Posted by dahsdebater on 2/18/2012 4:59:00 PM (view original):
Not particularly significant, actually.  Both Def*ATH and Blk*Ath were significant, also Def*SPD.  I still suspect they all actually interact to produce, at most, an outside and inside overall defensive efficiency rating, and I'm sure if I'd manipulated the model a bit I could have made it look however I wanted with regards to interactions.  I was more interested in regressions based on individual ratings for what I was looking at.

I'm with you on the importance of Blk on defense, at least with regards to bigs. Not 100% sure on the effect of blks for guards, partially because most guards I recruit have terrible blk. 

Posted by jck45 on 2/18/2012 1:26:00 PM (view original):
I was wondering if any coaches out there were feeling like they were maybe overvaluing any attributes. That is, placing too much value on a specific attribute during the recruiting process.  Conversely, are there any coaches that think they might be undervaluing, or not emphasizing enough, any given attribute?

Personally, I think I may tend to focus too much on passing and ball handling (even for my bigs) and let weak rebounding slide (again, even for my bigs and especially at the SF). For example, if a recruit can't BH or Pass well, I might be too quick to nix him BUT if he can't realize a high REB, I might be too quick to let him slide...

Any fellow coaches find themselves maybe overvaluing or undervaluing anything? 

bh and passing are of very little value (relatively) for bigs. rebounding is absolutely critical. if i had to make a list of my 5 cliches that sum up my advice to coaches, this one would definitely make the cut - if your team can't rebound, it doesn't really matter if they can do anything else. you just can't let weak rebounding slide for your bigs! i mean, its one thing to have a guy who is a great rebounder, another who is an awesome rebounder, and then let one dude be just a good rebounder so you could get a low post presence. but to let it go for bh/pass - that is going to cause a lot of damage!

on the other hand, bh/pass are very important for your 1/2, much more so than rebounding. but, rebounding is way more beneficial for your guards than bh/pass is for your bigs. just remember that bh/pass for your bigs are nice-to-haves, that you should not concern yourself with until you get the must-haves (for example sake, if you offered me an otherwise strong big man, with 70 reb and 50 bh and 50 pass, or a big with 90 reb and 1 bh and 1 pass, i would not hesitate to take the 90 reb big)

Posted by dahsdebater on 2/18/2012 3:13:00 PM (view original):
I think most coaches undervalue BLK.  They just kinda look at it as an afterthought and consider DEF and ATH the most important determinants of defensive ability, with SPD 3rd.  Based on some regression over hundreds of teams I ran a year or so ago BLK correlates to OPPFG% more than any other individual rating, even for guards.

be VERY careful. there was once a coach many looked to for advice, named jdtapp, and he ran a similar regression. it is SUPER misleading and flat-out wrong. funnily, he quit the game, claiming it was too simple from a statistics standpoint - but he was not overly successful, and his conclusions were so far off, i felt bad there were coaches reading them and believing them. at the time, i was a new coach who got no respect, and people already shat on me enough for claiming game planning mattered, so i just let it slide..

in truth, blocking can be an afterthought in the press with little impact. it is more important in the other defenses, but still not as important.

dahs - ill do you a solid favor, if you like. im assuming you have at least basic training in statistics. try to figure this one out - ill help you along if you need it - why is running a regression like you did, between individual rating and opp fg%, essentially totally and completely meaningless? hint - what assumptions were you making, implicitly and explicitly (most likely, implicitly) in your work? i believe jdtapp ran against overall points scored or something, which was actually worse than what you did - technically, not theoretically - why?

Posted by isack24 on 2/18/2012 4:00:00 PM (view original):

Posted by dahsdebater on 2/18/2012 3:13:00 PM (view original):
I think most coaches undervalue BLK.  They just kinda look at it as an afterthought and consider DEF and ATH the most important determinants of defensive ability, with SPD 3rd.  Based on some regression over hundreds of teams I ran a year or so ago BLK correlates to OPPFG% more than any other individual rating, even for guards.

