2021 Offseason Transfers

[Acymetric]

It is possible I've lost track of what the debate is even about...does anyone have a link to Kedsy's original post on the subject? I think maybe we're trying to answer different questions. It still feels (to me) like the point being made is basically just a tautology, but perhaps with more context and the original post/data it would make more sense to me.

Edit: Found Kedsy's post, no need for anyone else to go looking

Ok, one thought I have that would (I think?) solve the issue of semi-arbitrary cut-offs when analyzing transfers by time-period: chart it as a 4 year rolling average.

I maintain that you can't use the number of recruits to normalize the data across time, because the number of recruits and the number of transfers both impact each other (Have we ever gone out and recruited or reclassified a player last minute after a surprise loss? If yes, you can't use that additional recruit to normalize attrition because the additional recruit was the result​ of the attrition).

[CDu]

Ok, one thought I have that would (I think?) solve the issue of semi-arbitrary cut-offs when analyzing transfers by time-period: chart it as a 4 year rolling average.

I maintain that you can't use the number of recruits to normalize the data across time, because the number of recruits and the number of transfers both impact each other (Have we ever gone out and recruited or reclassified a player last minute after a surprise loss? If yes, you can't use that additional recruit to normalize attrition because the additional recruit was the result​ of the attrition).

I'll say again though: we haven't had any meaningful difference in the total number of transfers. The differences in the number of players we've rostered is almost entirely driven by the NBA early entry.

Our transfer totals in those 11 year buckets is 6, 7, and 9. But our total players jumped from 32 in the first 11 year range (when all but 3 players were 4 year guys) to 38 in the second range (when only about 5 or 6 guys went pro) to 52 in the most recent 11 years (when over 20 went pro early).

So it is totally reasonable to use the total number of players we've rostered over those spans, because it isn't the transfers that is driving the difference in the number of players.

[CrazyNotCrazie]

Can we please split off into two threads - one on 2021 Offseason Transfers and another on Trends in Transfers? I keep coming to this thread hoping for news on transfers and instead it seems kind of bogged down in hair splitting about whether or not Duke has a lot of transfers vis a vis other schools. I have my thoughts on that (there is not statistically significant data given how small basketball rosters are and there have been too many major paradigm shifts in the last 10-15 years to draw an educated conclusion). So I don't personally think it is really worth discussing. But just because I don't think it is an issue doesn't mean that others aren't allowed to discuss. I just wish you would do so elsewhere. Thanks.

[Acymetric]

I'll say again though: we haven't had any meaningful difference in the total number of transfers. The differences in the number of players we've rostered is almost entirely driven by the NBA early entry.

Our transfer totals in those 11 year buckets is 6, 7, and 9. But our total players jumped from 32 in the first 11 year range (when all but 3 players were 4 year guys) to 38 in the second range (when only about 5 or 6 guys went pro) to 52 in the most recent 11 years (when over 20 went pro early).

So it is totally reasonable to use the total number of players we've rostered over those spans, because it isn't the transfers that is driving the difference in the number of players.

So, I'm not really arguing with your conclusion. It doesn't appear that we have seen an increase in transfers, I agree with that (our transfers have always been unevenly distributed which messes with people's intuitions about what it "feels" like). My argument is that the variables in "transfers/rostered players" are too confounded (in the statistical sense) to be meaningful when used together that way, and the fact that the result happens to appear to support your position doesn't tell us much (and as you noted at one point, the sample size is small enough that you really can't determine anything conclusively in either direction regardless).

[CDu]

So, I'm not really arguing with your conclusion. It doesn't appear that we have seen an increase in transfers, I agree with that (our transfers have always been unevenly distributed which messes with people's intuitions about what it "feels" like). My argument is that the variables in "transfers/rostered players" are too confounded (in the statistical sense) to be meaningful when used together that way, and the fact that the result happens to appear to support your position doesn't tell us much (and as you noted at one point, the sample size is small enough that you really can't determine anything conclusively in either direction regardless).

