Moderators AHF Posted August 16, 2021 Moderators Report Share Posted August 16, 2021 4 minutes ago, Final_quest said: Yep, it was based off a foggy memory, and I didn't feel like finding the needle in a haystack. Even for something like summer league data tells a story. It's less predictive than other data sets, but performance is not completely meaningless. Will post if I find anything. Thanks, @Final_quest. I'm very skeptical of its predictive power due to the lack of structure, the talent level and small number of games but would love it read it. Link to comment Share on other sites More sharing options...
Final_quest Posted August 16, 2021 Report Share Posted August 16, 2021 Here's the best article I found in a few minutes of searching. https://model284.com/the-predictive-value-of-nba-summer-league-nba-rookies/ Link to comment Share on other sites More sharing options...
Moderators AHF Posted August 16, 2021 Moderators Report Share Posted August 16, 2021 17 minutes ago, Final_quest said: Here's the best article I found in a few minutes of searching. https://model284.com/the-predictive-value-of-nba-summer-league-nba-rookies/ So for projecting rookie performance, it looks pretty weak: Quote Projecting NBA Rookie Season Performance We’ve explored correlations and relationships of simple box score statistics from Summer League to NBA Rookie Season performance. But how much predictive power does NBA Summer League have on overall rookie performance? In order to capture NBA rookie season performance, we used a metric called Win Shares (though other all-encompassing metrics such as Value Over Replacement Player, Box Plus Minus, Real Plus Minus, and more were also explored). Win Shares is a player statistic that attempts to divvy up credit for team success to the individuals on the team. Win shares can technically be any value positive or negative, and a negative value would mean a player was so poor he essentially took away wins from his team. For more info on the calculation of Win Shares, see this breakdown. Here is an example of the top 5 Win Shares from 2015 NBA Rookies, to give us a better idea of what the Win Shares metric looks like: Player Win Shares Karl-Anthony Towns 8.3 Kristaps Porzingis 4.3 Willie Cauley-Stein 3.8 Willy Hernangomez 3.4 Frank Kaminsky 3.3 Now, using Win Shares as the response variable, I built a simple multiple linear regression model to explore how much variance in Win Shares can be explained by NBA Summer League Performance. After adjusting for minutes, the following predictive variables were used: Total Rebounds (TRB), Field Goal Attempts (FGA), Free Throw Attempts (FTA), Three Point Attempts (X3PA), Assists (AST), Steals (STL), Blocks (BLK), Turnovers (TOV), and Personal Fouls (PF). Note that each player’s numbers were standardized by position prior to fitting the model. The table below shows the coefficient for each variable in the regression model: INTERCEPT TRB FGA FTA X3PA AST STL BLK TOV PF 0.54 0.06 0.25 0.28 0.07 0.19 -0.05 -0.01 -0.02 0.01 One measure of so-called predictive power is adjusted R-squared. First off, R-squared, also known as the coefficient of determination, is the proportion of variance in the dependent variable (Win Shares) that is explained by the predictive variables (box score stats). Adjusted R-squared takes into account the number of independent variables included. In this case, our “simple” model explained roughly 10.9% of the variance in rookie win shares from the aforementioned Summer League statistics, which is not particularly high. In the final model, I explored a number of ratios and interactions, ultimately settling on two interactions between Field Goal Attempts and Free Throw Attempts as well as Assists and Turnovers (in addition to our previously included variables). This final model showed Summer League performance explaining 15.5% of the variance in Rookie Win Shares. While this isn’t astronomically high, it does show there is some predictive power in Summer League on NBA Rookie performance. Their for fun more anecdotal predictive experiment also showed pretty weak correlation with almost none of the players' actual performance particularly close to the predicted numbers: Quote Out-of-Sample Predictions Now let’s have a little fun and make some out-of-sample predictions. The idea of out-of-sample predictions stems from the concept of cross validation, which is the process of removing a sample of data, known as the testing dataset, and then training your model on the remaining data, the training dataset. Finally, we use the trained model to make predictions on the test data. In this case, I removed a single player from the dataset to test on and trained on the remaining players in the data. Below are the top 10 predicted Rookie Win Shares: Player Predicted Rookie Win Shares Actual Rookie Win Shares Thomas Robinson 4.82 -0.1 Damian Lillard 2.66 5.8 C.J. McCollum 2.62 0.2 Jaylen Brown 2.55 1.5 Doug McDermott 2.36 0.0 Blake Griffin 2.21 9.8 Karl-Anthony Towns 2.13 8.3 Gary Harris 1.90 -0.7 Montrezl Harrell 1.86 0.9 Jabari Parker 1.80 1.3 While we can