You apparently have the stats to back it up, but I have a really tough time believing BLK is more important to OppFG% than DEF.

good for you. the biggest thing i can advise, as my HD career slowly winds down, it to remain skeptical. i am probably the biggest skeptic on these forums (you can do it respectfully, despite what many coaches here would lead you to believe through their disrespectful and attitude-full posts). you CANNOT take what CS says as absolute. you CANNOT take what a vet, no matter how successful, it doesnt matter if its me or OR and lostmyth or anyone else, as absolute. you can take them as guidelines, as hints for your own research. i can't tell you how many times ive seen CS and vets be horribly wrong. i am lucky to have the problem solving background to figure a lot of this stuff out myself, and to have the confidence to believe something when the whole community and CS said the opposite. sure, ive been wrong (another reason not to listen to vets - few have a better mastery of this game than me, and i have been wrong plenty of times). but ive also been right, and ive also seen guys the whole world shat on triumph, when the whole of the community was against them.

example - for a while, pulldowns were a gigantic disaster, they kept changing, and CS kept saying the opposite. once, furry_nips, a guy who was as respected as a retarded crack-smoking monkey, and maybe 1 other coach, claimed pulldowns were back. EVERYONE i mean EVERYONE said they were wrong, pulldowns could not exist, as CS said so. this actually happened at least 2 maybe 3 times in that year of chaos. as soon as i read their post, i explored - they were right, pulldowns were back. it took a good 4 months for vets and CS to eat their words, although almost nobody admitted it. meanwhile, furry_nips, myself, and whomever else realized god gave you 2 ears and 1 mouth for a reason, held a huge advantage over others (although maybe in this case, i should say, god gave you 2 eyes and 1 mouth for a reason. or 1 finger. errrr.... i guess there are 10 of those. so um, yeah, 2 ears and 1 mouth =)

the key is, you should never write off a coach just because he disagrees with CS or OR or anyone else. but you should also never take something as an absolute, no matter how many of the games' best agree. for a year, when potential came out, i advocated repeatedly how you could safely practice 0 minutes on a capped out category. and virtually everyone, over 95% of the posts responding to mine, continued to **** on me, quoting CS nonsense. i already had over a dozen titles at the time. but that shouldn't matter, much at least. the fact is, i knew you could go 0 minutes, because i did it. the people all bashing me? they DIDNT EVEN TRY!! or else they would have known. its amazing how closed minded people can be. "i never tried it, but you MUST be wrong!". if people act that way to me, with 15 championships at the time and two of the greatest runs in the history of the game, what chance does a young, passionate, intense coach without a title to his name have of sharing an unpopular discovery? i have benefitted tremendously from giving those kinds of guys a shot, trying out what they said, and often getting a major advantage from it.

so, in conclusion - NEVER take anything as an absolute. never believe because someone has stats (no offense dahs, none at all) or because a guy works for WIS or because a guy has 30 titles, they are right. and NEVER believe because a guy has 0 titles or because you don't know of him, he is wrong. keep an open mind, question everything, test everything you can. that is the path to success. and, if i do say so myself, it is considerably more fun than wandering around like a sheep following his shepherd. baaaaaaa.

I'm not sure what, specifically, you're suggesting was wrong with my math.  I know of several significant flaws:

1) I only looked at press and man.  They looked generally the same, but I don't know for sure how anything applied to zone.  I went through thousands of box scores, and there's no way I was taking extra time to average defensive ratings or go through box scores.

2) This is the big one: since I didn't go through box scores, I just assumed starters played all their minutes against one another.  Obviously starters played some minutes and had some shots against bench players.  All of that throws off the analysis.

3) I made no effort to differentiate between inside and outside defense, which may well be different in the engine.

Some others, but I'm really tired right now.  Maybe will talk about it some more tomorrow afternoon.

"but you should also never take something as an absolute"

Hooray!  Hehe.

I used to greatly undervalue ballhandling, even for guards. 

I will comment on one of BillyG's points on Furry.

Billy is correct that Furry was outspoken in the forums that pulldowns did exist in the face of what was communicated down from CS.

But a possible reason that many coaches teamed up on him, myself included, was that he was a pompous little ***** about it and did exactly what Billy warns of and believed in absolutes and ignored differing opinions.  Furry was vocal (and correct) that pulldowns existed, but he also was quite vocal that there was only one way that a recruit could be pulled down and that this method was 100% effective at all levels.  Even though he hadn't tried it at all levels.

However contrary to what Furry said, or what CS said, there were still multiple methods to pulldown recruits.   Myself, and I'm assuming many other coaches also continued experimenting with multiple methods of pulling down recruits.   Once each coach discovered an approach it almost made recruiting boring, at least at the D2 and D3 levels. 
Remember when the recruits weren't revealed until 5pm the night of recruiting?  It wasn't that difficult to look at a recruit, and in 5 minutes gauge his 4-year improvement and then decide to pulldown using method A, which might cost $3500 in cycle 2, or wait and use method B which would cost $3000 in cycle 12 or $2350 in cycle 15.
  
