Posted by hughesjr on 2/3/2011 5:23:00 PM (view original):

Posted by coach_billyg on 2/3/2011 4:58:00 PM (view original):
i still think it is retarded to mess with d2/d3 recruit generation to fix a d1 problem. also, i fee like one of the biggest problems was not that there weren't enough players who could come in and compete with high end big 6 school on day 1... its that there aren't enough of these players who can come in and compete with high end big 6 schools as juniors and seniors. so, guys will start better and will max out earlier. i just don't see the benefit there... we can all go back to putting a little more practice time into team practice and a little less of our own time worrying about practice plans? but meanwhile, upperclassmen, which are what make teams great, really haven't gotten any better.

i do think this could impact the elite players - sounds like they got worse, with more potential. so, they are probably going to contribute even less when they leave early, which sucks. but, they might stay longer, which is good i guess, maybe. im not sure what the net result will be on top teams, guess we will have to wait and see.

all in all, it seems like this change is just an attempt to reverse what seble sought out to do in the first place - which was to create elite incoming players and create more room for growth for players. these are concepts that i supported, but it just went from elite players to mediocre players with very little in between, and that was the problem. so it seems like to me, instead of making those not elite but not mediocre players more plentiful, seble is reducing incoming elite players and creating less room for growth for players. just doesn't make a whole lot of sense to me.

From what seble said, they generate all the recruits at the same time.  So unless he is going to single out D-1 recruits with some "after generation" processing, then it is modify all recruits or modify no recruits.  Right???

First, if some "after generation" processing is feasible, then absolutely, let's do that.

Beyond that, the last change to recruits made the elite guys better and everyone else worse. So it does appear that they have the ability to make a change beyond simply making everyone better or everyone worse.

Posted by dahsdebater on 2/3/2011 6:29:00 PM (view original):
You guys keep talking about this recruit generation "curve."  I thought it was made clear in this dev chat that no such thing exists.  Recruits appear to be generated by having a mean and a variance applied individually to each one with everybody starting from the mean.  It looks like all that Seble probably did to change the generation was to perhaps increase the variance a bit, leading to more and bigger outliers.  There doesn't appear to be any funtion defining the number of players with any given overall talent range.  I'm rather glad that there isn't...  that would probably end up making recruit generation much more predictable over the long term.  But assuming that this is the way it functions, the only way to "change the shape" would be to totally overhaul the recruit generation process or insert an extra step to arbitrarily improve players seen as potential mid-major recruits as I described briefly above.  The "curve" currently associated with recruit generation is inevitably a bell curve and there's no simple way to change that.

if it was just mean and variance, you would see a similar effect on both ends of the curve. d3 recruits look roughly the same to me now as they always have. the top of d1 is totally different. that wouldn't make any sense in the scenario you describe. so i think you are really over simplifying things. likely, recruit generation is already significantly more complex than you've outlined. besides it just doesn't make any sense that seble would effectively have a "mean" and "variance" variable that he could just tweak. mean and variance is computed off a function. an infinite amount of functions can result in the same mean and variance. when seble talks about the mean and variance changing, he is just using those terms as a way to characterize the change he made to the function - the recruit generation function. and functions yield curves. so when i am talking conceptually changing the curve, i am talking about changing the output of the function in a certain way - not sliding something on some magical curve picture somewhere. curves are an excellent way to describe the output of a function, its a very reasonably way to talk about these kinds of things IMO.

Posted by moy23 on 2/3/2011 5:57:00 PM (view original):

Posted by gillispie on 2/3/2011 5:47:00 PM (view original):

Posted by oldresorter on 2/3/2011 5:34:00 PM (view original):
this is the Q&A for the 5th ?, seems to be a much more major change than the rhetoric might imply, currently if I am reading this right, 9 thru 11 average 715, 24 thru 26 average 614, a 101 difference, and 174-176 average 516, a 98 difference.

Under the new system, seble says the same differences will be 39 and 154, assuming the 9 thru 11 hasn't changed, this is how that would like side by side:

rank          old          new
9-11          715          715
24-26        614          676
174-176  516          522

those changed appear substantial, and very much in the right direction - I don't understand what the problem is, to me, seble fixed the issue - what am I missing?  

