howtothings.co.uk

For those of you that have seen The Social Network you will be familiar with the "Window" Algorithm, also known as the ELU algorithm. It was used by Chess companies, and still is by some, to rate chess players. It is very simple once you understand it. Basically it does this:

1. Calculates chance either candidate has of winning.
   
2. Winner's base rating goes up, losers goes down.
   

What's interesting is the method used to bring the base rating up or down. It is proportional. If somebody with a high probability beats somebody with a low probability (Going with the odds) The values only change a little. If it is the other way round, the values take a huge leap. This will be better demonstrated later.

The first thing you need to do is calculate the probability of each winning.



Both will give the percentage of a and b winning, respectively. Now, we need two things to compare, and we need to give them a base rating (Ra and Rb). So I've decided we should bring out a long fight for this one. Pack3t vs Sharn. Can I hear the crowd roar?


Given these base ratings lets calculate the chance of each winning



So we know who we'd pick to win in a competition based on those base ratings, but what happens if for some reason, Sharn Won!?!? The Base ratings would have to change. Look at what they would become.. (Note: T+1 means Time+1, it essentially means, what Ra/b will become)


But what if, as expected, Pack3t Won?

As you can see, the values don't move anywhere near as much as they did when Sharn won, this means that the equation accounts for the fact that somebody high up, beating somebody low down shouldn't give them as many points. I hope this guide has been useful, if you have any more questions about this equation. Ask, and I will answer to the best of my knowledge.

Alright, so what is the minimum and maximum score that is available? Sounds like a good program to write when one has a week off school

What are the odds of the people on this forum winning in a civil war, againsy each other. Being completely biased towards me, of course

Packet: There is no Minimum or maximum score, You give each contestant a base rating. If the same person keeps scoring, their score will keep climbing. Unless for whatever reason you choose to stop it using a program. I will definitely put some time into writing a program to do it. Possibly create a function for calculating base rating. Then another that will change Ra, using a pointer.

(22-02-2011, 01:29 AM)Drumm Wrote: [ -> ]Packet: There is no Minimum or maximum score, You give each contestant a base rating. If the same person keeps scoring, their score will keep climbing. Unless for whatever reason you choose to stop it using a program. I will definitely put some time into writing a program to do it. Possibly create a function for calculating base rating. Then another that will change Ra, using a pointer.

Right path I think. I would use one program with several functions. Having the main call upon the other ones. That way you can keep your main objective clear, while possibly creating some reusable code in called functions.

You can be a cunt at times...

Love you more Sharn!
http://mcompute.co.uk/showthread.php?tid=803

C functions for this algorithm above. The elupercentage function needs more brackets..

(22-02-2011, 12:44 PM)bigsharn Wrote: [ -> ]You can be a cunt at times...

I think giving you a rating which would signify you winning 24% of the time is rather generous.

(22-02-2011, 08:54 PM)Pack3t SynAck3r Wrote: [ -> ]I think giving you a rating which would signify you winning 24% of the time is rather generous.

How do you know who he was directing that at?

(23-02-2011, 01:55 AM)latch Wrote: [ -> ]

(22-02-2011, 08:54 PM)Pack3t SynAck3r Wrote: [ -> ]I think giving you a rating which would signify you winning 24% of the time is rather generous.

How do you know who he was directing that at?

(22-02-2011, 05:02 PM)Drumm Wrote: [ -> ]Love you more Sharn!

I used my powers of observation. Also, Meatloaf has never gone so far as to call me a cunt in any of our dialogues. It then seems unlikely a simulated one would finally drive him to it. More alpha, less beta.