BunnyPuncher
Joined: Aug 02, 2003
|
  Posted:
Aug 18, 2005 - 15:27 |
|
Laviak wrote: | we have an awful lot of data in the fumbbl database ... if someone clever enough were to put some time into it, I'm sure we could do some analysis of wins/losses of the various teams and skills, and make use of that in determining a 'more accurate' TS rating system. Might take a bit of work though
|
I've done some analysis on 64,000 games played on fumbbl during 2003.
The only significant predictors (in order of importance) of winning are:
Luck Meter Differential
Coach Ranking Differential
Cas Differential
Race, Matchup etc. all contribute nothing in both a log regression and a discriminant analysis. Obviously this grouping of races in one datapoint covers up the ineptitude of teams such as gobs and flings but overall there seems to be little difference between the races.
From my discussions with Christer many moons ago the database with the player skills included is insanely large to get enough games (a couple hundred) so I'm not sure that is a workable data set. |
_________________
|
|
nin
Joined: May 27, 2005
|
  Posted:
Aug 18, 2005 - 16:22 |
|
BunnyPuncher wrote: | I've done some analysis on 64,000 games played on fumbbl during 2003.
The only significant predictors (in order of importance) of winning are:
Luck Meter Differential
Coach Ranking Differential
Cas Differential |
Yes, those are exactly the significant ones, because if you are lucky, your opponent does funny things, like forgetting were he scores (difficult, but I keep trying) and you manage to make him kill all his players tripping while dodging and going for it (I nearly did that once) you are probably winning no matter who has the +37 TS (ranked, in unranked it can be even better) advantage. |
|
|
Dreamsplitter
Joined: Dec 01, 2004
|
  Posted:
Aug 18, 2005 - 16:38 |
|
um... the luckmeter really wouldnt show an opp forgetting to score |
|
|
sk8bcn
Joined: Apr 13, 2004
|
  Posted:
Aug 18, 2005 - 16:59 |
|
BunnyPuncher wrote: | Laviak wrote: | we have an awful lot of data in the fumbbl database ... if someone clever enough were to put some time into it, I'm sure we could do some analysis of wins/losses of the various teams and skills, and make use of that in determining a 'more accurate' TS rating system. Might take a bit of work though
|
I've done some analysis on 64,000 games played on fumbbl during 2003.
The only significant predictors (in order of importance) of winning are:
Luck Meter Differential
Coach Ranking Differential
Cas Differential
Race, Matchup etc. all contribute nothing in both a log regression and a discriminant analysis. Obviously this grouping of races in one datapoint covers up the ineptitude of teams such as gobs and flings but overall there seems to be little difference between the races.
From my discussions with Christer many moons ago the database with the player skills included is insanely large to get enough games (a couple hundred) so I'm not sure that is a workable data set. |
could I take a look on this. That luck meter has an influence on winning percentage is clear. CR diff also.
about race I am more sceptic. The team has IMO an influence but here you hide it that possiblely other influence such as str diff, TR diff and CR diff. You need to make some discriminant analysis with those factors considered as stable.
I would like to take a look on it after all. |
_________________ Join NL Raises from the Ashes |
|
BunnyPuncher
Joined: Aug 02, 2003
|
  Posted:
Aug 18, 2005 - 18:11 |
|
I forgot TS btw... it is between Coach Rating and Cas... needless to say TR means nothing...
sk8cn: I keep meaning to post this stuff on the blog I just need to spend an hour or so on it... I'll try for next week. |
_________________
|
|
Telcontar
Joined: Nov 12, 2004
|
  Posted:
Aug 18, 2005 - 18:34 |
|
Hmm if TR means nothing and TS is a major factor in winning and loosing. And people cherrypick in choosing teams with considerably higher TR but equal TS to get some handicaps, while playing an equal strong team then we should try to change handicaps in Ranked.
I dont know if this is a dumb idea but cant handicaps be coded over TS instead of TR?? |
_________________ Who is General Failure and why is he reading my hard disk? |
|
Uber
Joined: Mar 22, 2004
|
  Posted:
Aug 18, 2005 - 19:08 |
|
Telcontar wrote: | I dont know if this is a dumb idea but cant handicaps be coded over TS instead of TR?? |
Actually it would be very easy. The java client simply reads the TR on the computer generate rosters from Fumbbl. It would be a simple matter to switch the value being written to the file when they're generated.
