# Statistical proof TrueSkill failed me. 343 please take note...

Out of curiosity I took a look at the career K/D ratios of the last 100 people I was matched with. In all of these games I searched alone. The majority of these games were BTB and Team Slayer with a little bit of Team Objective, Action Sack and BTB All Stars in there too.

[o] My overall K/D ratio is 1.49 and has remained pretty much steady for months.

[o] The average K/D of the past 100 players I was matched with was 1.11

[o] 34 of these players had a K/D of under 1.0

[o] 13 of these players had a higher K/D than mine, 87 were lower.

[o] The highest was a Noble with 2.13, and the lowest was a Private with 0.49

I think these findings pretty much prove that TrueSkill didn’t do a very good job of matching me up with players of equal skill. My K/D is not exceptional at 1.49, and the playlists I join are always very full, so there’s really no “not enough population” excuse.

Without a constant challenge, the game becomes less fun. Without fighting people more skilled than myself, where’s the opportunity and necessity to improve as a player?

Do you guys feel the same way? I’d love to see someone with either a much lower or much higher K/D than mine to do the same analysis of their own recent matches.

You are aware Trueskill only takes into account W/L and not K/D, correct? I believe you are trying to assume players with a high W/L automatically have a high K/D, which is highly fallacious. In order to test Trueskill, why don’t you test each player’s W/L?

Trueskill never promised accurate skill match-ups. It only ensures the most accurate matches out of all possibilities. Taking this into account, if you want to prove Trueskill wrong, you have to prove it didn’t match players up as accurately as possible, in which case you would have to determine the past 100 player’s mu and sigma from their W/L. You would do this by looking at every game ever played in Reach of the past 100 players. And even then you would have to know the mu and sigma of all those players in all of those games, and then you would have to go into more player’s game records until you know the W/L of everyone in Reach. In short, you can’t disprove Trueskill as a bystander not employed by Bungie or Microsoft.

But perhaps your problem in finding even matches is the fault of the population numbers, or gametype/map design. Or maybe it is an issue with yourself in being more aware of the games you play. Sorry, but I am not understanding the leap of logic you use in narrowing down the problem of playing “even” games and assuming it is the fault of Trueskill.

> You are aware Trueskill only takes into account W/L and not K/D, correct? I believe you are trying to assume players with a high W/L automatically have a high K/D, which is highly fallacious. In order to test Trueskill, why don’t you test each player’s W/L?
>
> Trueskill never promised accurate skill match-ups. It only ensures the most accurate matches out of all possibilities. Taking this into account, if you want to prove Trueskill wrong, you have to prove it didn’t match players up as accurately as possible, in which case you would have to determine the past 100 player’s mu and sigma from their W/L. You would do this by looking at every game ever played in Reach of the past 100 players. And even then you would have to know the mu and sigma of all those players in all of those games, and then you would have to go into more player’s game records until you know the W/L of everyone in Reach. In short, you can’t disprove Trueskill as a bystander not employed by Bungie or Microsoft.
>
> But perhaps your problem in finding even matches is the fault of the population numbers, or gametype/map design. Or maybe it is an issue with yourself in being more aware of the games you play. Sorry, but I am not understanding the leap of logic you use in narrowing down the problem of playing “even” games and assuming it is the fault of Trueskill.

Daaamn right !