Posted by dacj501 on 12/5/2011 9:09:00 AM (view original):
7-20 should never be in the postseason. 22-6 with a 66 RPI seems like it ought to be good enough for the PIT IMHO.

100% agree on both points.

Posted by jtt8355 on 12/5/2011 9:01:00 AM (view original):
sorry if this is a repeat of someone else's post, but: one of the coaching 'talents' in this game is knowing how to schedule. there's more to it (or there should be more to it) than predicting who the best teams will be next season and garnering a tough schedule...part of the coach's job should be scheduling the toughest schedule that they can win - either because their team is more talented than the other teams or by out-coaching their opponents. to put a 7-20 team in the post season is, imho, rewarding mediocrity. similarly, if the only teams that a coach / team can beat are sub-par (bc they're less talented or coached by sim, etc), the system should not be rewarding them either.

There we go.

This.
 

Quick update:

I've looked over the logic and I'm fairly settled on the following tweaks:
- More value for winning the game
- Less weight on opponent strength
- Less weight on margin

With these tweaks, here are the rank changes for some of the Naismith teams that have been discussed here:
Manhatten 107 to 88
UNC 67 to 75
FSU 80 to 108
S. Carolina 48 to 59

So essentially, the teams with high SOS but not great records are dropping, and the lesser SOS teams with better records are rising some.  I should also mention that there is a cap on margin at 20 points, so anything over 20 is treated the same as 20.  That was in place from the beginning, but I never mentioned it.

Can I ask why this wasn't beta-tested, if in fact it wasn't?  It was very weird to see a new formula, still with a ton of kinks in it, get debuted and put into action immediately.

Is there logic in the code to make sure a team with a losing record does NOT get in the NT unless they win their conference?

Posted by dirtyjerk on 12/5/2011 10:32:00 AM (view original):
Is there logic in the code to make sure a team with a losing record does NOT get in the NT unless they win their conference?

Yes.  There always was.

Thanks 

Posted by seble on 12/5/2011 10:01:00 AM (view original):
Quick update:

I've looked over the logic and I'm fairly settled on the following tweaks:
- More value for winning the game
- Less weight on opponent strength
- Less weight on margin

With these tweaks, here are the rank changes for some of the Naismith teams that have been discussed here:
Manhatten 107 to 88
UNC 67 to 75
FSU 80 to 108
S. Carolina 48 to 59

So essentially, the teams with high SOS but not great records are dropping, and the lesser SOS teams with better records are rising some.  I should also mention that there is a cap on margin at 20 points, so anything over 20 is treated the same as 20.  That was in place from the beginning, but I never mentioned it.

good tweaks

Posted by seble on 12/5/2011 10:01:00 AM (view original):
Quick update:

I've looked over the logic and I'm fairly settled on the following tweaks:
- More value for winning the game
- Less weight on opponent strength
- Less weight on margin

With these tweaks, here are the rank changes for some of the Naismith teams that have been discussed here:
Manhatten 107 to 88
UNC 67 to 75
FSU 80 to 108
S. Carolina 48 to 59

So essentially, the teams with high SOS but not great records are dropping, and the lesser SOS teams with better records are rising some.  I should also mention that there is a cap on margin at 20 points, so anything over 20 is treated the same as 20.  That was in place from the beginning, but I never mentioned it.

this is nice, but what about the conferences who got 4 teams in one bracket, and none in another? the system needs to move teams from lines in the s-curve where this happens... was never an issue before.

I'm looking into region placement now.  It's not easy logic to write, but I'll do my best to distribute teams more.

Posted by seble on 12/5/2011 10:59:00 AM (view original):
I'm looking into region placement now.  It's not easy logic to write, but I'll do my best to distribute teams more.

Thanks for the quick responses, seble, and willingness to listen to feedback. This level of engagement with the community is greatly appreciated.

Agreed. Thanks to seble for his hard work and quick responses.

If margin is going to be used, 20 points is still too high for the cap.

I should also mention that there is a cap on margin at 20 points, so anything over 20 is treated the same as 20.  That was in place from the beginning, but I never mentioned it.


Is there any thought on trying to normalize the margin of victory to the pace of the game, or to total points scored? 
Or will a 50-30 win be viewed the same as a 100-80 win. 

Posted by dacj501 on 12/5/2011 11:52:00 AM (view original):
If margin is going to be used, 20 points is still too high for the cap.

It depends on how you do it.  I'm pretty sure 20 is what Sagarin uses for football (or used to)... but I have read that in his programming, every point of margin matters less than the previous one.  So in calculating the MOV factor for ratings, a 2-point win is less than twice as good as a one point win, a 3 point win is less than 1.5 times as good as a 2-point win, etc. 

I assume seble is doing this already, but if not this might help a lot.