However, it's been pointed out that the strength formula is not completely balanced. In my opinion, a few teams suffer especially from this. Lizardmen for example tend to be overvalued, while Vampires are a bit undervalued.
Another thing that is undervalued and under costed is High Armor. AV9 makes a big difference, especially for the teams that have a majority of players with it.
It would be interesting to playtest this kind of system in dX. Maybe go with the team value calculations from the vault, they're somewhat similar to TS calculation. Anyhow, there doesnt seem to be much work being done on improvement right now, but it's interesting to ponder... |
_________________ Recovering FUMBBL addict. |
|
sk8bcn
Joined: Apr 13, 2004
|
  Posted:
Aug 18, 2005 - 22:58 |
|
at Telcontar: well other problem is: Fumbbl follows LRB and in LRB there is no str formula.
I like the idea of following as much as possible official rules. Even if it would make sense to change handicap, I would be against. |
_________________ Join NL Raises from the Ashes |
|
sk8bcn
Joined: Apr 13, 2004
|
  Posted:
Aug 18, 2005 - 23:02 |
|
worf, forgot to answer Bunny: I know you are graduate in Uni but I don't know in what kind of domain. I am statistician and follow pure math cursus. Usually, I don't go into numbers (e.g. CR formula) because maths are for work, fumbbl for fun (basically). But know that I see your post, I am getting curious. I would like to see the maths and interpretation that followed not the public version. Actually I want maths... |
_________________ Join NL Raises from the Ashes |
|
jimimothybodles
Joined: Mar 31, 2004
|
  Posted:
Aug 18, 2005 - 23:11 |
|
im a maths graduate too...
i however NEVER want to see maths again! *curses it all*
|
_________________ "Give a man a match, and he'll be warm for a day.
Set a man on fire, and he'll be warm for the rest of his life."
~Ancient Snotling Proverb |
|
CircularLogic
Joined: Aug 22, 2003
|
  Posted:
Aug 18, 2005 - 23:18 |
|
Yeah... post the math. As a hobby math-freak and biologist I`d like to see the calcs, especially as I did an assay in school over statistics as an artform, interpreting the same numbers in oposite ways |
|
|
Pingo
Joined: Aug 11, 2004
|
  Posted:
Sep 06, 2005 - 16:42 |
|
i think the CR should be wiped every 6 months or so
still long enough to prove your worth, yet temporary to get some people to loosen up. |
|
|
MrMojo
Joined: Apr 17, 2004
|
  Posted:
Sep 06, 2005 - 16:46 |
|
No! I wanna see my CR rise mountain high! |
_________________ My post count
Jesus loves me this I know, 'cos my Bible tells me so. |
|
Mezir
Joined: Aug 02, 2003
|
  Posted:
Sep 06, 2005 - 16:48 |
|
The point of the ELO system is that it stabilises your ranking over time. As such a wipe every six months is counter-productive. |
_________________ Build a man a fire and he's warm for a day; set a man on fire and he's warm for the rest of his life. |
|
thechosenone
Joined: Jul 08, 2005
|
  Posted:
Sep 09, 2005 - 23:08 |
|
How very, very, very interesting,
When Christer asks you how ranked is, many of you have said it is A-OK, fine, no problems...
Yet when the next cherry picking thread appears, those ones will be the first to whinge about it with long tirades of text.
Is it cold feet? Or the fact one really can't justify whining about it at certain levels, e.g., Christer's level because they realise how petty it really is when viewed at that magnification?
e.g.,
Rich Westerner at resteraunt says angrily to his equally rich friend: "this Peking duck is a bit underdone, and the texture is lacking"
Homeless man who is walking past the outside tables: No good is it, sir?
Rich Westerner: (flushes red) No, actually it is very good, really.
(Please note, in no way am I implying that Christer is a homeless scavenger). |
_________________ - Wimp Lo: I'm bleeding, making me the victor
- Master Tang: Pay no attention to Wimp Lo, we purposely trained him wrong... as a joke. |
|
|
| |