ELO Discussion

[Raveen]

I think Grimm was suggesting doubling the number of ranks but leaving the underlying ELO intact. Each rank would be worth 50 ELO rather than the current 100.

Bax, surely a k factor would unbalance the system and push it further away from being a 0 sum system (which would be mathematically ideal)? I was thinking that a multiplier on ELO gains losses for the first few weeks of the new system might help, sure people would fluctuate wildly but it would push the system away from everyone being Vet1 more quickly. Maybe start with a few weeks at 2xELO then cut it back to 1.5x and then settle in to the final standard values.

[Grimmwolf_GB]

QUOTE (Raveen @ Nov 23 2006, 10:59 AM) I think Grimm was suggesting doubling the number of ranks but leaving the underlying ELO intact. Each rank would be worth 50 ELO rather than the current 100.

Yes, exactly.
Double the number of ranks and have a more accurate display of skills on a team.

[jgbaxter]

QUOTE (Raveen @ Nov 23 2006, 08:59 AM) Bax, surely a k factor would unbalance the system and push it further away from being a 0 sum system (which would be mathematically ideal)?

Actually it would still be a 0 sum system, the k-factor can be (and in my preference should be) a two edged mini-3. /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" />

Of course, this depends on a couple of things. Most importantly, if we think alleg elo now is a 0 sum system then that's great and we can keep in mind that the elo downgrads proportionally for a player so they may be an expert category player with a k-factor of 8 if they win a balanced game they gain the 4 pts, conversely the opfor loses an aggregate of 4 pts.

I'm actually happier if we use the k-factor in terms of games played rather then skill, so that newer players and less skilled players can ocillate faster to their true skill rank, and those with many games ocillate much slower as they're already approaching or at their true skill.

Some numbers for example...

k-32 New players... <150 games
k-24 Short timers... <450 games
k-16 uppity addicts... <1350 games
k-8 old farts >4050 games


4am, ouch, ok I hope that isn't gigo. /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" />

[Gappy]

QUOTE (jgbaxter @ Nov 23 2006, 03:06 AM) k-32 New players... <150 games
k-24 Short timers... <450 games
k-16 uppity addicts... <1350 games
k-8 old farts >4050 games
4am, ouch, ok I hope that isn't gigo. /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" />

Sorry, I just have to throw this in... /wink.gif" style="vertical-align:middle" emoid=";)" border="0" alt="wink.gif" />

Given that the most played games in the last 2.5 months is about 350 games, that averages out to about 4.5 games that count per day that you'll probably play.

Thus, to reach K-8 status even if you played this game like it was crack would take about 3 years /tongue.gif" style="vertical-align:middle" emoid=":P" border="0" alt="tongue.gif" />

[Raveen]

QUOTE (Gappy @ Nov 23 2006, 11:44 AM) Sorry, I just have to throw this in... /wink.gif" style="vertical-align:middle" emoid=";)" border="0" alt="wink.gif" />

Given that the most played games in the last 2.5 months is about 350 games, that averages out to about 4.5 games that count per day that you'll probably play.

Thus, to reach K-8 status even if you played this game like it was crack would take about 3 years /tongue.gif" style="vertical-align:middle" emoid=":P" border="0" alt="tongue.gif" />

Given that there are plenty of people who have been playing this game for 6 and a half years is that so unreasonable? Basically Bax's suggestion is that once you have that many games on record then your rating should not vary much as ELO will have found your correct rating already.

[Pook]

Uh... you mean like this?

Code: Select all

UPDATE ELO
             SET Member_ID = p.Member_ID,
                 ELORanking = CASE
                              WHEN ELORanking < 800 THEN ELORanking + (p.Adjustment * (p.AdjustedModifier * 4))
                              WHEN ELORanking BETWEEN 800 AND 1500 THEN ELORanking + (p.Adjustment * (p.AdjustedModifier * 2))
                              WHEN ELORanking BETWEEN 1500 AND 2200 THEN ELORanking + (p.Adjustment * (p.AdjustedModifier * 1))
                              WHEN ELORanking > 2200 THEN ELORanking + (p.Adjustment * (p.AdjustedModifier * .5))

It's already in there /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" />

[Raveen]

QUOTE (Pook @ Nov 23 2006, 02:31 PM) Uh... you mean like this?

Code: Select all

UPDATE ELO
             SET Member_ID = p.Member_ID,
                 ELORanking = CASE
                              WHEN ELORanking < 800 THEN ELORanking + (p.Adjustment * (p.AdjustedModifier * 4))
                              WHEN ELORanking BETWEEN 800 AND 1500 THEN ELORanking + (p.Adjustment * (p.AdjustedModifier * 2))
                              WHEN ELORanking BETWEEN 1500 AND 2200 THEN ELORanking + (p.Adjustment * (p.AdjustedModifier * 1))
                              WHEN ELORanking > 2200 THEN ELORanking + (p.Adjustment * (p.AdjustedModifier * .5))

It's already in there /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" />

Don't you think that basing it on games played (rather than current rank) would be better? What's the reasoning for higher rated players having less volatile ratings?

[jgbaxter]

Oh damn, Pook kinda listened to me months ago and I missed it?!

hehe

Well, using current rank or games played both have their advantages, I like the latter myself. /smile.gif" style="vertical-align:middle" emoid=":)" border="0" alt="smile.gif" />