SRP
Postdoc
Posts: 3525
Reg: 02-04-06
|
10-15-17 01:49 PM - Post#234006
In response to Silver Maple
Aside from fan partisanship, I think part of the issue here is that people are over-interpreting what Mike is saying. 6/7 of his ten-win-share elite were not "consensus three-star." So it's going to be very easy to think of examples of successful lower-rated recruits.
I also agree with skeptics that fit, player development, and synergies/anti-synergies with other players may create systematic error that could be modeled in some way. But without such a model I agree with Mike that the sensible predictive thing to do is to treat it as unsystematic error, already incorporated in the odds.
The relevant IL question to me is to what extent the marginal impact of high-level recruits declines for a single team [cough, Harvard, cough]. My intuition is that it eventually declines because of PT issues if nothing else. But I also think that the fact that three-stars are still pretty uncertain makes the marginal value of each one stay pretty high. Not sure exactly how I'd approach modeling that stuff, although brute-force Monte Carlo simulations would probably work.
|
SomeGuy
Postdoc
Posts: 4617
Reg: 11-22-04
|
10-15-17 03:38 PM - Post#234007
In response to Silver Maple
Part of the question on the role player characterization for Wang is where that information is coming from. Is it an assistant at another Ivy who just wasn't sold on Wang? Is it an outsider's impression that he might not be a system fit (this seemed to be part of why Mike thought Faulds was overrated)? Is it coming from Penn?
It seems to me that Mike does a good job sifting through this stuff (just guessing, he probably has multiple sources, and the overall crowd impression trumps the various reasons why some of the possibilities I list above might be suspect).
|
SomeGuy
Postdoc
Posts: 4617
Reg: 11-22-04
|
10-15-17 03:49 PM - Post#234008
In response to AsiaSunset
Forbes will be interesting, as wil Djuricic and Kirkwood. These are players where I don't think the ratings support Mike's perception that the Harvard guys are better. My guess is that he will turn out to be right, but we don't know that for sure. They will be good tests of whatever other stuff besides the simple ratings gets baked in when Mike tries to differentiate guys.
Mike, correct me if I am wrong about this. I see that you have Kirkwood ahead of Freedman, for example, yet Freedman seems to be in the consensus 3 star bucket while Kirkwood is not.
|
PennFan10
PhD Student
Posts: 1825
Reg: 02-15-15
|
10-15-17 04:00 PM - Post#234010
In response to SomeGuy
I’m pretty sure the Penn coaches are high on Wang and did not recruit him as a role player for this team.
|
SomeGuy
Postdoc
Posts: 4617
Reg: 11-22-04
|
10-15-17 04:46 PM - Post#234011
In response to PennFan10
That is my impression as well.
Could be a role player as a freshman of course -- seems like his skill set would give us an interesting 4th big option with AJ, Max, and Simmons.
|
mrjames
Professor
Posts: 5240
Loc: Montclair, NJ
Reg: 11-21-04
|
10-15-17 09:30 PM - Post#234025
In response to SomeGuy
On the women's model point - I'm not that educated about women's recruiting ranking systems and sites to pull advanced stats from, but as long as both exist, one could compile the data and test to see if there were relationships. I think ESPNW runs a system, right? Are there others?
As for Danilo and Noah vs Spencer, Canadians are a huge blind spot for the ratings systems, unless they're Top 50 kids or move to the US to play. Some of that is playing in Canada vs the US, but that's leveled a bit by AAU which has Canadian teams. A lot of it is the reclassing of kids due to Canada's schooling structure, which essentially gives kids the choice of reclassing forward or not. Danilo was a highly-rated recruit at ESPN, then reclassed from 2018 to 2017 in the fall after all the summer re-grading and was just given an NR. Noah averaged 16ppg on 54% shooting at Adidas Uprising and then 13ppg on 59% shooting at Adidas Nations (where Llewellyn lit the world on fire with 22ppg on 52% shooting). He will get a lot of publicity now at NMH, but I'm not sure anyone would know his high school exploits in Canada for anything.
Danilo was the most hotly pursued recruit in the 2017 class, though Much would have been. The 2017 class wasn't very good or deep though, so that doesn't necessarily mean he (or Much) are going to follow the same path as the bests from the loaded 2016 class.
Spencer's offensive value is clear. He can run an offense in a way that Harvard hasn't seen with any point guard it has had thus far. Can he play D1 defense? We'll find out. For Noah, there's not a lot to find out. If he could shoot, he'd be a Top 50 kid in the class and would be at a Duke, Kentucky, Kansas, etc. He doesn't have a remotely reliable jump shot. But otherwise, he's really tough to keep from producing and he can set up others (like a Wes Saunders). He's also a plus defender.
|
dperry
PhD Student
Posts: 1038
Loc: Oreland, PA
Reg: 11-24-04
|
10-18-17 11:17 PM - Post#234173
In response to JDP
Mike, how adaptable is your model to the women's game? I assume you would have to preform some recalibration (the recruit rankings may not be as robust) relative to the men's model? Does the data exist to undertake the task? does the will exist?
A couple of people have asked about women's recruiting lately; being the resident expert by default, I'm going to try to make some tine this weekend to talk about recruiting in general and how well I think the league did this year. In response to your particular question, however, the short answer is: not at this time. The only place you can find the more sophisticated stats for women right now is NatStat WBB, formerly WBBState. Since it costs $40 just for access to the current season, I'm not going to evaluate their quality in person (although when it drops to $20 on January 22, I might be tempted), but FiveThirtyEight thinks enough of them to have lamented their absence when their servers crashed right before the national tournament two years ago. However, their archive only goes back five years (and that will set you back $200.) I'll discuss the specific recruiting sites in my bigger post, but there are four good ones that I know of. The overall quality is inferior to that of the men's sites even for the power conferences, and it drops off precipitously when you get to the Ivies' level. At best, they are wildly inconsistent in their reporting; you have to look at all of them to get even some idea of what a team is getting, and they never seem to get entire classes unless they're tiny. Right now, you're lucky to get anything at all for the teams other than the P's and Harvard. Furthermore, they don't use the same ranking scales; how do you sum up a prospect who gets 3* from one site, is ranked 95th overall in the country by another, and 10th at her position by a 3rd? Lastly, they don't seem to preserve historical data for very long (unless it's behind the paywall); were one to do a project going forward, they would probably have to save all the evaluations themselves. In short, there's not enough data, and the will to collect it seems unlikely to manifest itself anytime soon. It certainly won't come from me; my number crunching interests run more towards team records over the years and other historical things.
David Perry
Penn '92
"Hail, Alma Mater/Thy sons cheer thee now
To thee, Pennsylvania/All rivals must bow!!!" |
|
|
