Monday, May 20, 2019

Damned lies and statistics?

After a bit of a dry spell, I've been getting back into playing League of Legends a bit more. Now, I never really left, but it's also not really the same. Not really. I've not brought it up lately, but I used to be really into the game, and now, not so much. They reworked almost all of my own champions and now my old methods don't work. I do still have, Karthus, Sivir, Sona, Amumu, Singed, and Udyr more or less intact. I'm probably also fine with Malzahar and Anivia. There are probably some others I can still play. But gone are the days when I had a system mapped out to play almost any champion. For most of 2018, I'd pop in and play some bot games occasionally and not too much else. I'm still leaving Ranked play alone for now. It's been all blink pick normals lately.

That's all a preface to something else, but the "something else" in this case goes way back, much further than the rest of this. I'll use my blind pick normals as an example. Without checking, I think the story is much the same for other gameplay modes. Anyway...

Back when I was really into LoL, I'd frequent the forums. A topic that would come up would be complaints about unfair matchmaking, bias in the algorithms, and such. Mostly, this was for Ranked play, but that point doesn't matter very much anyway. The complaint was across the board and so was the rebuttal. Typically, some player would describe how he or she was playing well and trying to win, but kept losing and falling in the rankings because they'd get toxic teammates. And some voice of wisdom would respond to this with a story. A brief story consisting of premises and conclusions to those premises.

Let's assume that you're playing the game and trying to win. You pay attention, you work with your teammates, you do your best, you try to hone your skills, etc. Barring rare outlier circumstances, you're not leaving the game or disconnecting from it. Now, in the game are nine other players. There's some chance that one of them is a problem, a liability. A player might have egregious connection problems, might abandon the game partway through, might never show up at all. Or perhaps someone else might grief or might rage or troll teammates, compromising the game. The matchmaking algorithms don't stop those from happening, but they do distribute such problems evenly. In a random game, each team has 5 players and any 1 of those 5 on either team might be the toxic element. But not you. As you are not going to be the toxic element, this means your team only has 4 players that might be the toxic element costing their team the game. So, in a game where there's a single toxic element, it's going to be on the other team 5/9ths of the time and on your team 4/9ths of the time. Sometimes there won't be any toxic element at all, but those games are 50/50 for you anyway. But over a long enough time, that 11.1% will grind the average of your opponents into the ground. Some games might have multiple toxic elements, but that doesn't change this anyway. Initially, the matchmaking algorithms might be causing you to lose by placing you too high, and you'll lose more for a while until you settle into your appropriate skill level or whatever. But once that part is out of the way, if you're diligent and non-toxic, your wins will overtake your losses. It's statistics.

And it's a fine story. It's a damn fine story, actually. Eminently persuasive. Logically sound. It just happens to have the misfortune of being wrong. Oh, it's convincing. But wrong is wrong.

My win percentage in blind pick normals, the mode I'm currently focused on, is 49%. If this was over the course of 100 games, I'd put it down to small sample size. 300 games and still, it's easy to imagine that I'm just on the wrong side of variance. It could happen. Someone out there is going to be in the wrong part of the confidence interval. But I don't have 100 of these games and I don't have 300. I have over 1,500 of these games recorded, and my win percentage has been below 50% for years. It's a pattern that seems to have carried over to other game modes too. I lose more than I win.

I have my suspicions as to why this is the case, but the important point in this post is that it is the case. Learning why is another matter.