SE4 Rating System

[geoschmo]

Originally posted by Slynky:
[PS: To Geo, I received both game results in the same message, so I did calculations based on the one listed first as the first one.]

Anyway you want to do it is fine with me. As long as I am not in Last place anymore. Woo Hoo!

[Slynky]

Originally posted by Phoenix-D:
You can consider the KOTH game verified.

Tks, Phoenix. (asterisk removed)

[Slynky]

Welcome to the SE4 Ratings system, Parabolize!

[Slynky]

OK, multiplayer game Ratings (what about this?):

Say A B C and D are in a game. Say they lose in this order: D C B and A.

D loses to everyone. Calculation: I average A B and C's rating and run it through the formula against D's rating. D's rating is updated with the number from the formula. A B and C's gain in rating is divided by 3 (the number of players that got the win) and posted to their stats immediately.

Then C is killed. I average A and B's rating and run it through the formula against C's rating. C's rating is updated with the number from the formula. A and B's gain in rating is divided by 2 and posted to their stats immediately.

Finally, B is killed. Here, I just run the regular formula.

Without going through all the calculations for 4 people rated at 1,000 exactly (figuring in my head where 1,000 vs 1,000 results in +16 and -16 points), this would be the results after the game:

A = +29 points
B = -3 points
C = -11 points
D = -16 points

The game count goes up 1 for every player. And, unless Lord Chane can modify the program, I'll be doing this on a spreadsheet.

So, what's good about this?

(1) Only slight modification to the chess formula and hopefully one that won't skew results.
(2) I can update results as people are eliminated from multiplayer games (someone mentioned that as being desireable).
(3) Losing won't have a "killer" impact on one's rating. In fact, you have to be in Last place to lose the same amount of points as losing in a one-on-one game.

What's bad about it?

(1) From the looks of it, only the winner gets points. Not a terrible problem, after all, he was the winner (but surely he had some help along the way). I'd like to see 2nd place get some points. Perhaps, 2nd place will get points if there are 5 or more rated players (I haven't checked it out but it looks like it).

I'll leave the game on the website until it has been completed.

Suggestions/comments?

[primitive]

Why not do it the easy way. Do as originaly planned, but add a "reduction-factor" to lessen the effect of the many win/losses.

With 40 % reduction for a 4 man game (and all keeping their 1000 points throughout the game):
A = +29 points
B = +10 points
C = -10 points
D = -29 points

[Slynky]

Originally posted by primitive:
Why not do it the easy way. Do as originaly planned, but add a "reduction-factor" to lessen the effect of the many win/losses.

With 40 % reduction for a 4 man game (and all keeping their 1000 points throughout the game):
A = +29 points
B = +10 points
C = -10 points
D = -29 points

And a 5-person game @ a reduction factor of 50% (5 x 10%) would result in:

A = +32
B = +16
C = 0
D = -16
E = -32

Right? (if I understand you).

[primitive]

That was the idea (if using 40 and 50 % as factors). But in hindsight, I don't think I like it very much

Coming Last or second to Last shouldn't really be much different scorewise, so your formula is better there. There should however always be a positive score for coming 2nd, so a little tweak may be in order.

The "perfect" formula for a 4 player game should give something like:
A = +30 points
B = + 5 points
C = -15 points
D = -20 points

And for 5 players:
A = +35 points
B = +15 points
C = - 5 points
D = -20 points
E = -25 points

Now thats a challenge for the math geniuses