Regarding D1 recruiting, Recruit Ranking Ave of Total Rating #1-#5 775 #9-11 715 #24-26 614 #48-50 581 #74-76 564 #98-100 555 #124-126 533 #149-151 525 #174-176 516 #198-200 508 The problem I see is there is the same difference #25 and #175, as #10 and #25. What is your view of this situation and how will the new recruits look? (reinsel - All-Star - 2:48 PM)

This change should address that issue. In the test world where I generated new recruits the difference between #9-11 and #24-26 is about 39 pts and the difference between #24-26 and #174-176 is 154 pts. Of course the rankings don't go strictly by overall rating, so there will be some variance to these numbers from world to world and season to season.

OR - i think what you might be missing is that there isn't a corresponding change to the maximum ratings. so before, at 24-26 position, if you had 614 players with an average of 100 points of improvement, then those players would turn into 714 players. now, it would be 676 players, turning into 714 players, instead of the intuitive, 776 players. i think the problem is not that freshman and sophs on mid majors aren't good enough - i argue that the problem is really with the juniors and seniors of those programs. if you are relying on freshman, you are dead anyway. so, mid majors will have better freshman, a little bit better sophs, and roughly the same juniors and seniors as before. i don't see how that fixes much of anything?

You can't have elite players and a ton of stud mid major type players and still have the elites players perform at elite levels in the sim... there has to be some sort of gap to make it happen. Otherwise it will be like before where all teams have strong 90+ core rosters and elites just look good on paper.

Moy, no one is saying there should be a ton of stud mid major players, only that they should be better than they are now. We're talking about a happy medium here, and there is clearly room for it.

Posted by moy23 on 2/3/2011 6:27:00 PM (view original):

Posted by coach_billyg on 2/3/2011 6:14:00 PM (view original):

Posted by moy23 on 2/3/2011 6:09:00 PM (view original):
Personally I would have just steepened the recruit generation curve at the front end so there are way less elite recruits and kept the curve the same in the middle to back end. Thus less elites to go around and more players with larger core attribute and potential variance. I.e. teams would have to choose between an 85 lp/65 reb post player or a 65 lp/85 reb post player.... rather than having more 85/85 post players.

you would have done that when, to recruit generation before seble's new engine, or just now (just before his last change yesterday)?

Just now. I think there are too many elite players as it is right now (just check out my u of I team - 9 points away from 11 players over 800 with some 10 games to go). Edit.... I don't even use fss and pretty much put 10 practice mins on each category so its not like I'm some practice minutes guru. I don't have to when I'm 'only' going after elite recruits. Bringing the # of elites down would force me to use fss like everyone else and to master practice mins.

well moy, i guess 6 months ago, when i was coaching in tark with 196 other human coaches, i would disagree that there are too many elite recruits. now, with 126 other human coaches, i tend to agree with you. it used to be roughly impossible for me at kentucky, surrounded by some great big 6 and acc teams, with a solid big 10, to recruit only elite players. just couldn't be done, at least not without taking at least 2 or 3 walk ons regularly, competition was just too hot. now, even though its still a pretty competitive area relative to the rest of tark IMO, that also means there are more elite recruits there, and with even a handful of quality opponents gone, it makes it a lot easier.

anyway, i think making an extremely limited number of elite players is in general, a bad idea. it essentially forces early entries to leave earlier. i think that would drastically diminish the value of elite players, while increasing the competition for them, and honestly they might not even be worth it anymore. if you want to basically get rid of good recruits all together, and have a steady curve the whole way like in d2 - i would be fine with that. but if you are going to have elite players, i want a smooth curve on down to the steady progression part, not for quality to fall off a cliff between elite and the steady progression. i just don't think the game would function right with say half as many elites, and the rest the same - the 90% of elites with the personality to leave early almost all would leave after freshman year, and the few that stuck around would dominate the game like never seen before.

