Why did we get the Kyrie Irving trade so wrong?

Home for all your discussion of basketball statistical analysis.
Post Reply
schtevie
Posts: 362
Joined: Thu Apr 14, 2011 11:24 pm

Why did we get the Kyrie Irving trade so wrong?

Post by schtevie » Mon Feb 12, 2018 1:18 am

Kevin Pelton offers comment on this topic (http://www.espn.com/nba/story/_/id/2239 ... pelton-nba) stipulating the case and goes about documenting how (and suggests why) Thomas' and Crowders' performances have cratered this year, according to advanced statistics.

Curiously, one factor not directly discussed is an "all else equal" condition that implicitly applies in this analysis, and it is coaching, in the instance, of which I speak. Specifically, Crowder and Thomas went from playing for Stevens to Lue, and Irving the reverse. What difference might that have made?

For those who might have some recollection of my comments here over the years on this subject, unlike some, I am not one who believes that the best coaches in the league put as many net points on the board as the best players. However, I do believe that if you subtract the value of the best coach from that of the worst, the plus-minus difference becomes non-trivial, and perhaps this situation has obtained.

So, what would be evidence along the line that Stevens minus Lue might be a non-small augment in perceived "player value"? One, broadly reputational, is that Stevens is understood (I think correctly) to be a very good coach. Conversely, Lue has only seen period upon period decline in the Cavaliers' on-court performance. His win-loss record was worse than Blatt's, within season, and it has gone downhill from there, the following two years.

Other suggestive evidence, to me, is that it is a really odd coincidence that not one but two players, who have had careers where they were consistently above-average, move to Cleveland and immediately see their RPM decline by around 6 points each. That, it could go without saying, is a lot. It is almost as though the receiving coach was unprepared to replicate the conditions of their previous success... And then on the opposite side, Kyrie Irving improves a bit in Boston, almost regaining his peak career RPM number (what was in 2014-15, under Blatt).

So to serve as a bookmark, here are today's RPM numbers for Thomas and Crowder (what includes one game for Thomas in LA). It will be interesting to see if and how they might move in the remainder of the season.

Isaiah Thomas: ORPM = -1.99, DRPM = -2.62, RPM = -4.61
Jae Crowder: ORPM = -0.56, DRPM = -1.51, RPM = -2.07

And I will leave it here for now...

Crow
Posts: 6206
Joined: Thu Apr 14, 2011 11:10 pm

Re: Why did we get the Kyrie Irving trade so wrong?

Post by Crow » Mon Feb 12, 2018 2:27 am

Coaching is probably part of it. Team fit is similar but a different part of it. 2 cases is dangerous to extrapolate from. Half a season, dangerous. Age inflextion point and injury are factors too. Player expectations / flexibility. Teammate reactions / interactions.

Lets see what happens with Hill, Hood, Nance, Clarkson. And Thomas and Crowder on new team. And everybody in a year or two. Estimate errors are some level of factor as is up n down play from player average seemingly independent from specific named causation.

Anybody want to check & summarize by team, coach, player quality level, role, etc. the in / out yr to yr RPM estimate changes for the last 500 player moves? This would be work but not huge on the scale of huge. That is what is needed. Some coaches might help / hurt a majority of all kinds, some might have mixed impacts, maybe related to mindset / experience.

Could look at how much change in box score stats yr to yr for Thomas and Crowder would explain of their RPM changes. What signs are there to support the non-box score component? How much might be estimate error? How volatile was the estimate during the season in absolute terms and compared to average for this season and the one before?

This analysis aside, don't we agree that multi-year and age adjusted would be closer on average for future estimates?

Coaching "quality" or goodness of fit or "luck" could vary year to year or month to month. How much of RPM is good lineup usage and how much of that was intelligent choice vs random? When coaches spray players into 500 lineups, most dink lineups does true player value display easily? Or does it face high friction & stress & randomness? Going back into RAPM splits, how does player RPM look in non-dink lineups vs dink and super-dink lineups (which get the most minutes)?

Looking at the raw data Crowder's biggest minute lineup was good. The next 9 bad, mostly horrendous. 7 minutes per game in the good lineup is something but why just 7 when the gulf in results is so huge? The big minute lineup in the abstract is more likely to measure effectiveness than the way small ones. Coach doesn't know or doesn't believe in the sample results? The Thomas case had less minutes and more half n half good and bad results. Of course adjusted lineup data would be preferable. Another potential impressionable RAPM that I'll mention would be multi -year RAPM by play with x # of starters. (Dozens of RAPM splits identified over the years) Did Thomas and / or Crowder see a marked change in frequency of time with a high number of starters from yr to yr? Probably. Did this matter? Got to look to guess and looking with RAPM splits worth checking alongside the raw over just the raw or not checking at all.

Looking at the raw data for top 10 most used Crowder lineups, this season was his worst set. Last season his best. Before that mixed under Stevens and Carlisle. Maybe last season was genius usage by Stevens with the benefit of experience. Was Lue's usage getting better or not? By raw data Crowder with the starters was good in first third of season and twice as good in next third. Next 5 lineups sucked in both periods. If you see and react to that I wouldn't call you a responsive coach or organization.

That is what I can ask, see and say in an hour.

Crow
Posts: 6206
Joined: Thu Apr 14, 2011 11:10 pm

Re: Why did we get the Kyrie Irving trade so wrong?

Post by Crow » Mon Feb 12, 2018 3:47 am

New hour...

Best bigger minute Thomas lineup? Without LeBron. Tiny sample but raises question what is the longer history of Thomas performance with other ball dominant, high usage players or stars in general. More RAPM splits to add to raw data. Thomas in Boston last season, better overall team plus minus in player pairs with low usage players vs. high usage players. Appears better when he doesn't have to share as much. Should Lue have known that and relied on that in his lineup usage more? I'd say so.

