A better way than Elo! (ASP)

[asdfghjklohhnhn]

ASP = Asdfghjklohhnhn's Scoring Policy

I think ASP should be Implemented into PS, but I have Insufficient Permissions Implement it by myself!
So I hope Antar will see this post, and recognize it!

n≈(y)*((((a*(x))*(-(y^2+2)+20*x+(a^((3/2)))))/(500*log(100*(x))))+(1500/(y+1)))-100(+-)|(o-r)(+-)(5, 10, or 15)|(+-)log(b!)*1.5  and  Minimum=1000  and  Start=1000

Where o = Opponents Rating at the Moment

Where r = Your Rating at the moment

Where n = Your new rating

Where a = x + y

Where b = starts at 1 changes every 5 wins by +1

Where x = Number of Battles Won

Where y = Number of Battles Lost

+- = + if won, - if lost

≈ = Rounded to the nearest 1's place

(5, 10, or 15) = 15 if not on the ladder, 10 if 500-201 on the ladder, 5 if 200+ on the ladder

Example:

If someone brand new wins 14 battles in a row and lost 1 afterwards,

 

[Zarel]

There's nothing wrong with trying to improve a rating system, but I should note that our current rating systems have lots of experience and research behind them.

Anyway, you might want to take a look at the GXE announcement thread:

https://www.smogon.com/forums/threa...layers-overall-rating-than-shoddys-cre.51169/

Notice how it has:

- A detailed explanation of why the formula is the way it is
- An analysis of the disadvantages of CRE and advantages of GXE
- Run simulations, and tables comparing each

 

[Nix_Hex]

Can't you just run your own server and implement this rating system, our is it client dependent? If the latter case, you can host a modified client.

 

[asdfghjklohhnhn]

Ok, thank you, if I ever get the chance to make a Server I will Implement this, and I will do more research, thank you!

 

[Antar]

asdfghjklohhnhn, explain to me what problem with Elo your system solves and how of solves it, and we'll talk. I highly suggest you read the "Everything you ever wanted to know about ratings" article linked in my signature. It explains not just the rating formulas but the theory behind them. Your rating system could very well be an improvement over Elo, but without explaining the reasoning behind your formula, I see no reason to expect that it would be.

Note that if I decide it's worth investigating, there's no need for you to start your own server as NixHex suggests--I can always calculate ratings offline from battle logs using arbitrary formulas. But my time right now is at an extremely steep premium, so I'm not investing any energy into something like this unless there's a really good reason.

 

[asdfghjklohhnhn]

Yes I am editing it now, just I was away!