Rigged competitive system is the reason for toxicity

[Outside-11127]

@Violence-11154

You need to demonstrate why being matched against similarly skilled opponents is synonymous with “rigging.”

Because it’s forcing the odds, as per definition of rigging.

@Violence-11154

If the system is adaptively working against the player, you need to show that higher elo players cannot climb to their appropriate rank.

No. You just need to show it takes fewer matches in a rigged system vs. a random one to claim everything is fine. It needs to be shown that rigging gets people to where they belong more efficiently than random SR only system. But the Virial theorem says for large populations (distributed normally by skill), random is best possible. So current laddering is handicapped and not fair.

@Violence-11154

It’s not “forced” 50-50 odds.

It is as per dev statements and patents. See for example:

@Kaawumba-1133

The game uses MMR to determine whether a match is fair, and attempts to match teams such that the game is fair (each team has a 50% chance of winning).

That’s forcing the odds “for match fairness”, therefore it’s rigging as per definition.

@Kaawumba-1133

If the matchmaking algorithm determined before the game that it was not fair, the higher ranked team will receive less SR/MMR on a win, and lose more SR/MMR on a loss.

Trying to determine the outcome of the game, and rigging for it, before it’s even played - is nothing like any other ranked environment. You can assign pbsr and sr bonuses for deltas incurred by randomly shipping two teams around an SR margin. No place for MMR except to rig “for fairness” - which makes the overall laddering process anti-competitive and unfair.

@Violence-11154

It’s giving you 6 fair coins against another 6 fair coins.

That would be random SR naive matchmaking. With MMR rigging, it’s cherry picking coins that are biased to fall one way or the other, and shipping 6 of them together to balance the other 6. It’s rigging the match outcome to be 50-50 by picking biased coins.

@Violence-11154

Yeah–just like defining the parameters of a fair chess board.

Chess is a two player zero sum sequential combinatorial game. It’s state space doesn’t change with balance patches, and it will soon be weakly solved (for example, a proof by induction on base games that white can always force a win, which is greatly suspected).
Chess is in a simpler complexity class than Overwatch. OW is an N-player stochastic differntial game you can transform to a fixed base game with perturbations and adapt epsilon Nash strategies for via e.g. pointer networks or quantum gans, like alpha.star.

@Violence-11154

Fair competitions pit equally or near equally skilled opponents against each other. Fair competitions don’t randomly assign players to matches.

They randomly seed or have you qualify for a bracket, then let you (and/or your team) progress naturally through the ladder, brackets, tournaments, etc. Even for single player ranking systems (like Chess), they use only 1 metric (elo), rank you accordingly, pay you out accordingly, and matchmake you naively around your current elo projection. If Hou Yifan is climbing a new account in Chess, the matchmaker doesn’t analyze her performance and cherry pick her subsequent opponents. Who she faces on her climb is entirely random around her elo and winrate. And there is no adaptation to how she is doing or what her gameplay was recently like.

@Violence-11154

57,289 accounts at 3000 SR are required to move the mean from 2266 to 2500. That’s 46% of the total accounts sampled initially. You then have to make the assumption that every single one of those accounts is played during each season. That is, you need a massive number of alt accounts playing at the same elo during the same season for negligible mean shift.

I did a rough calc and also got 57k accounts to move the population mean 250 sr. Where are those numbers from? Who cares…because that’s not the issue. If you are attempting to climb through a bracket of 7% of the population (by player % skill), but it actually contains 11% of the accounts, you’re playing a lot more matches to sieve through. You want to play against diamond players and outrank them, not diamond accounts, which could be over-represented by master players parking alts over and over, each one having a constant-sum amount of SR to distribute, imparting geometric amounts of additional effort for anyone attempting to ladder through.

@Violence-11154

Again, let me know if you’re actually going to present anything other than personal anecdotes.

Have you shown math to dispute the following?:

@ȐĔƆƐĨPŦŜ-1693

Every alt is initialized with ~7200sr to dispose of. That has to be absorbed by the ladder population. That’s a lot of constant-sum inflation. It’s potential energy that gets converted into kinetic (zero-sum) events as you win/loss. So it has to dissipate into the bath.

The issue is the scaling. Every alt can basically tax ~O(k(k-1)/2) neighbours with extra work, who in turn adjust their neighbours, with O(ln(k)) less but still extra work. It doesn’t equalize, it amortizes. It raises the temperature of the bath - which is small, sparse, and lowpop during offhours.

As well as:

@ȐĔƆƐĨPŦŜ-1693

The new players distribute normally. The alts almost surely don’t.
And both enter near the median, imparting disruption until their final destination.

As well as:

@ȐĔƆƐĨPŦŜ-1693

number of accounts > unique players

If you bloat the distribution with repeat accounts but label things by % players, you are imparting a false SR mapping. No person should be allowed to gain, lose, bank, trade, earn, dispose of duplicate amounts of SR without some zero-sum recalibration to the % players.

And he goes on to give example using race car drivers and race times:

@ȐĔƆƐĨPŦŜ-1693

He would be top 1% of racers but his races wouldn’t account for this. This top 1% race time would be disproportionally represented, i.e. showing up more than 1% of the time.

If you plotted the race times you wouldn’t get your 1% anymore, because this “alt” registered 1% data more than once.

So you really can’t have alts + rigging + 0 resets and call it “fair” laddering.
I have to side with Receipts because the logic holds. His examples are extremely conservative (they argue around a data-free base case). Which is the best you’re going to get unless you can show actual data.