I think the confounding issue is not a big deal in this case. It WOULD be a big deal if there was a lot of movement in both numerator and denominator. But in this case, the movement is all in the denominator, and all due to another factor. So for the purposes of the question asked, I think Kedsy's analysis is appropriate.

The odds of any Duke player transferring from Duke hasn't increased over the last 33 years. We've had some up and down blips at various points, but it's held pretty steady overall.

[Acymetric]

I think the confounding issue is not a big deal in this case. It WOULD be a big deal if there was a lot of movement in both numerator and denominator. But in this case, the movement is all in the denominator, and all due to another factor. So for the purposes of the question asked, I think Kedsy's analysis is appropriate.

The odds of any Duke player transferring from Duke hasn't increased over the last 33 years. We've had some up and down blips at various points, but it's held pretty steady overall.

Again, agree with the conclusion, but the confounding issue and the sample size issue make the analysis basically useless. I am all for browbeating people with stats when they're wrong, but in this case people are doing it with bad stats (we don't have enough data for good stats here) that can't actually reliably interpreted to mean anything because they happen to support the desired conclusion. The analysis also smuggles part of the conclusion ("all due to another factor") into the opening assumptions (and also treats transfer/go pro in a binary way that doesn't seem quite right...some cases of "going pro" look more like transferring than others*). The argument is stronger just looking at the 11 year buckets, if anything.

*Consider Zion or Tatum compared to Trevon Duval. All three were early entrants, but if we're trying to analyze trends I'm not sure it makes sense to treat them the same

[CDu]

Again, agree with the conclusion, but the confounding issue and the sample size issue make the analysis basically useless... The argument is stronger just looking at the 11 year buckets, if anything.

I disagree, especially with the last sentence. Because that is ignoring the huge increase in NBA early entry altogether. That's a bigger mistake than ignoring the impact on denominator that one or two transfers (or one* specious argument about a player who chose to go pro).

Another way to look at it would be to look at the odds ratio of transferring versus not transferring. That completely eliminates the denominator problem, as it essentially compares those who transferred with those who didn't transfer.

So we get 6/26 (OR = 0.231) for the first 11 years, 7/31 (OR = 0.226) in the second 11 years, and 9/43 (OR = 0.209) in the third 11 years.

*Consider Zion or Tatum compared to Trevon Duval. All three were early entrants, but if we're trying to analyze trends I'm not sure it makes sense to treat them the same

I'm not sure I agree with your logic here. Duval was a top-5 recruit. He was always going to be a one-and-done. He didn't transfer. And he played in the NBA as a rookie, albeit briefly. How is his decision notably different than, say, Vernon Carey?

By that same argument, Shav Randolph left early to a nonexistent NBA market. He didn't transfer either.

We don't need to look for ways to overcomplicate things. Either you transferred or you didn't, right? The "didn't" column includes anyone who didn't choose to spend one of their 4 years playing at another college/university. So the easiest way to tell if our transfer rate has increased is to look at the data and ask, "what is the probability that a player will transfer from Duke?" Or "what are the odds that a player will transfer from Duke?" And if that has held pretty consistent, you can say the transfer rate hasn't changed much. Which is the case here.

[Bay Area Duke Fan]

I'm not sure I agree with your logic here. Duval was a top-5 recruit. He was always going to be a one-and-done. He didn't transfer. And he played in the NBA as a rookie, albeit briefly. How is his decision notably different than, say, Vernon Carey?

By that same argument, Shav Randolph left early to a nonexistent NBA market. He didn't transfer either.

We don't need to look for ways to overcomplicate things. Either you transferred or you didn't, right? The "didn't" column includes anyone who didn't choose to spend one of their 4 years playing at another college/university. So the easiest way to tell if our transfer rate has increased is to look at the data and ask, "what is the probability that a player will transfer from Duke?" Or "what are the odds that a player will transfer from Duke?" And if that has held pretty consistent, you can say the transfer rate hasn't changed much. Which is the case here.

Big difference. Shav unpacked his bags and played at Duke for three seasons.

[CDu]

Big difference. Shav unpacked his bags and played at Duke for three seasons.

