I would say considering how many addional free bonus kills are due to funky deaths, that funky suicides should be deducted to account for balance. Also some SK really are just your own fault.

Ok, I have been trying to include opponent rank into my equation somehow, but can't think of a way at the moment.
The reason for this is to
A) Avoid people only playing newbs/cannon fodder and avoiding the higher skilled players to almost guarantee themselves all the kills and wins to get the top ranks.

B) To make it twice as nice when beating the top guys, first the satisfaction of the win, then joy of seeing your rank step up.

C) If you play at a time when the top players play, you will not be punished twice for losing and if you play at a time when the less accomplished players play, you will not gain too much in rank.

The problem to avoid is having it make TOO much difference as it is at the moment.
I am well aware that the current system for skill fluctuates too much and needs to stabilize more, but I am hoping that with the changes being proposed this will happen anyway.

I hope all those that are active in this thread and putting effort into improving the system understand what I am trying to get at and can come up with ways of incorporating my idea...

Any idea as to when the new version will be coming out so that we have some idea of when a poll should be made as to which method is most popular with the community to allow an admin decision to be made and implimented in time for the next stat series?
(Which is my prefered method on how to decide what change should be implimented).

The stats could be easily modified to limit the number of games to be counted towards rank during the week while continuing to allow all games played on the weekends.

Regardless of what method is choosen, please exercise due caution and do the work required before changing the established and time proven ranking system.

The method I have proposed does not prevent anyone from playing all day, if they so desire. "Now that your picture's in the papers being replicly admired and you can have anyone that you have ever desired, all you got to tell me now is why, why, why, why? Welcome to the working week!" - Elvis Costello

It simply limits the number of games the server will count towards our rankings.

I think the stats we're proposing should be considered for nothing else but to rank players for use in the current calculation for stats.

In other words, simply rank players on the server according to their rating, then use that rank when accounting for skill between players. The way the current skill is calculated is not a bad method it but I wouldn't say that the current system is 'proven'. It simply uses the wrong data imho.

I've asked gavin what he thinks of this and am waiting for reply.

Ah cool, can ya let us know if he's interested in a change pls.
I've been concerned that this has already been discussed but no changes implimented, I'd me more inclined to put more effort in if a change is definately possible.

it is certainly possible, gavin has said in the past if you look around hard enough, that he'd like to change the server rank to a skill based, rather than quanity based system. not many ppl have ever really taken much stock on the skill rating the way it is so I think an improvement would be needed before that happened.

Hmm, thought I had replied. Sorry.

Sure that sounds fine, as long as the community is ok with it.

there were two things, one was to change the way the calculation considers rank, it apparently worked by a difference in skill scores, not difference in server rank as I had thought it did for a long time.

Anyway, I think that to change the rankings displayed on the server to a new method, that method would need to be well insulated against cheaters. Using any skill calculation for that rank would make it slightly easier. Using total kills as the ranking means that said cheater would need to play a lot and typically we find them and ban them if they play that much. I have some formula's that could work for this. I've made it as strait forward as I think I can. I need to go back and look more at Peanut's method too.

Consider this sheet: