mrjames
Professor
Posts: 6062
Loc: Montclair, NJ
Reg: 11-21-04
|
Re: For Mike James--algorithm changes? 03-19-18 09:38 AM - Post#253769
In response to T.P.F.K.A.D.W.
It is indeed just that. I blend all of the ratings into a composite rating and then dice those features in slightly different ways to predict overall class win shares.
What I'd love to add are offers, which would add predictiveness (there are two reasons I get bullish or bearish on a player relative to the "stars" they get - legit offers and scouts from those that would know).
But the ratings are actually decently powerful to predict the overall performance of a class, because the errors at the individual level tend to cancel out a bit when aggregated, leaving some useful signal.
|
mrjames
Professor
Posts: 6062
Loc: Montclair, NJ
Reg: 11-21-04
|
Re: For Mike James--algorithm changes? 03-19-18 09:48 AM - Post#253771
In response to palestra38
I'd advise not using Assist to Turnover ratio, but rather to use the rates of each (assists per 100 made buckets and turnovers per 100 possessions).
Harvard's turnover rate in Ivy play was dead last in the Ivies at 21.1% and the Crimson was the only team above 20%. Its live-ball TO rate was up over 10% in conference play - also dead last.
Harvard's assist rate rose slightly in Ivy play, but probably actually fell when you account for the fact that most 3PT buckets are assisted, and Harvard was randomly cold in the non-conf driving down its assist rate for no good reason.
In Bryce Aiken's time on the floor in the non-conference, he had a 25.4% Assist Rate and an 18.5% Turnover Rate. His replacements in Ivy play were 20.9% Assist Rate and 24.9% Turnover Rate as well as 12.1% Assist Rate and 25.8% Turnover Rate.
Harvard also doesn't play an isolation offense. It plays a motion offense with an end-of-shot-clock high PNR. That it became more of an iso offense was due to personnel and having two players that are tremendous isolation players (Towns and Lewis). But both of those players could be effective with playmakers around them stretching the D. I suspect with more legitimate offensive threats next year, Harvard will be prepared to execute the motion offense it a way it hasn't since 2013-14.
|
SomeGuy
Professor
Posts: 6391
Reg: 11-22-04
|
Re: For Mike James--algorithm changes? 03-19-18 09:50 AM - Post#253773
In response to palestra38
Yes, one question was whether Aiken would play more like a point with Chambers gone. He turned out to be pretty similar to what he did with Chambers, so Harvard basically lost a guy with a great A/TO ratio and didn’t really replace him. McCarthy could have answered that some.
Whether Kirkwood and Freedman can answer it immediately is the main question— otherwise i’m Not sure Harvard has the answer.
|
Jeff2sf
Postdoc
Posts: 4466
Reg: 11-22-04
|
next year, the deepest of deep dives 03-19-18 10:10 AM - Post#253776
In response to SomeGuy
About this rapid improvement gene that Steve suddenly developed. Wouldn't we have said essentially the same thing about Jerome being able to improve guys after the 2nd place finish?
I see improvement for sure with lots of players, but generally speaking that happens in college, no? I don't know someone improves like Darnell, but I don't think even Steve puts that in his ledger. Darnell was incredible.
Brodeur improved on one thing, passing out of double teams. But that seems within the range of typical frosh to soph improvement.
Woods improved by playing a completely different role than he played before. Cool
Wood improved by playing a completely different role than he played before. Cool*
Betley I dunno, did he improve? If so, by more than a typical frosh to soph?
Rothschild improved and showed the power of body transformation. Cool.
Goodman did not improve, if anything looked like he regressed before snapping out of it.
Silpe. No.
Simmons seemed to regress.
Donahue, definite regression from freshman year.
I mean, maybe. I don't see this particular "skill" as something to bank on from Steve. Want to make sure he's bringing in the talent.
* but given just how awful Caleb was in that role, one has to wonder what Steve saw in practice that made him ever think he was suitable to be the starting pg.
|
mrjames
Professor
Posts: 6062
Loc: Montclair, NJ
Reg: 11-21-04
|
Re: For Mike James--algorithm changes? 03-19-18 10:12 AM - Post#253777
In response to SomeGuy
Umm... Bryce Aiken's assist rate this year was 25.4%. That would have been 2nd in the league this year and was basically even with Siyani's sophomore and junior year numbers.
|
palestra38
Professor
Posts: 32685
Reg: 11-21-04
|
Re: next year, the deepest of deep dives 03-19-18 10:25 AM - Post#253779
In response to Jeff2sf
Just about everything you mention is coaching--changing roles, learning to pass out of a double, even body transformation, as is sitting on the bench because you cannot do the things necessary to win.
And Jerome had his one good year because of 3 players he inherited that for a short time were healthy together. His own recruits --Tony Hicks being the poster boy--did not show that they were coached up.
|
Jeff2sf
Postdoc
Posts: 4466
Reg: 11-22-04
|
03-19-18 10:29 AM - Post#253780
In response to palestra38
His own recruits very definitely did not show they were coached up. No argument. The argument is with what happens if you just look at a point in time.
Changing roles is coaching, absolutely. But kids, very definitely, do improve from freshman to senior year at most colleges. Some kids don't. Most kids do. So what is it about Steve that shows he does it more than other coaches. Cus I don't think there's documented evidence of this.
|
10Q
Professor
Posts: 23199
Loc: Suburban Philly
Reg: 11-21-04
|
next year, the deepest of deep dives 03-19-18 10:33 AM - Post#253781
In response to Jeff2sf
This is Q signing off till next season. I just can't follow these hypotheticals. Enjoy.
I'm predicting another Ivy title next year, with an 11 seed and a first round win.
|
PennFan10
Postdoc
Posts: 3580
Reg: 02-15-15
|
03-19-18 10:34 AM - Post#253782
In response to mrjames
AJ got a lot better passing, 3pt shooting (though not good enough to be taking very many) and using his left hand 9particularly late in the season). That's a lot more than just passing out of double teams. A big part of the "improvement" of players is 1-due to assistant coaches working with these guys, and 2-players eliminating things that are negative (AW: turnovers and bad 2pt shots, Caleb: turnovers and porous D, Darnell: too many to list, Max:Defense, assists, etc).
As far are Harvard's "suffering" from bad recruiting outcomes, Penn didn't have a single player in those classes ranked ahead of those Harvard recruits did they? Maybe Matt Howard? So doesn't that help with the idea that recruiting rankings are limited in the predictive capabilities? Penn has gone 4-3 over the past 3 years vs Harvard while using players in the same classes as Harvard. This year we used 3 sophomores a junior and 2 seniors to go 2-1. And none of those players are ranked higher than any of Harvard's starters or some of their bench players.
I am in total agreement that better players win games and Harvard has probably 3 of the 5 best players in the league on their roster for next year, maybe even 5 of the top 8 so I get why they will be highly rated. But the under performance issues over the past 3 years are more than just lack of recruits panning out in my view.
|
Jeff2sf
Postdoc
Posts: 4466
Reg: 11-22-04
|
03-19-18 10:40 AM - Post#253784
In response to PennFan10
1. I'm wincing hard at the idea of you using record against as opposed to KenPom for the last 3 years.
2. Mike keeps saying it wasn't an underperformance, that the classes truly weren't very good and you keep ignoring that.
Switching topics, great, AJ improved at other things. Freshmen not coached by Steve Donahue do improve right? I mean, that's a thing. Jerome's an outlier in how bad he was, no?
|
palestra38
Professor
Posts: 32685
Reg: 11-21-04
|
Re: next year, the deepest of deep dives 03-19-18 10:41 AM - Post#253785
In response to 10Q
You mean no silly joke posts? You won't be able to resist.
|
palestra38
Professor
Posts: 32685
Reg: 11-21-04
|
03-19-18 10:47 AM - Post#253786
In response to Jeff2sf
Obviously, after having his first winning season, it is possible to claim that Steve Donahue will not be able to sustain the level of improvement shown from Year 1 to Year 2 to Year 3. And of course, talent means a lot (which was the point of my entire Antonio Woods argument with you last year). But the ball movement improvement was not limited to the guards---everyone moved the ball to the open man this year. And if we had slightly better shooters (and much better FT shooters), we would not have lost many games. So I don't think that it's a stretch to say that the vastly improved ball movement (it's fading from memory how poorly the Jerome teams moved the ball, relying on bad shots at the shot clock buzzer all the time) will survive the Foreman graduation. But correct me if I am wrong---don't we have 3 fairly highly touted sophs, none of whom played (other than garbage minutes) in the Ivy season? Won't our talent level be higher next year, even if we only have one touted frosh?
|
mrjames
Professor
Posts: 6062
Loc: Montclair, NJ
Reg: 11-21-04
|
03-19-18 10:53 AM - Post#253787
In response to PennFan10
Here are the win share predictions for those Harvard and Penn classes:
Harvard 2013 16.5 Win Shares
Penn 2013 18.6 Win Shares (Best that year was Brown at 21.5 Win Shares)
Harvard 2014 11.7 Win Shares
Penn 2014 16.5 Win Shares (Best that year was Princeton at 18.6 Win Shares)
Harvard 2015 19.5 Win Shares
Penn 2015 16.5 Win Shares (Best that year was Cornell at 24.3 Win Shares)
Harvard 2016 54.3 Win Shares
Penn 2016 18.6 Win Shares (Best that year was Harvard)
Harvard 2017 10.8 Win Shares
Penn 2017 14.7 Win Shares (Best that year was Columbia at 22.5 Win Shares)
Now, Harvard's 2013-2015 classes happened to also underperform those targets (to some extent because they got bowled over by the 2016 class), but expectations-wise over those years they were a step behind Penn to begin with.
Preliminarily, Harvard's 2018 class is in the 30 Win Share range, though this gets run off the final, so that could change.
|
10Q
Professor
Posts: 23199
Loc: Suburban Philly
Reg: 11-21-04
|
Re: next year, the deepest of deep dives 03-19-18 10:56 AM - Post#253788
In response to palestra38
All rights reserved
|
Jeff2sf
Postdoc
Posts: 4466
Reg: 11-22-04
|
next year, the deepest of deep dives 03-19-18 11:00 AM - Post#253789
In response to palestra38
The point of the Antonio Woods debate was talent? The guy is the 5th best player on a KenPom 130 team. Congratulations? "special" snort.
Jarrod, whom I so very much want to be good, was touted but was pushed to the bench in favor of players like Jake Silpe and Matt Macdonald. This was very definitely not a good freshman year for him. This is in no way a point in his favor.
Eddie got hurt. He had a wonderful game against Monmouth, a perfectly nice game against a high school/d3 team and was pretty lousy in the rest of the games (no seriously, go check). He was not touted to begin with. I sure hope we see more Monmouth, but if he shoots a still good but more normal 6 for 8, are we saying this was a good freshman year?
Jelani was touted. He also suffered a devastating knee injury. Many people come back from that. Some people don't. Some of the many who come back don't improve until their 2nd year playing. I hope we get two of Wang/Jelani/Jarrod/Eddie to make an impact and play in the Ivy Tournament. That would be a good thing. Yet it would be virtually impossible to argue that their impact will exceed Foreman and Wood's contributions next year. I think the improvement we can see from the other returnees will allow for an improvement. So I'd guess we'll be net positive as a whole, but not owing to the freshmen (edit - freshmen plus Jarrod who's a sophomore and coming off a not great freshman year) replacing the seniors.
Edited by Jeff2sf on 03-19-18 11:01 AM. Reason for edit: No reason given.
|
palestra38
Professor
Posts: 32685
Reg: 11-21-04
|
Re: next year, the deepest of deep dives 03-19-18 11:04 AM - Post#253790
In response to Jeff2sf
Reminder to Jeff---KenPom is not a measure of "talent".
And second reminder--You thought we would not be able to replace Howard.
|
Jeff2sf
Postdoc
Posts: 4466
Reg: 11-22-04
|
03-19-18 11:16 AM - Post#253791
In response to palestra38
I predicted we'd be better this year than last year. So obviously I thought we'd improve. I just resented people not recognizing Matt's great talent and contribution.
|
PennFan10
Postdoc
Posts: 3580
Reg: 02-15-15
|
03-19-18 11:30 AM - Post#253792
In response to Jeff2sf
1. I'm wincing hard at the idea of you using record against as opposed to KenPom for the last 3 years.
2. Mike keeps saying it wasn't an underperformance, that the classes truly weren't very good and you keep ignoring that.
Switching topics, great, AJ improved at other things. Freshmen not coached by Steve Donahue do improve right? I mean, that's a thing. Jerome's an outlier in how bad he was, no?
Ok. Let's be more specific. I think we are more in agreement than you think.
1-We had worse KP over 3 years, and still beat them (split 2 years won this year). This year they came in as KP 110 and went to KP 190 (finished at 141) while at full strength. Isn't that under performance?
2-Their classes weren't any good, ours were worse, we still beat them, why?
3-Not sure your point on the freshman not coached by Donahue? Can you rephrase?
|
PennFan10
Postdoc
Posts: 3580
Reg: 02-15-15
|
Re: next year, the deepest of deep dives 03-19-18 11:33 AM - Post#253793
In response to Jeff2sf
The point of the Antonio Woods debate was talent? The guy is the 5th best player on a KenPom 130 team. Congratulations? "special" snort.
Jarrod, whom I so very much want to be good, was touted but was pushed to the bench in favor of players like Jake Silpe and Matt Macdonald. This was very definitely not a good freshman year for him. This is in no way a point in his favor.
Eddie got hurt. He had a wonderful game against Monmouth, a perfectly nice game against a high school/d3 team and was pretty lousy in the rest of the games (no seriously, go check). He was not touted to begin with. I sure hope we see more Monmouth, but if he shoots a still good but more normal 6 for 8, are we saying this was a good freshman year?
Jelani was touted. He also suffered a devastating knee injury. Many people come back from that. Some people don't. Some of the many who come back don't improve until their 2nd year playing. I hope we get two of Wang/Jelani/Jarrod/Eddie to make an impact and play in the Ivy Tournament. That would be a good thing. Yet it would be virtually impossible to argue that their impact will exceed Foreman and Wood's contributions next year. I think the improvement we can see from the other returnees will allow for an improvement. So I'd guess we'll be net positive as a whole, but not owing to the freshmen (edit - freshmen plus Jarrod who's a sophomore and coming off a not great freshman year) replacing the seniors.
I think all of our predictions can be considered "accurate" as well as "way off" based on the data. If we are honest, there isn't a single one of us who would have said we would be KP 120 without any meaningful production from our freshmen, led by defense, and with Darnell as our best guard. I certainly didn't think so.
|
Jeff2sf
Postdoc
Posts: 4466
Reg: 11-22-04
|
03-19-18 11:45 AM - Post#253794
In response to PennFan10
1. I'm wincing hard at the idea of you using record against as opposed to KenPom for the last 3 years.
2. Mike keeps saying it wasn't an underperformance, that the classes truly weren't very good and you keep ignoring that.
Switching topics, great, AJ improved at other things. Freshmen not coached by Steve Donahue do improve right? I mean, that's a thing. Jerome's an outlier in how bad he was, no?
Ok. Let's be more specific. I think we are more in agreement than you think.
1-We had worse KP over 3 years, and still beat them (split 2 years won this year). This year they came in as KP 110 and went to KP 190 (finished at 141) while at full strength. Isn't that under performance?
2-Their classes weren't any good, ours were worse, we still beat them, why?
3-Not sure your point on the freshman not coached by Donahue? Can you rephrase?
You may be right about us being closer to agreement but...
1. They weren't at full strength, they lost their good point guard and got stuck with a not good point guard (even if you like Juzang generally, he may not fit as a point). At no other position do I fear a drop off between starting player and bench player. I expect us to be better next year than we were this year. If we are not, I expect it's because we couldn't replace Darnell's ball handling and steadiness. I would say if there were two players like AJ, one a big man and one a point guard, I'd much prefer to lose the big man version than the point guard.
2. Were our classes worse? If I'm reading Mike's 2013/2014/2015 class data right, in two of the three years we were predicted to have a better class.
3. Rephrasing: Steve and his staff certainly deserve credit in helping many of our players improve (though please note a few did not improve). But I think the default expectation at any school is that coaches are going to help players improve. Because players generally improve between freshmen and senior year. I think we are giving Steve more credit than is supported because we saw a coach like Jerome not improve the players he recruited. My position is that Jerome was especially bad at that and a true outlier.
I'm further saying that it's possible Steve is better than the average coach at coaxing improvement out of players. I don't think the evidence is there for that across his coaching career. I suspect he's typical.
|