SeraphimLeftNut wrote:

Col_Walter_Kurtz wrote:Skill does have the tendency to level off the higher you get, though. So even though skill is not a constant you have to assume that large jumps are not warranted, the more games are on record. But I guess this is obvious already. Are there better systems out there, that correct for large jumps but allow for a more fluid skill development?

I completely disagree with your opinion.
Generally skill is always increasing unless the player has other real life issues to skew this.

It would be quite remarkable if real life issues exactly balance the natural skill improvement in time to yield a constant skill.

Deafening Milk wrote:
For those of us that aren't buffs on this kind of terminology or looking at formulas could someone please explain how trueskill works?

Am I right in saying (from skimming through what has been said) that we have our rating (I'll use myself as an example) of 900 and let's say I have a deviation of 50 maximum so if I played a 1500 rating and I won I'd gain 50 points, if I played a 1200 and won I'd gain let's say 25 and if I played a 900 and won I'd gain let's say 5 (the larger the skill gap the greater the reward or loss if it were the other way around if you follow what I mean) am I right so far?

Then if I win vs that 1500 or 1200 because it was a result the system didn't expect due to me being the lower skill the deviation number would increase from 50 to let's say 60 as a result and if I was to lose/win where expected (lose vs 1200 or win vs 600) the deviation would drop let's say from 50 to 45 as I am doing what the system expects so it believes I am already closer to my "true skill level" again is this section right?

And would the deviation grow by a larger amount if I had an absolute shocker win let's say vs an 1800 than if I won vs a 1000?

I would personally (if I got the grasp of it right) like to be able to see the deviation amount I have somewhere easy to find and also something to tell me "if you win this match you will go up x amount of rating points and deviation will go up/down by y amount and the same but an if you lose.

tatsu wrote:

Voodoo wrote:The Elo-system is the best one in my eyes. It's easy to understand and you know before the game starts how much points you will lose\win.

disagreed.

ELO has no means byu which to value individuals in working together as a team and that not ELO's only weakness

ELO < TRUESKILL < *something we should make*

Voodoo wrote:
The Elo-system is the best one in my eyes. It's easy to understand and you know before the game starts how much points you will lose\win.

Sheeo wrote:

Morax wrote:I think I said this before but when this all goes down can there be an "all-time" leader board where ratings never decay and then an "active board" where it results at a fixed period of time that ties into the ladder match-making system?

Why would you want this? I'm not sure the implementation cost outweighs the benefit.

Sheeo wrote:
simply filtering out inactives on the leaderboards (We still do this, no need to introduce)

Morax wrote:
I think I said this before but when this all goes down can there be an "all-time" leader board where ratings never decay and then an "active board" where it results at a fixed period of time that ties into the ladder match-making system?

SeraphimLeftNut wrote:

Sheeo wrote:...This could be a simple factor of the amount of days since last ranked game.

Now that sounds like a good idea. The less you play the less certain your rating. Not playing means taking a larger risk than playing. Exactly what is needed.