Maelstrom wrote:
Quitch wrote:
It doesn't reflect a "living" opponent in the same way that I don't reflect Tiger, and that Tiger doesn't reflect madmax, and that madmax doesn't reflect Pitor, etc.
This is a very interesting view, I just don't get it
![:)](./images/smilies/icon_smile.gif)
. Since when do AIs get ranked in the same category as players? While it'd be interesting to see what glicko score the AI players will get, that would be a novelty to me, but not one that makes any sense to me to put in practice.
Every major ranking engine in Counter-Strike Source defaults to ranking bots.
Maelstrom wrote:
My point is that there is a learning curve in this game, and within a few games you have a basic understanding of how it works. Why should your inital learning curve fit in your score?
If that were true then every player would have peaked after three or four games, something which blatently isn't true.
You talk as if people are suffering from losing points early on, but I think this shows you don't understand Glicko. It is designed to show how good a player you are as accurately as possible. To do this it needs as much data as possible. If you withhold your early games from it all you're doing is decreasing the number of points somone earns when they win. The amount they gain will depend as well on who they beat.
Maelstrom wrote:
I understand that it will work out over time, but MA isn't exactly a game where you can get a lot of games in to fix your score in a short period of time. This totally gives the advantage to people that have played MA before and have gotten over the initial learning curve.
And that makes including the learning games all the more important. Veterans already have historical Glicko scores as the system has been running unofficially for ages. That system can (and does) track every game ever played on MA, so yes, Tiger/Pitor etc first games are all included.
I suspect the new Glicko system might well do the same thing.
Again, losing some points in your early games is not a handicap, they are essential data. I think you're massively overstating the cost of a couple of losses on Glicko. No one at the starting figure is going to lose enough points to mean much, especially if they play someone who is likely to beat them as they'll have a much higher rating.
The AIs will probably have a higher rating than your average starting player too.
Maelstrom wrote:
That is the main failing of Glicko with MA... it really requires a lot of games before it can stabalize itself reasonably.
Perhaps we'll have to agree to disagree on this point.
Which is why you're supposed to include the "score could be off by up to this much" factor.
Glicko isn't perfect for MAN2, but I haven't thought of a better system yet.
I lost my opening games on MA, but it didn't stop me hitting the top 10.