While true, it is not meaningful in the context of this discussion.

[sagegrouse]

So, I'm not really arguing with your conclusion. It doesn't appear that we have seen an increase in transfers, I agree with that (our transfers have always been unevenly distributed which messes with people's intuitions about what it "feels" like). My argument is that the variables in "transfers/rostered players" are too confounded (in the statistical sense) to be meaningful when used together that way, and the fact that the result happens to appear to support your position doesn't tell us much (and as you noted at one point, the sample size is small enough that you really can't determine anything conclusively in either direction regardless).

Again, agree with the conclusion, but the confounding issue and the sample size issue make the analysis basically useless. I am all for browbeating people with stats when they're wrong, but in this case people are doing it with bad stats (we don't have enough data for good stats here) that can't actually reliably interpreted to mean anything because they happen to support the desired conclusion. The analysis also smuggles part of the conclusion ("all due to another factor") into the opening assumptions (and also treats transfer/go pro in a binary way that doesn't seem quite right...some cases of "going pro" look more like transferring than others*). The argument is stronger just looking at the 11 year buckets, if anything.

*Consider Zion or Tatum compared to Trevon Duval. All three were early entrants, but if we're trying to analyze trends I'm not sure it makes sense to treat them the same

There is no "confounding issue" that makes the data on gains and losses of players useless. For example, look at the long-term averages of "recruited roster" size versus average number of recruits. It tells you how many years you are getting per recruit -- useful for some purposes (and the cost accountants would love it). The "feedback" that poses some statistical complexity doesn't seem very important to me -- feedback, meaning that the scholarship limits prevent teams from recruiting more players. Most teams have fewer "recruited players" than the scholarship limits, so that if an outstanding prospect or transfer presents himself, then he can be put on scholarship.

[Natty_B]

Rocket Watts in the portal and some dude from Kansas. It’s a staggering number of transfers.

[MChambers]

Rocket Watts in the portal and some dude from Kansas. It’s a staggering number of transfers.

Hope Watts changes his mind, because I really wanted to hear “Rocket Watts of the MSU Rocket Mortgage Spartans” next year.

[Acymetric]

I disagree, especially with the last sentence. Because that is ignoring the huge increase in NBA early entry altogether. That's a bigger mistake than ignoring the impact on denominator that one or two transfers (or one* specious argument about a player who chose to go pro).

Another way to look at it would be to look at the odds ratio of transferring versus not transferring. That completely eliminates the denominator problem, as it essentially compares those who transferred with those who didn't transfer.

So we get 6/26 (OR = 0.231) for the first 11 years, 7/31 (OR = 0.226) in the second 11 years, and 9/43 (OR = 0.209) in the third 11 years.

I can't quite figure out how you got from Kedsy's numbers to these numbers (obviously you're including/excluding some of the edge cases, but I can't figure out which ones to arrive at these numbers). Anyways, it doesn't really matter. I'm not going to convince you that these numbers are basically meaningless (again, I agree with the conclusion this is just a bad way to support it), and you aren't going to convince me that this is valuable (or even usable) statistical analysis. I suppose we can just drop it.

I'm not sure I agree with your logic here. Duval was a top-5 recruit. He was always going to be a one-and-done. He didn't transfer. And he played in the NBA as a rookie, albeit briefly. How is his decision notably different than, say, Vernon Carey?

By that same argument, Shav Randolph left early to a nonexistent NBA market. He didn't transfer either.

We don't need to look for ways to overcomplicate things. Either you transferred or you didn't, right? The "didn't" column includes anyone who didn't choose to spend one of their 4 years playing at another college/university. So the easiest way to tell if our transfer rate has increased is to look at the data and ask, "what is the probability that a player will transfer from Duke?" Or "what are the odds that a player will transfer from Duke?" And if that has held pretty consistent, you can say the transfer rate hasn't changed much. Which is the case here.

