BuzzerBeater Forums

BB Global (English) > Skill cap testing

Skill cap testing

Set priority
Show messages by
This Post:
33
155261.52 in reply to 155261.51
Date: 9/6/2010 9:31:30 PM
Overall Posts Rated:
155155
I just want to also make the comment (if I have not already, I can't remember) that any data sent to me will be kept strictly confidential. Even if I get help from others on this study, I will only discuss data at the macro level and I will not release skills from individual players. And hopefully it also goes without saying that I will not discuss individual players in any public forum.

Run of the Mill Canadian Manager
This Post:
00
155261.54 in reply to 155261.51
Date: 9/7/2010 5:00:16 AM
Overall Posts Rated:
102102
I thought about something.

The GM dameisel proposes to gather data thanks to his firefox add-on FoxyBB, and now also on the website (http://foxybb.tk/) in order to improve some formulas for this add-on.

Perhaps a partnership might be possible with people that would like to lead studies about skills like yours

Last edited by Dunker Joe at 9/7/2010 5:01:07 AM

BBF, le forum francophone : = (http://buzzerbeaterfrance.forumpro.fr/)
This Post:
00
155261.55 in reply to 155261.54
Date: 9/7/2010 8:02:23 AM
Overall Posts Rated:
155155
I thought about something.

The GM dameisel proposes to gather data thanks to his firefox add-on FoxyBB, and now also on the website (http://foxybb.tk/) in order to improve some formulas for this add-on.

Perhaps a partnership might be possible with people that would like to lead studies about skills like yours


If he wants to share some information about players who are capped (and provided he knows a bit of their training history), I am more than happy to take it. But because I have guaranteed confidentiality, I will not be sharing the other way. Sorry.

Run of the Mill Canadian Manager
This Post:
00
155261.56 in reply to 155261.53
Date: 9/7/2010 8:07:11 AM
Overall Posts Rated:
155155
An interesting suggestion. However, this data would not be useable in the model since it requires the precise moment of capping in order to be useful. At least that is how I see it.

I would be open to such a thing if someone could explain to me how to model effectively with uncapped players.

I could see some small use for it (ie: if I could get enough players, maybe by observation I could determine if sub-levels on potential exist). But just the fact that sub-levels exist does not get me any further along in the model (although, determining that they don't exist might).

Run of the Mill Canadian Manager
This Post:
00
155261.57 in reply to 155261.55
Date: 9/7/2010 8:08:40 AM
Overall Posts Rated:
102102
I am not connected to his project
(I just participated in the french translation)

BBF, le forum francophone : = (http://buzzerbeaterfrance.forumpro.fr/)
This Post:
44
155261.60 in reply to 155261.58
Date: 9/7/2010 10:58:03 AM
Overall Posts Rated:
155155

Why is it important to ask the managers if their players are capped are not.


Because this is a regression model, something like y = B1x1+B2x2+... where y is the potential and the xs are the skills. If a player is capped, I know y=the player's potential (plus or minus some fuzz due to sub-levels).

If the player is not capped, I have no idea what y is. It is probably something below the player's potential (if they are not capped), but beyond that, I do not know.

...
What is the highest possible REB (for example) of all the benchwarmer potentials. ...


This is kind of suggesting to just take players off the transfer list. Again, I have the same issue as I described above. Even if it seems certain that they are capped, there is nothing to say that they were not trained beyond their cap. Indeed, it is even impossible to say that they are capped. The only way to know is to have an idea of how training went.


Given a large data pool


I still stand by what I said before. With a large pool of bad data (or data with too many question marks), all you are going to get is a bad model.

Anyhow, no worries. I started to get more data. It may take some time, but I believe I will get there. And the last thing I want to do is take a bad path just to get answers faster.

Run of the Mill Canadian Manager
This Post:
11
155261.61 in reply to 155261.60
Date: 9/7/2010 11:57:31 AM
Overall Posts Rated:
409409

Anyhow, no worries. I started to get more data. It may take some time, but I believe I will get there. And the last thing I want to do is take a bad path just to get answers faster.


+1

This is highly recommended when trying to understand something in opposition of just giving an answer.

Last edited by Zero, the Magi. at 9/7/2010 11:57:51 AM

This Post:
00
155261.62 in reply to 155261.61
Date: 9/7/2010 12:29:42 PM
Overall Posts Rated:
155155

I just wanted to share a bit more - this one is more something I am starting to believe from looking at player skills rather than something I modelled, though. As I get more and more data, I am starting to believe more in the idea of sub-levels on potential. This is tough for me to admit, since I have been adamantly against them for a long time. But hey, even smart people can get things wrong sometimes. ;-)

When I talked to Joseph Ka, he said something that I thought was interesting: that if sub-levels do exist, maybe all-time great is just a limit case of hall of fame potential. I also heard of a rare potential called "BB developer". I assume that would be the opposite (ie: a limit case of announcer). In order words, something like this:

BB developer = 0 (which possibly means capped no matter what)
announcer = 0-1
bench warmer = 1-2
role player = 2-3
6th man = 3-4
starter = 4-5
star = 5-6
allstar = 6-7
perennial allstar = 7-8
superstar = 8-9
MVP = 9-10
hall of famer = 10-11
all-time great = 11 (which possibly means never capped)

Anyhow, I do not know anything for sure yet, but if true it seems pretty interesting.

Run of the Mill Canadian Manager
Advertisement