Microsoft's Trueskill System

Recommended Videos

Paragon Fury

The Loud Shadow
Jan 23, 2009
5,161
0
0
Ok, lately I've been thinking about Microsoft's Trueskill system, the basis for most all Matchmaking services on XBL. For those of you who don't know how it does what is does, basically it uses a modified Bell Curve to try and determine where players fall on it to make matches.

Most people would agree its a flawed system. Yet I think the basis of most of their complaints falls to something even more basic; the Bell Cruve itself. It uses a single curve, where I believe a Double Bell curve would've been more appropriate. And why is it inappropriate? Because the single curve relies on there being some defintion of an "Average" or "Middle" player; which I believe cannot exist in any game, much less a competieve one.


Now think real hard while going through this. Think about all the games you've ever played. No lying to yourself either; can you, anyone you know, or played with, be objectively described as "Average"? Is it even possible to have such a definition? If you really think about it, it isn't - you're either good or bad at a game. There isn't any middle ground. You can't tie every game; you can have a perfect 1/1 K/D every game. Not everyone wins all the time; nor does everyone lose all the time. You're either good at a game the majority of the time, or bad at it the majority of the time. Which is why a Double Bell Curve is more appropriate; it seperates the two from each other, reflecting this difference.

While yes, it is perfectly possible to have different levels of "good" and "bad", the two don't cross.

And while no, I don't know how I would go about fixing this issue, or dealing with the complication of parties, I do think that starting with a more acccurate basis from where to start would definitely improve all MM systems in the future.


In addition, this not intended to be critical of people considered "bad" at a game. I believe everyone has a right to play any game they buy regardless of how good they are at said game (the exception to this being those who would cheat, whom I believe should have their controllers and consoles explode, but thats another story). In addition, I am referring to the objective ways of measuring good and bad, not the subjective ones. K/D, W/L, Objective caps etc, not the BR/M4A1/Ken/Zerg etc. tussles people have on forums and in-game. In your intend to turn this thread into one of those fights, don't even post. You're just wasting everyone's time.
 

AnOriginalConcept

New member
Jan 7, 2010
187
0
0
I don't understand your point. Of course you can be average at a game. Even if you don't have a 1-1 KD spread every game, you can on average.
If you chart the K/D of players in first person shooters, I'm certain it will be a bell curve.

Why can't someone be average? Why isn't the bell curve appropriate?
 

Agayek

Ravenous Gormandizer
Oct 23, 2008
5,175
0
0
It would make much more sense to just use a modified Elo rating system as has been used successfully for oodles of different games since 1960. It's not perfect, but it's as close as I've ever seen.
 

Chamale

New member
Sep 9, 2009
1,344
0
0
I don't see why player skill is best represented by a double curve. Player skill does, in my experience, tend towards an average, with certain exceptions on the high or low ends. I wouldn't say that most people are generally bad or good, just average. If there's a server of above or below average skill, these players will look like they belong in another place, of course.

I think Left 4 Dead has better matchmaking than Microsoft, secretly keeping track of dozens of statistics in order to match players with those who have similar playstyles. It even programs AI based on the player's skill level and personality to be more similar, and also to act as foils to the team.
 

Paragon Fury

The Loud Shadow
Jan 23, 2009
5,161
0
0
AnOriginalConcept said:
I don't understand your point. Of course you can be average at a game. Even if you don't have a 1-1 KD spread every game, you can on average.
If you chart the K/D of players in first person shooters, I'm certain it will be a bell curve.

Why can't someone be average? Why isn't the bell curve appropriate?
How you can have an average player? In most games, there isn't a middle ground; you either are A or B. There isn't a "C" in there.

You can win or lose. You can kill more than you die, or die more than you kill. You either are a help to your team most of the time, or you hinder them most of the time. Saying there is an "average" implies that there is a middle; which there objectively cannot be. You can have different levels of good and bad (One person can hinder your team more than another would for example), but at the end of the day, that person has a definable positive or negative effect on your team for the majority of his or her games, or on everyone in the case of FFA.
 

Chamale

New member
Sep 9, 2009
1,344
0
0
Paragon Fury said:
AnOriginalConcept said:
I don't understand your point. Of course you can be average at a game. Even if you don't have a 1-1 KD spread every game, you can on average.
If you chart the K/D of players in first person shooters, I'm certain it will be a bell curve.

Why can't someone be average? Why isn't the bell curve appropriate?
How you can have an average player? In most games, there isn't a middle ground; you either are A or B. There isn't a "C" in there.

You can win or lose. You can kill more than you die, or die more than you kill. You either are a help to your team most of the time, or you hinder them most of the time. Saying there is an "average" implies that there is a middle; which there objectively cannot be.
I disagree. Plenty of people can be close enough to the middle ground that they're considered average. Sure, it varies by game or by year. But so do stats in baseball. This year, Alfonso Soriano is an above average player, with .260 batting average and 19 home runs so far. Last year, Alfonso Soriano was pretty average, with a low .241 batting average but a respectable home run count of 20. However, his stats are always roughly the same each year. This shows that baseball players can be classified into a categorical skill level. Alex Rodriguez = Good (But very overpaid), Mario Mendoza = Bad (but even at age 60, he'd kick your ass in a baseball game).

Sure, on one day, an average player might do very well or very badly. Johnny van der Meer was definitely an average pitcher - in his career, he had a 119-121 record. However, he had an unexpected spike in skill when he pitched 2 consecutive no-hitters. This is a fine example of an average player who unexpectedly does very well, and is always remember as a much better player than he is.

[/rant on someone who doesn't understand statistics.]

EDIT: By the way, I found a source of some video game statistics. If we consider player skill to be measured by the amount of time it takes to finish Half-Life 2: Episode 1, we find that this time is indeed a bell curve [http://www.steampowered.com/status/ep1/]. Not a standard bell curve, but not the double bell curve predicted by the OP.
 

Jamash

Top Todger
Jun 25, 2008
3,638
0
0
Not every game has a fatalistic win or lose, kill/death outcome to competitive matches.

Sports games like Pro Evo Soccer can have matches that end in a draw (and league tables), and racing games end with the players finishing in 1st-8th place (or 1st-3rd podium places), not 1 winner and 7 losers.

Also, I'd consider myself to be an average player. I've been playing videogames for 27 years so I'm well versed in the concepts and can pick up any game really easily (it's very rare that I find any game challenging), however I don't have the competitive urge or desire to hone my skills through constant competition and become anything better than average.

Because I play a large variety of games for fun and because I'm not very competitive, I'm better than the person who has just picked up the game and isn't very good, but I'm not as good as the leaderboard player who plays that game constantly and exclusively to become the best at that game, therefore I must be somewhere in between the two extremes, or average.