37 coaches online • Server time: 16:00
Forum Chat
Log in
Recent Forum Topics goto Post Creating a custom to...goto Post Secret League Americ...goto Post DOTP Season 4
SearchSearch 
Post new topic   Reply to topic
View previous topic Log in to check your private messages View next topic
Christer



Joined: Aug 02, 2003

Post   Posted: Oct 20, 2017 - 21:37
FUMBBL Staff
Reply with quote Back to top

JackassRampant wrote:
Using current “all divisions” CR:
175+: 33 coaches
170+: 87
165+: 119
160+: 131
155+: 223
150+: 525
145+: 492
140+: 103
<140: 29


That's roughly 1750 coaches. These are currently put into brackets with the top 50 as legends, and the remaining 1700 split into 5 equal groups of roughly 340 coaches each.

That makes the CRs roughly as follows:

Experienced: 0-148
Veteran: 148-150
Emerging Star: 150-153
Star: up to 152-159
Super Star: 159-174
Legend: 174+

What I want to do is to redo this so that the emerging star bracket is squarely centered on 150 (150.03 is the actual average of ratings at the moment), and then use standard deviation (Roughly 5.4) as a basis for the brackets. Without much consideration:

Experienced: below -3 sigma (up to 133.5)
Veteran: -3 to -1.5 sigma (up to 141.75)
Emerging Star: -1.5 to +1.5 sigma (up to 158.25)
Star: +1.5 to +3 sigma (up to 166.5)
Super Star: +3 sigma to +4.5 sigma (up to 174.75)
Legend: 172.5+ (above 4.5 sigma, or top X coaches) (174.75+)

Would end up with loads of people in emerging star though, so possibly narrow that one slightly.. Would be nice to have a nice mathematical foundation to this.
JackassRampant



Joined: Feb 26, 2011

Post   Posted: Oct 20, 2017 - 21:39 Reply with quote Back to top

Emerging Star should be large. That's fine. Maybe tighten it on the low end?

-0.75 sigma?

_________________
Veni, Vidi, Risi
Grod



Joined: Sep 30, 2003

Post   Posted: Oct 20, 2017 - 21:51 Reply with quote Back to top

Christer wrote:

Would end up with loads of people in emerging star though, so possibly narrow that one slightly.. Would be nice to have a nice mathematical foundation to this.


Can you plot a histogram of the distribution of CRs? You might find some funny effects like a huge spike near 150 because of the number of coaches that never played or played only once. Would you exclude them when you calculate the std to work out the bands? Also if the distribution isn't normal, plotting it might give you some ideas where to stick the bands.

_________________
I am so clever that sometimes I don't understand a single word of what I am saying.

Oscar Wilde
Christer



Joined: Aug 02, 2003

Post   Posted: Oct 20, 2017 - 21:53
FUMBBL Staff
Reply with quote Back to top

Grod wrote:
Can you plot a histogram of the distribution of CRs


FUMBBL->Statistics->Rankings tab.

Shows a CR distribution for everyone with 10 or more R or B games played.
happygrue



Joined: Oct 15, 2010

Post   Posted: Oct 20, 2017 - 21:55 Reply with quote Back to top

What about adding "rookie" to make 7 brackets, then emerging star could be centered and you can work with an even number of brackets above and below?

_________________
Come join us in #metabox, the Discord channel for HLP, ARR, and E.L.F. in the box!
Image
JackassRampant



Joined: Feb 26, 2011

Post   Posted: Oct 20, 2017 - 22:06 Reply with quote Back to top

happygrue wrote:
What about adding "rookie" to make 7 brackets, then emerging star could be centered and you can work with an even number of brackets above and below?
If "Legend" and "Rookie" are both calculated differently from the others, he's already got that.

_________________
Veni, Vidi, Risi
happygrue



Joined: Oct 15, 2010

Post   Posted: Oct 20, 2017 - 22:13 Reply with quote Back to top

Yeah, upon reflection I think exactly as Christer wrote above looks pretty good.