Will Walton? First game, lots of minutes with low usage Hart and Brewer. Minutes with high usage Lopez and Randle but little if any with both at same time. Intelligence or random? We'll see but so far Walton looks good on this.

Crow
Posts: 6206
Joined: Thu Apr 14, 2011 11:10 pm

Re: Why did we get the Kyrie Irving trade so wrong?

Post by Crow » Mon Feb 12, 2018 5:40 am

What people or people using RPM got least wrong about "the Kyrie trade" was with Kyrie, especially Offensive Kyrie. His ORPM estimate is the same and his DRPM is about a point better. More effort, better system, maturing could explain that as could randomness.

His basic role didn't change that much. He elevation to #1 fit his personality / desires better. Thomas' context changed a lot and the role was less and less desirable. Crowder's average context changed a lot.

Will Thomas' RPM for this season end up close to the average of his last 2? How about the following season? Would anyone have a problem concluding that most of the change is related to injury and aging?

J.E.
Posts: 818
Joined: Fri Apr 15, 2011 8:28 am

Re: Why did we get the Kyrie Irving trade so wrong?

Post by J.E. » Tue Feb 13, 2018 1:54 pm

I've mentioned it on Twitter a couple of times, but I think there are several factors that contribute to their diminished performance. Some of them out of anyone's control, some of them were not weighed as heavily as they should have been

You already mentioned the coaching change. Stevens wasn't rated highly before this season in coach RAPM (but now is), but Lue has always been rated low if not last.
Thomas' injury made me more aware that small players have to be close to 100% to still be effective. If they don't have all their speed, it will be a struggle.
I'm no expert on hip injuries, but I'm sure they impact speed significantly.
I don't think Thomas was 100% when he played his first game. Also, he obviously didn't go through pre-season, and NBA teams don't really practice during the season
For Crowder, his mom dying hours after he told her that he got traded must've been tough

Still, there's no denying that RPM expected different contributions from both of them. That said, with Irving not deviating much from his past rating, and with Thomas also cratering in essentially all other metrics, it was only Crowder where RPM was more "off" than other metrics. Given Crowder's stellar +/- performance during most his career, even in Dallas, that's a mis-prediction I don't feel too awful about

schtevie
Posts: 362
Joined: Thu Apr 14, 2011 11:24 pm

Re: Why did we get the Kyrie Irving trade so wrong?

Post by schtevie » Sun Feb 18, 2018 3:30 pm

Jeremias, should you have the time and inclination, I (and perhaps others) would be very interested in better understanding the context of the effect of changing teams on expected player RPM. There are what, four full seasons now of RPM? Could you provide a graph and/or summary statistics for this period for players changing teams of expected RPM vs. (full-season) actual RPM. (And within-season team changes would be another set of interesting data.) I am supposing that expected RPM is just last year's RPM plus an age adjustment (yes?)

Some questions:

(1) is there an average transfer effect? On the one hand, trades, in theory, ought be mutually beneficial, but improving direct on-court performance isn't the basis for the vast majority of player transfers (those being salary cap considerations, draft picks, etc). Accordingly, I expect that the simple average is probably negative.

(2) Relating to larger issues of this particular trade, what are the priors used in RPM for rookies? My recollection from xRAPM (likely incorrect) is that there was a standard -3.0 for all, regardless of draft position or round. Is that the case, or are there now gradations?

Thanks, in advance, for any enlightenment on these issues.

J.E.
Posts: 818
Joined: Fri Apr 15, 2011 8:28 am

Re: Why did we get the Kyrie Irving trade so wrong?

Post by J.E. » Thu Feb 22, 2018 4:05 pm

Could you provide a graph and/or summary statistics for this period for players changing teams of expected RPM vs. (full-season) actual RPM. (And within-season team changes would be another set of interesting data.)
I probably won’t have time to do that until the summer
(1) is there an average transfer effect? On the one hand, trades, in theory, ought be mutually beneficial, but improving direct on-court performance isn't the basis for the vast majority of player transfers (those being salary cap considerations, draft picks, etc). Accordingly, I expect that the simple average is probably negative.
https://twitter.com/kmedved brought it up first that players that switch teams don’t always reproduce their +/- part of their impact. To what extent that’s the case I will have to look into
(2) Relating to larger issues of this particular trade, what are the priors used in RPM for rookies? My recollection from xRAPM (likely incorrect) is that there was a standard -3.0 for all, regardless of draft position or round. Is that the case, or are there now gradations?
Rookies were only treated differently in an old, non-ESPN version. The -3 prior became obsolete when BoxScore data came into play

Dr Positivity
Posts: 298
Joined: Thu Sep 20, 2012 6:44 pm

Re: Why did we get the Kyrie Irving trade so wrong?

Post by Dr Positivity » Tue Feb 27, 2018 8:49 pm

+/- stats suggested Isaiah was an abomination on defense, Kanter level. If you looked at the rest of their roster last year, on/off and team DRTG, it all looked like Isaiah was so bad to be taking them from 1st level defense to middle of the pack defense just on his own. And their results this year have backed that up perfectly.

Both Crowder and Patterson on Toronto were +/- heroes for their team that haven't been missed much. Their fit on the team correlated very well with their success, but talent wise I guess they were easy to replace. It also gets worse for them when you realize the goal is to win a title and not the regular season. As a Raptors fan it became obvious in recent years that even if Patterson was all that in terms of regular season value, we had to start thinking of what talent you over the hump in postseason play and Patterson was certainly not that

Post Reply