Greedy Goblin

Monday, January 2, 2017

1-2 more weeks please

I promised the League of Legends matchmaking rig proof for today and instead I just provide a peek:
As you can see on this relatively small sample, I can predict the outcome of the games very well now before the start. And no, teamscore is not "high_winrate_players - low_winrate_players". That would be just bad design. I keep you in suspense because I want two more things before publishing.

The first is refining the teamscore formula, this was created with an initial bias which merged two variables that proved to be not only separate but completely opposite (one increases win chance, other decreases).

But the bigger issue is a rig is defeated if it's exploitable. If I just create charts, convince a few people to not play, couple million players will remain. But if I show a way how to exploit the rigged matchmaker, I don't only provide undisputed proof (if it wouldn't be rigged, the exploit would obviously decrease winrate), but players who want to continue playing will use the exploit, ruining the game for those who don't exploit, forcing the company to stop rigging (or rather refine their algorithm).

The obvious solution could be simply dodging queue when teamscore is low. But the matchmaker has some anti-dodging mechanism:
This chart was created with my first, very bad teamscore formula. The point is that as time went on, I got more and more "acceptable" games with my winrate decreasing. It seems the matchmaker learns the patterns of queue dodgers and baits them, giving them games they take but lose. Remember, it can give you a good team and just give a better team to the enemy.

So my final goal is to not only prove the rigging (the first chart with the teamscore formula would be enough), but figure out how to increase my winrate without getting better and without intensive queue dodging. At the cost of 50 games with 30% winrate I figured out the fundamentals, now I have to refine it and win-win-win.

Within two weeks I come back to this topic with the exploit. Or I accept that the matchmaker is too good to break and settle with only the statistical proof.

11 comments:

maxim said...

Not sure I understand what teamscore is and how it can be crested before the match.

Not sure how you are going to prevent yourself from getting better without going full on M&S.

Not sure how the fact that some matchmaking puts you at a disadvantage is proof positive of "rigged". Imperfect matchmaking will happen. Especially after disbalancing stuff like winning streaks.

Gevlon said...

@Maxim: Of course you don't, because I didn't tell the formula

By playing only one, beginner-suggested, out-of-meta champion

As soon as you see the formula, you'll know it's rigged.

Anonymous said...

Whatever your formula may be I suggest verifying it with someone who has no idea what you are after. Reason being self suggestion ie. if you think you are likely to win you will be play differently than if you believed you are likely to lose.

Anonymous said...

Also of interest to you might be the historical mmr (match making rating) which lol used to apparently show before the current system.
Guesses are that current system was implemented on top of the old one and the old mmr was just hidden.
Now the question is, are you just trying to figure out the formula on how mmr is calculated?
Various websites such as op.gg can also show their version of mmr. They don't tell how they calculate it but perhaps they would tell you should you inquire it.

Gevlon said...

@Anon: and this is why figuring out how to break the system and getting high winrate would be the ultimate proof.

@Next anon: I don't think that MMR even exists. I believe games are created ad hoc including both unranked and gold players.

Anonymous said...

From my personal experience the MMR shown in op.gg has some relevance to what happens in the client. Ie. when it has shown to me 'your MMR is quite a big higher than the average in this division (higher by 2 divs or something)' -> After the next promos I have skipped a division.

Pheredhel said...

One thing I miss here is the definiton of a working exploit of the system.
That is something you should declare now, not when publishing the results.

As you use ranked, my suggestion would be to use the official metric of that system:
how high do you rank?. There is no toplist for dps, winpercentage etc...

The only thing that would get people to really see it as an exploit would be if you manage to raise in the rankings through it.

This however would also mean that queue dodge penalties work against your exploits (getting other to dodge would work in your favour again though).

So can you define criteria now that we can later measure the success of your exploit against?

Just having "some strange numbers" that do not affect any metric that players usually care for won't do much in my opinion.

Gevlon said...

@Pheredhel: obviously. The only working exploit is that can you elevate significantly higher than you could climb with normal play.

I'm not sure I can do that, because there are many countermeasures to protect the system. That's why there is delay. I'm afraid I can't do it and have to settle with just statistical evidence. But I'm trying and thinking how to do it.

Anonymous said...

Casino always win.
Make your own game

Gevlon said...

Of course the casino always wins. I have no problem with the casino paying double for red, double for black, but red+black =/= 100% due to zero being green. That's fine. But rigging the game so it gives red when buddies put on red is wrong. Actually it's a felony in a casino.

Anonymous said...

"But the bigger issue is a rig is defeated if it's exploitable. If I just create charts, convince a few people to not play, couple million players will remain. But if I show a way how to exploit the rigged matchmaker, I don't only provide undisputed proof (if it wouldn't be rigged, the exploit would obviously decrease winrate), but players who want to continue playing will use the exploit, ruining the game for those who don't exploit, forcing the company to stop rigging (or rather refine their algorithm)."

I have to say, that is a brilliant application of your "Be constructive instead of destructive" goal. No sarcasm.