New way to calculate standings
I think there is a better way to show the data that is being compiled than is being done now. Currently, when a given player places using multiple characters, the points for that placing are divided evenly among them. I propose that instead of dividing the points, no points should be awarded instead.
Consider the following three games:
Game 1
-A B C
A- 5 5
B- - 5
Game 2
-A B C
A- 9 1
B- - 9
Game 3
-A B C
A- 5 9
B- - 9
If you were to look at tournament data for the three games, you would expect to see all characters represented in Game 1, about equally. This is also the case in Game 2, as the three characters share a rock-paper-scissors relationship. In Game 3, however, you would only expect to see A and B placing.
Game 1 is clearly the most balanced game, but is Game 2 or 3 the second most balanced? Game 3 has more match ups that are even, despite effectively having fewer characters to choose from. In the double blind first game of a set, for Game 2 there’s only a 1/3 chance the players will select the same character and have an even match. Otherwise, the first game, and as a result the whole set, is decided at the character select screen. Even when they do pick the same character, the extreme lopsidedness of the match ups causes the set to be decided by who won the first game. In Game 3, as long as the two players stay away from C, every round of the set will be an even match up. It is for these reasons that I believe Game 3 is the second most balanced game, despite effectively having fewer characters.
So, looking at which characters place in tournaments for Game 1 and Game 2, both games will have all three characters placing about equally. So how do we tell the difference between a very well balanced game and a very poorly balanced one using tournament data? The difference is in the number of characters a player uses to place.
In Game 1, no one ever needs to counter pick. In fact, you can’t get a match up advantage by switching characters. Furthermore, learning two characters takes a lot more effort than just learning one character, as the number of match ups you need to know doubles. So by attempting to learn two characters, you’re likely just putting yourself at a disadvantage. As a result, the overwhelming majority of players who place in Game 1 will likely be using only one character.
Contrast this with Game 2. The match ups are so lopsided, people who place will have to learn all three characters to overcome the large percentage of unwinnable pairings. You would therefore expect the top placers in Game 2 to be using all three characters on their way to the top.
Now, most games are closer to Game 3. There is a set of characters with no unwinnable match ups, and a set of characters with unwinnable match ups. In practice, an unwinnable match up is one where it’s easier to just learn another character than it is to learn how to win the match up, and this tends to be at around the 7-3 mark. So now we can have a functioning definition of what a “viable” character is: a viable character does not have any unwinnable match ups. This means that a person playing a viable character doesn’t have to counter pick to win a tournament. It is possible for a player to use multiple unviable characters, with the resulting combination allowing him to win, but if one is used alone, the player will be unable to win the tournament (at least one that has a decent level of competition).
Therefore, by not awarding points for placings that used multiple characters, we can get a clearer picture of which characters are viable. Characters that can’t place alone won’t be getting points, revealing which characters really are viable. The resulting numbers will much more accurately reflect the balance of the game.
There are other benefits to making this change. Worst case, it requires the same amount of number crunching, but most of the time, lines can be stripped out of the placings due to the use of more than one character. Depending on how the data has been stored and processed, it might even be fairly easy to recalculate the previous standings. This would supply pretty powerful evidence one way or the other on whether Metaknight is truly ban worthy.
I guess all that’s left to do now is find out if I hit the character limit or not….