pearldrumbum
New member
- Joined
- Sep 13, 2013
- Messages
- 3
This isn't really for anything important, but every now and then my brain gets locked on something and won't let go.... you know how it is, I'm sure.
I started thinking about how video game rating systems work, and set out to devise my own simple version of one, just so I could test it and simulate some games.
So say it's a racing game, with 4 players. Video game rating systems are meant to attempt to measure the "true skill" of a player, so for my simple version I'm going to assume that it does accurately enough.
I want to simulate the finishing order of 4 players with the following ratings of "true skill." The rating numbers themselves are arbitrary for now, but I've tried to disperse them in a way that there should be an obvious trend in how they finish, leaving the last two closer together to attempt to see more "flip-flopping" of positions than the first two.
Player A - 1500
Player B - 1250
Player C - 1000
Player D - 900
To clarify, these numbers just represent relative skill. ie - Player A is ~18% better player than Player B (so they should be scalable). I want to be able to calculate the probability of the different outcomes so I can reward players more for doing better than they "should." For example, with the ratings above, Player C would get a bigger rating boost for winning the race than Player A would, because it is expected that A would win. Conversely, Player A would get a huge rating penalty if he came in last, relative to the penalty Player D would get for the same result.
I started thinking about how video game rating systems work, and set out to devise my own simple version of one, just so I could test it and simulate some games.
So say it's a racing game, with 4 players. Video game rating systems are meant to attempt to measure the "true skill" of a player, so for my simple version I'm going to assume that it does accurately enough.
I want to simulate the finishing order of 4 players with the following ratings of "true skill." The rating numbers themselves are arbitrary for now, but I've tried to disperse them in a way that there should be an obvious trend in how they finish, leaving the last two closer together to attempt to see more "flip-flopping" of positions than the first two.
Player A - 1500
Player B - 1250
Player C - 1000
Player D - 900
To clarify, these numbers just represent relative skill. ie - Player A is ~18% better player than Player B (so they should be scalable). I want to be able to calculate the probability of the different outcomes so I can reward players more for doing better than they "should." For example, with the ratings above, Player C would get a bigger rating boost for winning the race than Player A would, because it is expected that A would win. Conversely, Player A would get a huge rating penalty if he came in last, relative to the penalty Player D would get for the same result.