Duval was a bad example, I always forget how highly rated he was (I wasn't very enthused about him going in from what I had seen/heard and tend to forget that didn't reflect the recruiting consensus at the time). I do think any roster turnover analysis that treats Shavlik Randolph, Trevon Duval, and Zion Williamson's departures as the same probably isn't telling you anything very interesting about your roster turnover (although that isn't directly related to my criticisms of the transfer analysis, I realize it is a separate question, but also one with a more useful answer).

[Acymetric]

There is no "confounding issue" that makes the data on gains and losses of players useless. For example, look at the long-term averages of "recruited roster" size versus average number of recruits. It tells you how many years you are getting per recruit -- useful for some purposes (and the cost accountants would love it). The "feedback" that poses some statistical complexity doesn't seem very important to me -- feedback, meaning that the scholarship limits prevent teams from recruiting more players. Most teams have fewer "recruited players" than the scholarship limits, so that if an outstanding prospect or transfer presents himself, then he can be put on scholarship.

When you say "long-term averages" are you talking about some kind of rolling average across multiple years? Because that's exactly what I'm saying would be useful/valuable.

[907bluedevils]

Can we please split off into two threads - one on 2021 Offseason Transfers and another on Trends in Transfers? I keep coming to this thread hoping for news on transfers and instead it seems kind of bogged down in hair splitting about whether or not Duke has a lot of transfers vis a vis other schools. I have my thoughts on that (there is not statistically significant data given how small basketball rosters are and there have been too many major paradigm shifts in the last 10-15 years to draw an educated conclusion). So I don't personally think it is really worth discussing. But just because I don't think it is an issue doesn't mean that others aren't allowed to discuss. I just wish you would do so elsewhere. Thanks.

I agree 100% agree, are mods able to help out?

[CDu]

I can't quite figure out how you got from Kedsy's numbers to these numbers (obviously you're including/excluding some of the edge cases, but I can't figure out which ones to arrive at these numbers).

I didn't include/exclude anything. Kedsy's analysis had 6 of 32 players transferring in the first 11 years, 7 of 38 in the second 11 years, and 9 of 52 in the last 11 years. So the odds for each is:

First 11 years: (6/32) / (26/32) = 6/26
2nd 11 years: (7/38) / (31/38) = 7/31
3rd 11 years: (9/52) / (43/52) = 9/43

Pretty standard odds calculations: p/(1-p).

As for the comment about roster turnover in general, yes that IS a separate question. If we were talking about roster turnover, then confounding would indeed be a huge problem. But it's really not in this setting.

And regardless, using the odds ratio approach eliminates the confounding when talking about the very narrow question of transfer rate. It of course doesn't help with the different question about roster turnover in general, but again that's a different question. And it doesn't get into qualitative discussions about what is or is not different between various early exits (which isn't terribly interesting to me, but to each their own).

[scottdude8]

Rocket Watts in the portal and some dude from Kansas. It’s a staggering number of transfers.

Rocket will be an interesting one to watch. He may have disappointed compared to the sky high expectations he had coming into East Lansing, but he's shown he can be a solid scorer at a major conference level. He just clearly isn't a point guard, which is what Izzo shoehorned him into being (side note: this was writing on the wall after MSU got the transfer from Northeastern a couple days ago). Rocket could very easily have a career path like Quentin Grimes at Houston (after leaving Kansas) or Quinnerly at Alabama (after leaving Villanova), both of which seem to have blossomed while not having to play as a traditional point guard.

He's a good enough player that I'm glad to see him leaving my hated Spartans.

[jaywilliams22]

Very nice get for Kentucky...grad transfer Kellan Grady from Davidson

https://twitter.com/GoodmanHoops/sta...102601737?s=20

[Utley]

Agreed - I think he would have been a great get here too. Did we pursue or are we hand tied until the Keels announcement?

Very nice get for Kentucky...grad transfer Kellan Grady from Davidson

https://twitter.com/GoodmanHoops/sta...102601737?s=20

[SavDukeGrad]

Very nice get for Kentucky...grad transfer Kellan Grady from Davidson

https://twitter.com/GoodmanHoops/sta...102601737?s=20

Darn. I was kind of hoping we would offer him if Keels doesn’t pick us. Great get for UK, unfortunately.