*if* we had national recruiting, and elites with iq, essentially making them 1 and dones who actually could have a huge impact, then i would be totally on board with something like what you propose. i have no problem with almost all of the recruits being "mediocre" by todays standards, with only like 10 or 20 elite recruits who leave after 1 year anyway. that would allow mid majors to compete with bcs schools, while preserving the advantage of the elite school to enough of a degree. i just want some balance, i don't particularly care about the form it comes in.

gill - I simply didn't get the impression that there was any overwhelming function.  Obviously a mean and variance can be computed off a function, but they can exist statically ie. start at the expectation value and apply the variance, so the variance itself would be the function involved.  Your assertion that we would have seen the results in D3 is, as I'm sure you know, a bit ridiculous.  Sure, there are probably some outliers in D3, but they're going to be the guys who are now virtually useless.  Who's going to look at them?  Honestly, would any human coach have noticed if the bottom D3 players had gotten worse than they were before the update?  I actually did notice that right before the update I was able to recruit some players that were quite useful by their junior and senior seasons out of the initial D3 pool and haven't been able to do so since the update, so I would argue that the entire initial D3 pool may have gone downhill a bit.  Since most of the better coaches barely even look at the players who start in D3, much less the bottom end of that grouping, there could easily have been a change to the bottom of D3 that nobody noticed.  Say those 10 guys above 800 are balanced by 10 guys starting below 340.  Is anybody paying attention?  Doubtful.

The answer to the first question in the dev chat - your question - strongly implies to me that recruits are generated rating by rating by starting with a mean value for the rating and applying a standard deviation, presumably using some input from a RNG to determine the impact of the normal curve.  The elite recruits are just the ones that get lucky in a lot of categories...

Posted by dahsdebater on 2/3/2011 6:58:00 PM (view original):
The answer to the first question in the dev chat - your question - strongly implies to me that recruits are generated rating by rating by starting with a mean value for the rating and applying a standard deviation, presumably using some input from a RNG to determine the impact of the normal curve.  The elite recruits are just the ones that get lucky in a lot of categories...

what you are describing sounds to me like something that would yield a bell curve... i just don't think that is the case. what we have really is not very bell curvey at all is it?

Posted by dahsdebater on 2/3/2011 6:50:00 PM (view original):
gill - I simply didn't get the impression that there was any overwhelming function.  Obviously a mean and variance can be computed off a function, but they can exist statically ie. start at the expectation value and apply the variance, so the variance itself would be the function involved.  Your assertion that we would have seen the results in D3 is, as I'm sure you know, a bit ridiculous.  Sure, there are probably some outliers in D3, but they're going to be the guys who are now virtually useless.  Who's going to look at them?  Honestly, would any human coach have noticed if the bottom D3 players had gotten worse than they were before the update?  I actually did notice that right before the update I was able to recruit some players that were quite useful by their junior and senior seasons out of the initial D3 pool and haven't been able to do so since the update, so I would argue that the entire initial D3 pool may have gone downhill a bit.  Since most of the better coaches barely even look at the players who start in D3, much less the bottom end of that grouping, there could easily have been a change to the bottom of D3 that nobody noticed.  Say those 10 guys above 800 are balanced by 10 guys starting below 340.  Is anybody paying attention?  Doubtful.

im not sure what exactly you mean by apply the variance... that doesn't really make a whole lot of sense. variance describes the deviation from mean - but it doesn't tell you anything about the distribution of those points that deviate from the mean. and its not something you can really apply.

anyway, are you basically saying, start at 50, and then randomly add or subtract 1-50? using a RNG? that is kind of what it sounds like but its hard to tell. and then repeat for each category - thus the elites are the guys who got lucky in a lot of RNG outcomes?

Posted by coach_billyg on 2/3/2011 6:47:00 PM (view original):

Posted by moy23 on 2/3/2011 6:27:00 PM (view original):

Posted by coach_billyg on 2/3/2011 6:14:00 PM (view original):

Posted by moy23 on 2/3/2011 6:09:00 PM (view original):
Personally I would have just steepened the recruit generation curve at the front end so there are way less elite recruits and kept the curve the same in the middle to back end. Thus less elites to go around and more players with larger core attribute and potential variance. I.e. teams would have to choose between an 85 lp/65 reb post player or a 65 lp/85 reb post player.... rather than having more 85/85 post players.

you would have done that when, to recruit generation before seble's new engine, or just now (just before his last change yesterday)?

