you to win or lose a game.

We only have direct evidence that it matches people by MMR (and averages the MMR of all 5 players in the team in the match to be as close to the enemy average MMR as possible).

An exception is Quick Match, which patch notes show, also take into factor hero levels of the heroes picked (in addition to MMR).

They hired a guy who has a background in machine learning in his resume for the sole purpose of creating an “engaging” matchmaking.

And it doens’t have to be as fancy as minority report…omg…it cld be as simple as your last few games

FYI I don’t really care if you agree or disagree, you are free to have your opinion and likewise me too. Readers can decide for themselves.

You mean they hired a biology major to work on the AI (Kevin Gu)?

Again, I have no opinion one way or the other - I just prefer citing patch notes or developer comments rather than making up conspiracy theories.

(He also has nothing to do with matchmaking)

Take a breath please, you’re too green to understand how data works. The “evidence” you’re looking for is right under your nose on hotslogs. What you do not understand is how it used to calculate winnings and how it factors into the end result in.

Let’s make it simple:

  • Li Li as support on your team, Lucio on the other team.

Fundamentally Lucio is the stronger healer, obviously. This is a determinted matchmaking by its own. However, you can get deeper than this.

  • Talent pick rates.

Every player plays their characters different. The matchmaker can determinate if your Li Li will likely pick the LV10 heroic for several heals or the damage heroic. Simple enough, yes? The matchmaker will likely match someone with you who prefers to use the damage heroic to make your game harder. Hotslogs registers your winrate for every hero you play and their talent picks, this is not so different here but it will factor in the overall matchmaking in the rock-scissor-paper system in where you face characters who are strong against your ones in every mode.

This is also why you should try and check your mates profiles and enemy profiles at the end to see what their favorite picks are. If you prefer to play a non-% based damage dealer against people who will very likely pick a Cho’gall against you (because the data indicates that two players who always play together use Cho’gall as their pick) in 6 out of the 10 games, then you know how odds are calculated. It was never about certain truths (which is true in QM however because you are outdrafted by default).

But it doesn’t end here. Map win/lose ratio, available heroes for counter-play, preferred roles and skill level also contribute into the matchmaker. You can everything measure through data. Even without your guys “machine learning” argument it’s pretty easy to see how data is calculated.

check joshmenke linkedin

https://www.linkedin.com/in/joshmenke

Let me “machine learn” what you are goin to say next…“he was not in HOTS” ?

First off- “Forced 50%” is an excuse so as to not own up to your skill or lack thereof

Second- One-tricking is a choice, play more heroes.

True, he never worked on HotS.

He worked on D3 and WoW matchmaking only. Many years ago. And very, very early SC1.
Has nothing to do with Heroes of the Storm.

He has worked on practically nothing but Call of Duty and Halo for the last 5 years.

“Matchmaking and ranking for all of Activision Blizzard.”

That’s more like it

1 Like

That guy hasn’t spent a day working on hots. Get your facts straight. Consulted maybe but worked on this game never. He is an activision guy unless you count blizz games from the early 2010s

You are correcting his linkedin resume ?..really?

He’s just bending the truth.

I think the last time I heard something about conspiracy theory, I was really amused. This was about Joe Biden having an electronic device because he has a trial pending.