Woodstock
Joined: Dec 11, 2004
|
  Posted:
Nov 25, 2013 - 13:21 |
|
This suggestion doesn't make sense, fluff-wise or in the logical world... |
|
|
Smeat
Joined: Nov 19, 2006
|
  Posted:
Nov 25, 2013 - 16:42 |
|
There is some sense there, and depends if you're talking "overall" or race-by-race. And (I assume) the suggestion is not that 6 games makes a coach "skilled", but they cannot become "skilled" until they have played at least 6, regardless of their win %.
It would stop a short lucky streak (or quick early cherry-picking) to achieve a title. For those coaches who go after empty titles.
Having to play 176+ games sounds like a lot, and not every top-caliber coach needs or wants to go there (or even has the time to if they did!), but it would give recognition to those who do, and carry added significance to those who saw the title.
This would be more painful for individual races than all combined. Few of the top coaches on this site got that good by playing fewer games than that. Even if they played a lot elsewhere, this format takes some getting used to. |
_________________ Let's go A.P.E.!
(...and what exactly do you think they do with all those dead players?...) |
|
Chainsaw
Joined: Aug 31, 2005
|
  Posted:
Nov 25, 2013 - 17:12 |
|
I think Coach Rank needs a fundamental rethink. It should be renamed Chainsaw Rank, I should always be #1, and everybody else gets ranked accordingly. "He's #2 to Chainsaw, he must be ace!" Sounds like a system that I can get behind. |
_________________ Coach Chainsaw's Dugout
Free Gamer - blog - community |
|
Christer
Joined: Aug 02, 2003
|
I appreciate the feedback, but the thing is that when I pass by the code for CR again (which is on my todo list), I want to go the other way around.
The way the system works now, with 5 games minimum to pass beyond Rookie, it's impossible to tell if the person is someone who's played very few games, or simply someone who doesn't win alot for whatever reason (be it skill or team choice, it doesn't matter).
Currently, the CR system is ELO based and functions sort of as follows:
- A win probability (P) is calculated for each coach (Pa and Pb, where Pa+Pb = 1)
- The match is played and each coach gets a score (S), where S=1 for a win, 0.5 for a tie and 0 for a loss
- A new ranking is calculated: Rnew = Rold + K * (S - P)
The K value is generally speaking fixed (K=2), but is increased in case there's an "upset victory". This is defined by the event where a coach in a lower ranking bracket wins against a coach in a higher bracket (where bracket is one of the Rookie, Veteran, etc levels).
This bracket special handling is why I want to rework the system slightly, because coaches who are "fixed" to the lower brackets due to few games will be handled differently from others, which isn't ideal. This behaviour isn't terrible in itself but an early win can push people too high in an artificial manner.
What I would like to do instead is to switch to using a Glicko based system (which is similar to ELO on its own), where there's an additional factor involved called the "ratings reliability" or "ratings deviation" for each player. In simple terms, this allows a new coach to quicker find their "actual" level by (in ELO terms) use a higher K value early on. The system is also designed to slowly increase the value over time, so coaches who return after a lengthy time will end up with a more volatile ranking.
In terms of the bracket names, I would completely reserve the Rookie level to players with few number of games, and have Experienced -> Legend be the actual rankings showed once the number of games are high enough. The system would, however, consider the actual level of the player rather than rookie in terms of calculating rankings.
This could also allow me to implement things like "seasons" where this deviation parameter is periodically pushed high to stir things up a bit.
But as with many other things, this is still on my todo list and not fully decided on and there are plenty of things higher on said list.. We'll see |
|
|
Smeat
Joined: Nov 19, 2006
|
  Posted:
Nov 25, 2013 - 18:06 |
|
thx for the explanation! Carry on w/ things more important, by all means. |
_________________ Let's go A.P.E.!
(...and what exactly do you think they do with all those dead players?...) |
|
Badoek
Joined: May 17, 2009
|
  Posted:
Nov 26, 2013 - 08:12 |
|
I really feel like Christer is ignoring the Chainsaw Rating. Quite disrespectful tbh.
He could have replied "Shut up Chainsaw, or I'll rename you to Pony1337". |
_________________
|
|
Kelkka
Joined: Aug 14, 2009
|
  Posted:
Nov 26, 2013 - 08:25 |
|
Juts out of curiosity, is the win probability based on your overall title/CR or the racial title/CR involved in the match? |
|
|
BillBrasky
Joined: Feb 15, 2005
|
  Posted:
Nov 26, 2013 - 10:53 |
|
Maybe you should only be a legend if you beat BillBrasky 1000 times!!!!
Otherwise you are just a lame hipster or something |
|
|
Rabe
Joined: Jun 06, 2009
|
  Posted:
Nov 26, 2013 - 11:46 |
|
And you can never become legend (again) once you've been beaten by Rabe! \o/ |
_________________ .
|
|
DrPoods
Joined: Nov 14, 2013
|
  Posted:
Jan 22, 2014 - 16:15 |
|
I'm not sure if the is the right place to ask but I have been looking around and cannot find what I am looking for.
With regards to Coach ratings, I am just wondering why my overall ranking is so low. I realise I struggle a bit at the moment but the lowest team rating I have is 139.88 (Khemri), all the others being higher than 140, yet my overall is 132.44.
Being absolutely hopeless at mathematics I probably am missing something but I just don't know how the overall ranking can be that much lower than all of the racial ratings.
Can someone help explain this to me? Thanks. |
_________________ "Gallifrey falls no more"
Do your part! Join the Adoption Agency NOW! |
|
Hitonagashi
Joined: Apr 09, 2006
|
  Posted:
Jan 22, 2014 - 16:17 |
|
Sure.
Let's say that you start at 150. Each loss costs you 1 CR (big hit!).
You play 5 games, one with each race, and lose all 5. All your racial averages are now at 149, and yet your "combined" is at 145.
That's basically how CR works AFAIK; a game affects the racial and combined/league stats whenever it is finished. |
_________________ http://www.calculateyour.tv - an easy way to work out specific team builds.
|
|
p4m
Joined: Aug 02, 2003
|
  Posted:
Jan 22, 2014 - 16:17 |
|
Rabe wrote: | And you can never become legend (again) once you've been beaten by Rabe! \o/ |
this is a false statement |
|
|
DrPoods
Joined: Nov 14, 2013
|
  Posted:
Jan 22, 2014 - 16:21 |
|
If it helps, I found the thread "coach rank confusion" but I am still helplessly confused! It doesn't hugely matter in the grand scheme of things but I would like to know. |
_________________ "Gallifrey falls no more"
Do your part! Join the Adoption Agency NOW! |
|
DrPoods
Joined: Nov 14, 2013
|
  Posted:
Jan 22, 2014 - 16:23 |
|
Hitonagashi wrote: | Sure.
Let's say that you start at 150. Each loss costs you 1 CR (big hit!).
You play 5 games, one with each race, and lose all 5. All your racial averages are now at 149, and yet your "combined" is at 145.
That's basically how CR works AFAIK; a game affects the racial and combined/league stats whenever it is finished. |
Ah OK. So essentially because I have been crappy it hurts the overall ranking, but because I mixed and matched teams a lot each individual race stays reasonably strong?
But why then would 1 game with a certain race go from starting at 150 to say 146 or so? Isn't that a large drop? |
|
|
backelie
Joined: Jul 20, 2010
|
  Posted:
Jan 22, 2014 - 16:47 |
|
Rabe wrote: | And you can never become legend (again) once you've been beaten by Rabe! \o/ |
Sounds horribly unfair considering you're a box necro legend |
|
|
|
| |