certainly spend more time building a “better” predictive model using various variable selection techniques and machine learning algorithms, the point of this exercise was less about prediction and more about understanding the relationship between summer league performance and NBA Rookie Season performance. Sometimes the simplest models are the most effective in that they provide easier interpretation. So for these players here is the % and # of WS by which each prediction was wrong: Thomas Robinson 4920% // 4.92 WS Damian Lillard 118% // 3.14 WS C.J. McCollum 1210% // 2.42 WS Jaylen Brown 70% // 1.05 WS Doug McDermott Can't be computed with 0 but sufficient to say way wrong // 2.36 WS Blake Griffin 343% // 7.59 WS Karl-Anthony Towns 290% // 6.17 WS Gary Harris 371% // 2.6 WS Montrezl Harrell 107% // .96 WS Jabari Parker 38% // .5 WS In a league where most players put up less than 4 WS in a total season, those are huge variations whether as a % or as a total win share value. Bear in mind they missed on WS by an average of 3.17 win shares per person. Wow. How much worse could that possibly be than blind guessing? For the 2020 draft class, you could simply guess 0.0 win shares per person and be much closer to the mark. (Simply guessing 0 without making any effort to adjust for individual players would have been closer than 3.17 win shares off for 13 of the 14 2020 lottery picks and 56 of the 60 overall players drafted in 2020.) Color me unimpressed. Link to comment Share on other sites More sharing options...
Popular Post TheNorthCydeRises Posted August 16, 2021 Author Popular Post Report Share Posted August 16, 2021 23 hours ago, MarylandHawk said: He did not. The year at Duke did not go as any of them expected. Jalen Johnson was going to be a one and done player from the outset. Coach K knew that. The plan for Jalen was to help them get a far as possible on a national championship run. Things got off to a great start, he had. 19 and 19 game in his debut appearance. Then injury hit, Coach chose not to play in several games due to the pandemic, Duke lost more games, the students were locked up in hotels and couldn't experience campus life. The team was not going to make the tournament. Jalen forced his way back on the court after the injury and had one of the most dominate performances in Duke history...in a loss. At that point, his foot was re-aggravated and the next couple of games he did not play well. THIS IS WHEN COACH K decided to go in a different direction to develop some of the younger guys on the team that he knew would be returning who also fit better into his system. At this point Jalen had to start considering what was best for him. Stay at Duke in a reduced role playing limited minutes. Stay at Duke rehab a foot injury. Or start preparing yourself for your first real job in life and get the best possible jump start that you can. He did that latter. Non of that sounds like quitting. His teammates supported the decision because it meant more playing time for them. Coach supported it because he could develop for next year. His agents supported it because he was protecting his draft stock. It worked out for all involved. Although he wasn’t a top 10 pick, he was clearly overjoyed to be drafted by the Hawks because there was a real moment where he felt his dream slipping away. There is a paradigm shift about college BB relevance as a path to the NBA. Kids have different options now and the traditional college supporters can’t wrap their minds around that concept. Nobody is mad that Steve jobs or Bill gates dropped out of college to pursue their dreams. Obviously there were a lot of emotions involved especially when you believe you are the best prospect in the class. We will sit back and enjoy the outcome as he develops into the person he believes he can be. 2 1 6 Link to comment Share on other sites More sharing options...
TheNorthCydeRises Posted August 16, 2021 Author Report Share Posted August 16, 2021 1 hour ago, txsting said: Let's not put all that weight on him just yet. He's a rookie, a prospect with potential. He's a project too, because he needs to install some upgrades in order to reach that potential. It's up to him how good he wants to be. That project tag will be thrown away after the 2nd preseason game. That guy is good. Real good. Link to comment Share on other sites More sharing options...
Final_quest Posted August 16, 2021 Report Share Posted August 16, 2021 Trust me. I’m not trying to impress anybody, AHF. If you want to maintain scoring 2 points or 20 points in 25 minutes has no predictive value, I’m not gonna try to change your mind. I literally only found two articles on the subject and posted the more comprehensive one. I actually didn’t think either article had a great focus. So, I’d still like to see a better effort from someone to tackle this subject. Would like to see something at least cover those who average over 20 points versus those who average under 10. I’d be shocked if there wasn’t something predictive about summer league. Link to comment Share on other sites More sharing options...
