Wednesday, January 30, 2019

At the end of a tank battle in World of Tanks you get a window showing the contributions of the 30 players involved in the battle, in terms of kills, damage dealt, and experience gained. It is very obvious that within a team there is a large variance in these data, from the guy who died before he could achieve anything, to the guy who achieved multiple kills. However if I look how much talk there is of "carrying games" on the forums or on YouTube, I must say that I think that the idea is somewhat spurious, at least in as far as we are talking about random battles.

World of Tanks now has more different play modes than it used to have when I first started playing. Most of the new modes involve you playing with some sort of non-random team, a clan / guild. In that sort of organized team play, it is perfectly possible for one team to be consistently better than another. Thus if as a player you play a lot of those games, you can have win rates which are very different from the statistical 49% of wins (due to about 2% of games ending in a draw).

But if you play only in random battles, your win rate will be somewhere around those 49%. Mine always has been, as I never joined a clan. This is not just a reflection of me being an average player. But it is a mathematical problem: As there are 29 other players in a battle, a single player would need a very extreme performance in order to outdo the contribution of all those other players. If you spent a full day playing battles in which you deliberately drove off a cliff at the start of every battle, you statistically would still end up with a win rate of above 45% in spite of never having contributed anything to your team. Thus in the other direction, if you were a really good player and played perfectly in every game, you would still have difficulties reaching a 55% win rate.

What applies very much in World of Tanks is a phenomenon of selective memory. If things go very bad in a battle, they do so very quickly, and are then quickly forgotten. If you watch replays on YouTube, you'll see that the large majority of posted videos are victories, or at least losses in which the player posting the video still did very well. People just usually don't post the videos of the game in which they decided to go in one direction which just happened to be the one where the enemy was making his big push and they killed after 2 minutes of battle. But these games happen all the time, they are just not recorded or remembered.

If I hadn't a mod running that showed my win rate at two digits behind the comma, and showed me how many wins I need to do more than losses to reach 49% win rate, I probably wouldn't be able to tell at the end of long session whether my win rate had been above or below 49% that day. Without tools, humans aren't good at probabilities and statistics, and 45% doesn't feel all that much different than 55%. Yesterday evening I had a 100% win rate, but I only played 3 battles. But I know that as long as I stick to random battles, over time I will win "about half" of them. And that is okay with me, because there are battles in which I am having fun despite my team losing. And fun is the greater purpose of playing games.