But when it was brought up that other methods also worked for pulldowns and no single method was 100% effective at all levels, Furry basically called those with differing views names and told them they didn't have a clue about recruiting.   I think this attitude towards others helped greatly in converting almost every Furry post into an automatic hot topic.

Posted by dahsdebater on 2/18/2012 11:38:00 PM (view original):
I'm not sure what, specifically, you're suggesting was wrong with my math.  I know of several significant flaws:

1) I only looked at press and man.  They looked generally the same, but I don't know for sure how anything applied to zone.  I went through thousands of box scores, and there's no way I was taking extra time to average defensive ratings or go through box scores.

2) This is the big one: since I didn't go through box scores, I just assumed starters played all their minutes against one another.  Obviously starters played some minutes and had some shots against bench players.  All of that throws off the analysis.

3) I made no effort to differentiate between inside and outside defense, which may well be different in the engine.

Some others, but I'm really tired right now.  Maybe will talk about it some more tomorrow afternoon.

you are thinking little. i am not nit picking these things. they, over large enough sample size, could arguably all even out (arguably not, but that is not where i am going :)

forget this game. what key assumption are you making any time you do a regression like you are doing? it been a number of years, so maybe i am not thinking of it all right, but there is at least one critical assumption you are making about the nature of the variables in question. if true, a regression is a perfectly viable tool. if not... not so much.

Posted by oldwarrior on 2/19/2012 2:07:00 PM (view original):
I will comment on one of BillyG's points on Furry.

Billy is correct that Furry was outspoken in the forums that pulldowns did exist in the face of what was communicated down from CS.

But a possible reason that many coaches teamed up on him, myself included, was that he was a pompous little ***** about it and did exactly what Billy warns of and believed in absolutes and ignored differing opinions.  Furry was vocal (and correct) that pulldowns existed, but he also was quite vocal that there was only one way that a recruit could be pulled down and that this method was 100% effective at all levels.  Even though he hadn't tried it at all levels.

However contrary to what Furry said, or what CS said, there were still multiple methods to pulldown recruits.   Myself, and I'm assuming many other coaches also continued experimenting with multiple methods of pulling down recruits.   Once each coach discovered an approach it almost made recruiting boring, at least at the D2 and D3 levels. 
Remember when the recruits weren't revealed until 5pm the night of recruiting?  It wasn't that difficult to look at a recruit, and in 5 minutes gauge his 4-year improvement and then decide to pulldown using method A, which might cost $3500 in cycle 2, or wait and use method B which would cost $3000 in cycle 12 or $2350 in cycle 15.
  
But when it was brought up that other methods also worked for pulldowns and no single method was 100% effective at all levels, Furry basically called those with differing views names and told them they didn't have a clue about recruiting.   I think this attitude towards others helped greatly in converting almost every Furry post into an automatic hot topic.

i don't disagree that furry had it coming. that was just an extreme example of what happens on the forums all the time. and, for what its worth - i think he was usually right that a single method was 100% effective at all levels. there were all these crazy theories going around, often wrong, that lead to a lot of bad conclusions as to if things worked reliably or not. with a true understanding of the methods at hand, mostly, they always worked. but there was certainly a time that was not the case either. im not saying fury was always right - however, he was righter than virtually every veteran, championship-laden coach who was posting - and infinitely more on target than admin's posts at the time. and im not suggesting he went about it the right way, nor that he deserved better treatment than he got. all im saying is, even him, the way he was then ( he got a lot better ), had the capacity to be right when CS and the veteran coaches all disagreed. a guy who most would claim was the most off-base coach in the whole game. if he had that capacity - then doesn't any coach? doesn't any coach deserve a fair shake on the forums, instead of being immediately written off for presenting a contrary or unpopular idea? (ok, maybe any coach but a few, that you can count on one hand, who obviously have no interest in a real dialogue).

Posted by dahsdebater on 2/18/2012 11:38:00 PM (view original):
I'm not sure what, specifically, you're suggesting was wrong with my math.  I know of several significant flaws:

1) I only looked at press and man.  They looked generally the same, but I don't know for sure how anything applied to zone.  I went through thousands of box scores, and there's no way I was taking extra time to average defensive ratings or go through box scores.

2) This is the big one: since I didn't go through box scores, I just assumed starters played all their minutes against one another.  Obviously starters played some minutes and had some shots against bench players.  All of that throws off the analysis.

3) I made no effort to differentiate between inside and outside defense, which may well be different in the engine.