Moderators AHF Posted August 16, 2021 Moderators Report Share Posted August 16, 2021 7 minutes ago, Final_quest said: Trust me. I’m not trying to impress anybody, AHF. If you want to maintain scoring 2 points or 20 points in 25 minutes has no predictive value, I’m not gonna try to change your mind. I literally only found two articles on the subject and posted the more comprehensive one. I actually didn’t think either article had a great focus. So, I’d still like to see a better effort from someone to tackle this subject. Would like to see something at least cover those who average over 20 points versus those who average under 10. I’d be shocked if there wasn’t something predictive about summer league. I don't think 5 NBA games is a meaningful sample size from which to draw a lot of meaningful conclusions on players let alone 5 summer league games. The lack of structure and the quality of competition are just added factors on top of the already extremely small sample size. I'm sure there is some limited relationship between the summer league numbers and NBA production but I expect them to be very weak in the aggregate which is what I saw in this. That isn't a commentary on you at all. Just the numbers and the article which I view as being very limited in application. To the contrary, I really appreciate you sharing this. In my mind, the summer league is better for coaches to see some of the tools their players have (positioning, form, etc.) and get together a list of things for that player to work on than it is for GMs to be able to pencil in production from these players based on their summer league stats. 1 3 Link to comment Share on other sites More sharing options...
kg01 Posted August 16, 2021 Report Share Posted August 16, 2021 16 minutes ago, AHF said: I don't think 5 NBA games is a meaningful sample size from which to draw a lot of meaningful conclusions on players let alone 5 summer league games...... I can't believe my longtime pal, @TheNorthCydeRises came out with this thread. smdh Geez, 'Cyde ... 1 3 Link to comment Share on other sites More sharing options...
Premium Member Diesel Posted August 16, 2021 Premium Member Report Share Posted August 16, 2021 On 8/15/2021 at 7:08 AM, Diesel said: Summer League. The reason I say Summer League is because SL is filled with players trying to make a squad. No attention to a defensive scheme. I'm glad that he's hooping...but his numbers would have looked the same had he gone down to Grant Park at Noon on yesterday. ... and they would have been just as meaningless. Let's just wait and see what we have. The regular season will tell the whole story. Link to comment Share on other sites More sharing options...
HawkItus Posted August 16, 2021 Report Share Posted August 16, 2021 James Worthy esque Link to comment Share on other sites More sharing options...
Popular Post RedDawg#8 Posted August 16, 2021 Popular Post Report Share Posted August 16, 2021 I understand everyone's hesitation and their reasoning behind it, especially when based upon past examples and whatever. However, if I were a betting man, just based on eye-ball test alone, this class looks really talented and deep overall, more-so than others recently. We can wait 3 years before we decide it's ok to form an opinion them, and by that time we can look back on this thread and say some folks were right or wrong. I don't really need that long, I can see there is a little something different to this group. Im calling it now. Jalen Johnson will at least put up John Collins type production in the NBA at some point in his career. Situation plays a huge factor, and right now he is not in a great spot to get consistent minutes on this roster. That doesn't change my outlook on him and his talent. The only factors that can derail his potential would be injury, or some other off the court stuff. Those factors can not be predicted, so why even worry about them. Some guys just have the goods, and it doesn't take all day to recognize. Jalen's got the goods. He can play basketball. There are far less talented cats making careers in the NBA right now. Find me later when this is no longer a hot take lol 6 Link to comment Share on other sites More sharing options...
Final_quest Posted August 16, 2021 Report Share Posted August 16, 2021 3 hours ago, AHF said: I don't think 5 NBA games is a meaningful sample size from which to draw a lot of meaningful conclusions on players let alone 5 summer league games. The lack of structure and the quality of competition are just added factors on top of the already extremely small sample size. I'm sure there is some limited relationship between the summer league numbers and NBA production but I expect them to be very weak in the aggregate which is what I saw in this. That isn't a commentary on you at all. Just the numbers and the article which I view as being very limited in application. To the contrary, I really appreciate you sharing this. In my mind, the summer league is better for coaches to see some of the tools their players have (positioning, form, etc.) and get together a list of things for that player to work on than it is for GMs to be able to pencil in production from these players based on their summer league stats. Good points. Probably why no one has made much effort to do data analysis with summer league. Early returns look good for our rookies, though. 2 Link to comment Share on other sites More sharing options...