One other random idea would be to make super star exactly symmetrical to experienced... Unless you have won a major and are over 4.5 sigma - then you get legend. That would make legends really rare, but might also drive up tournament participation for coaches who really want that tag... Wink

_________________
Come join us in #metabox, the Discord channel for HLP, ARR, and E.L.F. in the box!
Image
koadah



Joined: Mar 30, 2005

Post   Posted: Oct 20, 2017 - 22:17 Reply with quote Back to top

Christer wrote:
Grod wrote:
Can you plot a histogram of the distribution of CRs


FUMBBL->Statistics->Rankings tab.

Shows a CR distribution for everyone with 10 or more R or B games played.


Has that filtered out people who played 10-15 games and were never seen again?

_________________
Image
O[L]C 2016 Swiss! - April ---- All Stars - Anniversary Bowl - Teams of Stars - 13th March
thoralf



Joined: Mar 06, 2008

Post   Posted: Oct 20, 2017 - 23:52 Reply with quote Back to top

happygrue wrote:

One other random idea would be to make super star exactly symmetrical to experienced... Unless you have won a major and are over 4.5 sigma - then you get legend. That would make legends really rare, but might also drive up tournament participation for coaches who really want that tag... ;)


That or beating a Legend in a best of 12 match.

Another idea would be to give titles more "inertia" - for example, a Star could keep his title for a while. This could nudge him to experiment, or perhaps play in an ARRRR!

After that while, either his rating swings back or he's demoted.

_________________
There is always Sneaky Git.
Rags



Joined: Nov 09, 2008

Post   Posted: Oct 21, 2017 - 01:12 Reply with quote Back to top

PurpleChest wrote:
I have always (perceptions aside)defended CR as a reasonable attempt at defining coach rating. Go check, I stand by everything i have ever posted.

I struggle to keep up with the pure math, but can usually follow the reasoning and hence eventually get the math too.

But I am lost as to why a 4-0 win is better than a 1-0 win.
It would seem contrary to your stated aims.
Increasing, as it does, the value of unfair games with high score outcomes.
Promoting certain styles of play and certain races while devaluing others,
Bringing 'perception' of a 'good' win into something seeking a rational outcome.

So I am interested as to why you feel this is an appropriate measure? It feels like a step toward 'personal preference of play style' and away from 'how likely is a win' to me.


Maybe because it's more difficult and risky to win by margin of 4? Going for margin 1 is a definite style of play so heavily favoured it's the default
Rags



Joined: Nov 09, 2008

Post   Posted: Oct 21, 2017 - 01:17 Reply with quote Back to top

MattDakka wrote:
I agree, there is not necessarily direct and strict correlation between winning with high TD difference and coach's skill (of course, there could be exceptions).


There's not necessarily direct and strict correlation between much, but if a coach consistently wins by higher margins against comparable opposition, then that would seem to indicate greater skill
MattDakka



Joined: Oct 09, 2007

Post   Posted: Oct 21, 2017 - 02:02 Reply with quote Back to top

Probably a coach playing vs opponent coaches of comparable skill with average dice would not try/risk/be able to win by large margins.
Balle2000



Joined: Sep 25, 2008

Post   Posted: Oct 21, 2017 - 04:01 Reply with quote Back to top

A problem with the previous CR formula: people are discouraged to even touch sub par races (and team builds) because of the immediate negative impact it is likely to have on their CR.

Without having taken the time to read everything, hopefully this has been addressed this time around, and we might have more variation.
Balle2000



Joined: Sep 25, 2008

Post   Posted: Oct 21, 2017 - 04:03 Reply with quote Back to top

I like the idea of major win unlocking the Legend tag.
koadah



Joined: Mar 30, 2005

Post   Posted: Oct 21, 2017 - 16:09 Reply with quote Back to top

Balle2000 wrote:
I like the idea of major win unlocking the Legend tag.


Maybe they could be Legend* to differentiate them

I think that Major winners should have something special on their home page. Maybe a trophy or rosette for each Major won.

_________________
Image
O[L]C 2016 Swiss! - April ---- All Stars - Anniversary Bowl - Teams of Stars - 13th March
Display posts from previous:     
 Jump to:   
All times are GMT + 1 Hour
Post new topic   Reply to topic
View previous topic Log in to check your private messages View next topic