Just now. I think there are too many elite players as it is right now (just check out my u of I team - 9 points away from 11 players over 800 with some 10 games to go). Edit.... I don't even use fss and pretty much put 10 practice mins on each category so its not like I'm some practice minutes guru. I don't have to when I'm 'only' going after elite recruits. Bringing the # of elites down would force me to use fss like everyone else and to master practice mins.

well moy, i guess 6 months ago, when i was coaching in tark with 196 other human coaches, i would disagree that there are too many elite recruits. now, with 126 other human coaches, i tend to agree with you. it used to be roughly impossible for me at kentucky, surrounded by some great big 6 and acc teams, with a solid big 10, to recruit only elite players. just couldn't be done, at least not without taking at least 2 or 3 walk ons regularly, competition was just too hot. now, even though its still a pretty competitive area relative to the rest of tark IMO, that also means there are more elite recruits there, and with even a handful of quality opponents gone, it makes it a lot easier.

anyway, i think making an extremely limited number of elite players is in general, a bad idea. it essentially forces early entries to leave earlier. i think that would drastically diminish the value of elite players, while increasing the competition for them, and honestly they might not even be worth it anymore. if you want to basically get rid of good recruits all together, and have a steady curve the whole way like in d2 - i would be fine with that. but if you are going to have elite players, i want a smooth curve on down to the steady progression part, not for quality to fall off a cliff between elite and the steady progression. i just don't think the game would function right with say half as many elites, and the rest the same - the 90% of elites with the personality to leave early almost all would leave after freshman year, and the few that stuck around would dominate the game like never seen before.

*if* we had national recruiting, and elites with iq, essentially making them 1 and dones who actually could have a huge impact, then i would be totally on board with something like what you propose. i have no problem with almost all of the recruits being "mediocre" by todays standards, with only like 10 or 20 elite recruits who leave after 1 year anyway. that would allow mid majors to compete with bcs schools, while preserving the advantage of the elite school to enough of a degree. i just want some balance, i don't particularly care about the form it comes in.

If I recall... part of sebles grand vision was that 'elites' would not go nutty for ee. An example would be john Keller from my roster - 950 rated Jr should have already gone especially after 2 elite 8s.

dahs, i guess what i am getting at is, straight randomization off starting values could not yaild recruits like we see. there has to be some frequency or something in the randomization to get the kind of curve we have... very sharp at the top, then very even the rest of the way.

I guess I disagree with your general premise that it's "very even the rest of the way."  Given that it is now abundantly clear that players are actually initially generated by their cap values the only way to even come close to making a good estimate of the shape of the curve that would be generated by overall recruit ratings would require you to FSS every state and would still require a great deal of guesstimation.  But it seems to me that the really good D3 teams could beat some of the sim D1 teams even in the absence of large numbers of walk-ons, and that most of D2 falls in that same range.  I think a huge proportion of the overall player pool will finish with an overall rating between 675 and 750 or 775, and it flattens out considerably from there.  Yes there seem to be enough seemingly legitimate D1 players for the top D1 schools, but there are significantly less Big  6 D1 schools than D2 schools and it's well publicized that mid-majors and sometimes even Big 6 programs are dipping down to take recruits from D2 schools.  If in fact there was a very flat curve such as what you described, I don't think we'd see nearly as much of this sort of thing happening.  But that's entirely conjecture since a complete study would be virtually impossible.

Posted by jjboogie on 2/3/2011 5:29:00 PM (view original):
wronoj - seble and i are talking in some tickets so we have his attention at least and i know some pretty well respected coaches echo our sentiments.  don't give up yet! :)

glad to hear it. it really has worried me... and his responses in the dev chat certainly didn't make me feel any better.