Some others, but I'm really tired right now.  Maybe will talk about it some more tomorrow afternoon.

not really suggesting anything wrong with the math - your math could very well have been perfectly correct. just to be safe - i am definitely not trying to criticize you or anything you did! just pointing out an important flaw. thought it might be interesting to hear what you would think of.

anyway - what i am thinking of is this. you have a model that sort of goes like this -
you have 10 variables attached to a player, call them a b c d e .... (like ath, spd, def, lp, passing all are. the name is irrelevant). then, you are looking at - really, comparing against some part of the players performance - it could be their points scored, rebounds, opponents' rebounds, etc... and saying, what impact does each individual variable have. well, players with 50-60 of variable a, maybe score 10 points, with 80-90, maybe score 15 points. the same might be true for variables b and f. so, you might conclude, all have an equal importance in how effectively a player can score.

the above conclusion could be challenged because there could be other factors impacting the results - for example, maybe a guy with 80-90 of variable a or b or f is getting treated differently by his coach - or the opponent's coach. if that was true, there is really a hidden variable - the actions of the coach. if the actions of the coach are *dependent* on one of these variables - then you cannot know what really caused the outcome. for example, is it because the player went from 50-60 in a rating to 80-90, he got that much better at scoring? or is it because the coach played him more, so he got more opportunities? or did the coach play him the same amount of time, but let him take more shots? 

well, the above principle doesn't just apply to the numerous hidden variables present... it also applies to the named ones we are studying, a, b, c, etc... what if a is always half of b? then when you compare a player's performance at 30a vs 40a, you are also getting the impact of 60b vs 80b. a might not mean a damn, but if b does, then it makes a look important. this kind of thing can happen no matter what the relationship of a and b - if there is a relationship. the only way the model makes sense is if all the variables are *independent*. i am trying to explain in a way people not trained in statistics may follow...

so yeah - it really comes down to, are the ratings you are considering independent of each other? of course, we KNOW there are "hidden" variables that they are dependent on - im sure everybody reading can think of a dozen ways a coach might adjust his behavior as his players' ratings change. well, in HD - the answer is no.

people talk about height and weight not mattering. that is true in games - but it matters in recruit generation. for example, taller players generally have better sb and reb. its very significant - if you've never seen it, you can look through big men recruits for about 2 minutes and you should be able to see it. well, along those lines - sb definitely impacts reb. a tall player will generally have high reb and high sb - therefore, to some extent (and it is significant, but possibly lesser), a player with high sb is likely to have high reb. there are quite a few other correlations as well, in recruit generation. (if you struggle to think of any, check out def vs ath/spd - see anything?). of course, its not just height and weight that don't matter in game, but impact recruit generation - there is also the position. as a result, there are tons of little relationships in the variables. think about it for a second - a guy with 80 ball handling is BOUND to have better passing, on average, than a guy with 1 ball handling. would anybody dispute that? so, how do you know, when you look at the impact of bh on fg% - if it is the bh or the passing that is causing that? 

IMO WE is the biggest attribute that is not properly evaluated.

I think that the confounding you are suggesting is much more of a problem analyzing offense than defense.  That's a big part of why I didn't bother.  Also because it's easier to learn about offensive efficiency based on the eye test, just watching what happens.  The game compiles offensive stats for us.  No such advantages for defensive stats.  Also, other than double teams, which are uncommon enough not to be a major confounding problem, the only defensive setting that coaches can adjust is +/-.  I figure over a large enough sample those should largely balance each other.  Also, a large portion of the sample was sim coached, so they always come out in a 0.  I really think the biggest flaw in the analysis is the assumption of starter-on-starter matchups.

Posted by dahsdebater on 2/19/2012 4:36:00 PM (view original):
I think that the confounding you are suggesting is much more of a problem analyzing offense than defense.  That's a big part of why I didn't bother.  Also because it's easier to learn about offensive efficiency based on the eye test, just watching what happens.  The game compiles offensive stats for us.  No such advantages for defensive stats.  Also, other than double teams, which are uncommon enough not to be a major confounding problem, the only defensive setting that coaches can adjust is +/-.  I figure over a large enough sample those should largely balance each other.  Also, a large portion of the sample was sim coached, so they always come out in a 0.  I really think the biggest flaw in the analysis is the assumption of starter-on-starter matchups.

R-squared for your regression?

For the record, I believe that running proper statistical tests is quite helpful for WIS games. In GD, slid64r ran regressions with thousands of observations, and pretty much figured out the old GD engine.