What does all of this mean?
There are a lot of stats flying around the site. Some you may be familiar with, like the kill/death spread or total headshots. Others may require an explanation, which we list below.
Elo rating (μ)
A player/team's Elo rating, also known as the Elo μ value, is a measure of player/team skill in comparison to the other teams they've competed against. It is a function of wins/losses, not individual kills/assists/deaths. Elo-based rating systems are well-known in chess, where Elo is the official player rating system for FIDE (the World Chess Federation) and USCF (United States Chess Federation), as well as countless 1v1 video games. Check out the Wikipedia article for more information.
In our implementation of Elo, we used the formula outlined in the TrueSkill paper, which is why our Elo ratings aren't numerically similar to FIDE chess ratings. Also, we used a Gaussian distribution instead of a logistic distribution since we do not know enough about Halo 3 high-level player skill to know if a logistic distribution would be more appropriate.
Glicko rating (μ, σ)
Similar to Elo, Glicko is a player/team skill rating system based on wins/losses, but one of its key advantages compared to Elo is the notion of rating volatility (denoted by σ). Rating volatility can be thought of as a measure of rating certainty - the higher this value is, the less certain we are of a player/team's actual skill value. Again, the Glicko skill rating itself is the μ value, but in conjunction with σ, we not only have a sense of how good a player/team is, but also how certain we are of how good they are. If that makes sense. Glicko also allows you to discount the weight of older games in the rating, but we have not implemented that yet. It's on the to do list, though. Find out more about Glicko at its Wikipedia article.
Adjusted Glicko rating
This is a player/team's Glicko rating adjusted for the amount of rating volatility they have, which is calculated with the following formula: adjSkill = μ - 3*sqrt(σ). Essentially, we are adjusting the Glicko skill value downward by 3 standard deviations of rating volatility, which corresponds to the 1% lower quantile and, by the central limit theorem, means that about 99.7% of a player/team's skill is represented in this adjusted skill rating. In English, the adjusted Glicko skill rating is the "safe" estimate of a player/team's skill. This is the one you'll wanting to be caring about the most.
KAmD/M and KAmD/G
The purpose of this statistic is to differentiate between strong slayer/support players (i.e., those with high KA/M or KA/G) who stay alive and those who die often. The higher this number is, the greater the average margin is between the total number of kills/assists and the number of deaths. Only calculable for games where game length is known.
Slayer average
Slayer average quantifies a player's average slaying/support contribution (kills and assists) in slayer games relative to the maximum score in an MLG slayer game (50). Think of it as a "batting average" for MLG Slayer games — a player having a slayer average of .400 means that in slayer games, that player accounts for 20 kills/assists.
At the team level, it's useful to think of this as a proxy measure for "teamshot" in Slayer games. A team in which every player tends to have a lot of kills and/or assists will have a high slayer average. Also, because the ideal winning score in a slayer game is included in the denominator, teams which don't often reach the 50 kill goal in a slayer game are penalized compared to those who do (that is, their average isn't proportional to the team's score, it's proportional to a "perfect" score).
CTF Foundry average
Similar to slayer average, CTF Foundry average measures a player's flag capture contribution relative to perfect scores in Foundry CTF games (5 caps).
CTF non-Foundry average
CTF Other average measures a player's flag capture contribution relative to perfect scores in non-Foundry CTF games (3 caps).
CTF average
CTF average is a weighted average of CTF Foundry average and CTF non-Foundry average.
Oddball average
Oddball average measures a player's oddball time contribution relative to perfect scores in Oddball games (250 seconds).
King of the Hill average
King of the Hill average measures a player's hill time contribution relative to perfect scores in King of the Hill games (250 seconds).
Objective average
Objective average measures a player's objective contribution relative to perfect scores in each objective gametype (see above), weighted by gametype prevalence (e.g., if a team plays 25 Oddball games and 100 King of the Hill games, more weight is given to the King of the Hill average in the overal calculation).
Offensive average
This is the total weighted average of all 5 gametype averages (slayer average, foundry CTF average, non-foundry CTF average, oddball average, and king of the hill average). Players who score high in this statistic are excellent at adapting their in-game performance to the task at hand: kills/assists in slayer games, flag caps in CTF games, ball time in oddball, and hill time in king of the hill. Put differently, these players "play the game as it needs to be played". For teams, a high offensive average means the team is basically good (or great) at every gametype.
