Registration Date: 04.10.2009
Posts: 1,590
Location: Ljubljana, Slovenia
A new hall of fame
Now that I'm all in scripting I could do some scripts that get the data from the webpage and crunch it. The first idea is to do something similar to what Radiance was doing back in the fall - a hall of fame based only on latest galaxies. Ok, all nice and dandy, but there is some stuff to be defined:
- which galaxies count (ranked, populated by enough people)?
- am I to sum up normalized fame or total fame, score or is there some fine formula in existance?
- should i throw away team galaxies?
- some other criteria
so please help me define the problem, so I can waste time or else I'll install some stupid game and waste time.
ps. if you have some other idea for a script based on the data from the CS web just spit it out.
It's a scheme to render a relative skill level from 0 to 1...
In a nutshell it came down to this:
1) New players start at some arbitrary assumed skill level. Say .5 or .25 whatever really since its not at all accurate with 0 games played.
2) When a game starts one calculates the zValue of each player's skill... call this zSkill [zS for shorthand]. This is how will the player is expected to do in the game.
[wiki can explain how zValue = (V - uV)/sV normalizes scores http://en.wikipedia.org/wiki/Standard_score] Note that this can be problematic when a game has noithing but first time players in it since the stdev = 0 and so the zValue = divByZeroError... so in that exceptional case just use zSkill = 0
Come to think of it I should probably incorporate a standard error into the skill values and permit that to influence the stddev... particularly in games with a mix of first time and repeat players
3) When a game ends one calculates the zValue of each player's score... call this zPerformance [zP for short... actually I used to call it zScore but I am right now thinking zPerformance is better so that the shorthand zP makes more sense and for no other reason... its just a name]
4) Then determine how well they did vs expectations and assign a "Performance" value for that particular game. This was the meat of it. Think of this is establishing what skill level the apparently played this game at.
5) Adjust each player's skill accordingly which was just average in with all previous games.
Notes on step 4:
I think this is where it became seen as complicated.
Determining how well someone performs at such a complex task is a non-trivial exercise. The eqn's to describe such a value would be just as non-trivial.
I used trigonometry TAN's and arcTAN's to produce nice curves. Hmmm... I suppose if I had a reasonable stddev even with all new players I could just reverse the zS calc and use P = zPsP+uP... but then I would have to find the spreadsheet with the raw data again... ummm...
It was pointed out that this system has the possibility of a Pyrrhic victory. This occurs when an apparent ace plays against a number of less skilled players and while (s)he might well win the game, (s)he fails to do so with the crushing margin they should have done so with.
Many IIRC did not like this feature. I was not one of those.
After the Skill was derived there was a secondary step that had to do with assigning a "rank". I was a fan of the notion of a complete K-ary Tree [see http://en.wikipedia.org/wiki/K-ary_tree] where there was exactly 1 Top Ranked player with exactly K subordinates.
It was pointed out that this system has the possibility of loosing ranks just because someone does better than you. Many IIRC did not like this feature. I was not one of those.
And I am sure I should get back to work now
__________________ "I do not agree with a word you say, but I will fight to the death for your right to say it". --Voltaire
Registration Date: 22.02.2007
Posts: 708
Location: Third Star To The Left, straight on til morning
quote:
Originally posted by Nongolf
hehe. Excellent. You've been here two days and already we're argueing.
It wasn't actually about a new fame or reputation system, but about how to correlate the score charts into something useful.
Ah... I may well have been led astray by the title of the thread... "A new hall of fame".
I may also have misinterpreted:
quote:
Originally posted by mizzihood
The first idea is to do something similar to what Radiance was doing back in the fall - a hall of fame based only on latest galaxies.
My best guess was that "what Radiance was doing back in the fall" also had something to do with this: http://cosmicsupremacy.com/forum/thread.php?threadid=6621
[which in itself is a good idea if one sticks with the absolute ranking system that fame currently is].
I do hope you can forgive me for thinking that the OP was therefore speaking about "A new hall of fame".
Perhaps it comes down to defining terms. I was taking "the hall of fame" to be a system about how to correlate the score charts into something useful. Granted I have been away for some time... but it is still based primarily on the resulting scores from each galaxy is it not?
__________________ "I do not agree with a word you say, but I will fight to the death for your right to say it". --Voltaire
This post has been edited 1 time(s), it was last edited by Gedrin: 11.04.2012 15:44.
Originally posted by Gedrin
Granted I have been away for some time... but it is still based primarily on the resulting scores from each galaxy is it not?
Yup. Nothing changed with the rep/fame/score systems.
