So Following Up with real numbers from Allen world generated last night:

Ranking    OLD      NEW    Change
1 to 5          775        775
9 to 11       715         722      +7                         
24 to 26     614        628       +14
48 to 50     581        595       +14
74 to 76     564        559       -5
98 to 100   555        553        -2
124 to 126 533       539       +7
149 to 151  525       528       +3
174 to 176  516       523        +7
198 to 200   508      513       +5

Seble's 40 point gap between #10 and #25 in his test world certainly didn't translate to Allen world last night where is still 94.  (Down from 101 pre-update).  #25 to #175 is now 105, up from 98. 

So a little tiny bit better, but my thesis, that the gap from #10 to #25 (94pts) is the same as #25 to #175 (105pts) is still pretty much true. 

I guess I have to say #10 to #25 (94pts) is equal to #25 to #140 or so now.  Overall its better, recruits in the 15-50th rating got a boost, which was needed.  Maybe not as big a boost as I would have liked.

This is based on 15 player samples, so its not 100% accurate.

Thoughts?

I actually would like to see this done in small amounts like it appears here over 2-3 seasons. I really hate it when a "small" tweek throws the game out of whack for 4 seasons. If the recruits are alot better... and I happen to have 6 open schollys... I'm a happy guy. If I filled 6 last season with the last of the worst recruits the game has ever created... I'm not a happy guy.
Now if your numbers reflect all recruits in Allen... and this is the only bump... then nothing much has changed for anyone.
But if this isn't enough... and Seble adds this much again next season... and possibly again once more to reach what he's looking for. I'm 100% behind the "subtle" change.

my 2cents

i put together a histogram of the overall ratings of the top 300 d1 recruits in allen. without going through the trouble of posting an image of it, i can tell you that so far it does NOT look like a normal distribution.

it looks like there's a dearth of players from 634-684 and a surplus from 684-784. not to mention the a bunch of the highly rated players appear WAY out there.

again, this isnt a huge dataset, but anyone who knows what a z-score is would see an issue with the distribution.

with that size sample, couldn't even one unusually good player throw it off?  i mean, 9 - 11 and 24 - 26 is a sample size of six players total, only three in each set.
 

 

I used 15 players (3 C's 3 PF's etc) at most of the zones.  Not a ton, but  better than just 3.

Posted by jetwildcat on 3/21/2011 12:53:00 PM (view original):
i put together a histogram of the overall ratings of the top 300 d1 recruits in allen. without going through the trouble of posting an image of it, i can tell you that so far it does NOT look like a normal distribution.

it looks like there's a dearth of players from 634-684 and a surplus from 684-784. not to mention the a bunch of the highly rated players appear WAY out there.

again, this isnt a huge dataset, but anyone who knows what a z-score is would see an issue with the distribution.

Z-score?  Another stats guy.  Love it.   

For sure its not normal distribution, but I don't think that is a good idea.  (to make it normal).

I don't think it should be a normal distribution. It should be skewed right. 

I'd like it to be closer to linear from #1 to #200. 
NOTE:  I do NOT want linear from 1-200.  That sucks since there isn't enough difference at the top.

Now its like 1+[(1-10x)/(1+10x)] and I want it to be like 1+[(1-2x)/(1+2x)]

Something like that. 

Posted by tianyi7886 on 3/21/2011 3:01:00 PM (view original):
I don't think it should be a normal distribution. It should be skewed right. 

i think it should be a far-right portion of a skewed right distribution, actually. where if you took a normal distribution of the ratings for everyone that age in the country, it would be skewed right, and the recruiting pool is the 99.8th percentile (approx 2 million males at each age in US, assuming the 4000 recruits or so are the best)

Posted by jetwildcat on 3/21/2011 4:26:00 PM (view original):

Posted by tianyi7886 on 3/21/2011 3:01:00 PM (view original):
I don't think it should be a normal distribution. It should be skewed right. 

i think it should be a far-right portion of a skewed right distribution, actually. where if you took a normal distribution of the ratings for everyone that age in the country, it would be skewed right, and the recruiting pool is the 99.8th percentile (approx 2 million males at each age in US, assuming the 4000 recruits or so are the best)

thats kind of what we see in hd, no?

Posted by tianyi7886 on 3/21/2011 4:54:00 PM (view original):

Posted by jetwildcat on 3/21/2011 4:26:00 PM (view original):

Posted by tianyi7886 on 3/21/2011 3:01:00 PM (view original):
I don't think it should be a normal distribution. It should be skewed right. 

i think it should be a far-right portion of a skewed right distribution, actually. where if you took a normal distribution of the ratings for everyone that age in the country, it would be skewed right, and the recruiting pool is the 99.8th percentile (approx 2 million males at each age in US, assuming the 4000 recruits or so are the best)

thats kind of what we see in hd, no?

what i was seeing is a steep dropoff after 634 then a plateau out to 784. then a few dots for the high players.

of course it's a relatively small sample size.