Moderators AHF Posted August 16, 2021 Moderators Report Share Posted August 16, 2021 2 hours ago, Final_quest said: Good points. Probably why no one has made much effort to do data analysis with summer league. Early returns look good for our rookies, though. I was thrilled on draft day and remain extremely pumped on our draft class. 2 Link to comment Share on other sites More sharing options...
Hawkmoor Posted August 17, 2021 Report Share Posted August 17, 2021 Better than Hunter and Cam? Jalen is a consistent jump shot from being the best player on the TEAM. Link to comment Share on other sites More sharing options...
ATLHawks3 Posted August 17, 2021 Report Share Posted August 17, 2021 Johnson should really work to improve on his jumper, screen setting, shot creation ability in the halfcourt, and team defense before the start of training camp if he wants to really threaten Gallo for that backup PF role. 1 Link to comment Share on other sites More sharing options...
Popular Post RandomFan Posted August 17, 2021 Popular Post Report Share Posted August 17, 2021 2 hours ago, Hawkmoor said: Better than Hunter and Cam? Jalen is a consistent jump shot from being the best player on the TEAM. 3 1 4 Link to comment Share on other sites More sharing options...
NBASupes Posted August 17, 2021 Report Share Posted August 17, 2021 43 minutes ago, ATLHawks3 said: Johnson should really work to improve on his jumper, screen setting, shot creation ability in the halfcourt, and team defense before the start of training camp if he wants to really threaten Gallo for that backup PF role. Bro, I never thought I would read it but from what I've seen, JJ is far from challenging for mins based on the fact that he's simply not good enough. His talent level is sensational but so is Marvin Bagley III and no one here would put him above Gallo or JC right now. He has a lot of work to do, once he does it and improves a lot than the only thing left is giving him the minutes. 3 Link to comment Share on other sites More sharing options...
ATLHawks3 Posted August 17, 2021 Report Share Posted August 17, 2021 1 minute ago, NBASupes said: Bro, I never thought I would read it but from what I've seen, JJ is far from challenging for mins based on the fact that he's simply not good enough. His talent level is sensational but so is Marvin Bagley III and no one here would put him above Gallo or JC right now. He has a lot of work to do, once he does it and improves a lot than the only thing left is giving him the minutes. Well, he's a rookie, so it is expected that there will be growing pains for him to adjust to the NBA. But like I said, he doesn't just have to wait until training camp starts, he (and Coop) should be working until then on improving to really impress the coaching staff. Link to comment Share on other sites More sharing options...
Popular Post kg01 Posted August 17, 2021 Popular Post Report Share Posted August 17, 2021 It absolutely amazes me how far y'all have gone based off some random SL games. Do better, squawk. Were it up to ya'll, Trae would've been traeded after his SL stint. 5 1 1 Link to comment Share on other sites More sharing options...
Premium Member MarylandHawk Posted August 17, 2021 Premium Member Report Share Posted August 17, 2021 2 hours ago, kg01 said: It absolutely amazes me how far y'all have gone based off some random SL games. Do better, squawk. Were it up to ya'll, Trae would've been traeded after his SL stint. We are not all necessarily agreeing with the subject of this thread especially based on summer league. However, when you factor everything in, his ceiling looks to be just as high if not higher. The advantage that Cam and Dre have over Jalen is that we were willing to lose NBA games in order to let them develop. (Because they didn’t develop fast enough, loyd got fired. The one thing we don’t talk about is how the veterans were forced to play heavy minutes this past season when they were injured and as a result we won more games.) Jalen won’t get the opportunity to play 30 mins a game and his rope will be really short. His minutes won’t be consistent and his role will be the 4th or 5th option when he is on the floor. Apples and oranges type situation. If he does somehow manage to earn minutes and play well despite all of that, I will be even more impressed. Hell, even if you look at Trae’s development, it took him a half of a season playing 35 mins a game to round into the player we know and love. The one last thing I want to mention about Jalen that I like so much is he brings things this team is missing and that we can benefit from even with him playing a limited role. Play making defense. Fast break opportunities. Offensive play making big. 50/50 ball winner. Uber athleticism (in addition to John). High motor. Those are things that can help a team win even in limited minutes. 3 Link to comment Share on other sites More sharing options...
Recommended Posts
Please sign in to comment
You will be able to leave a comment after signing in
Sign In Now