Raptors analytics... and an internal rift? [post #67]

[hateslosing]

It was very thought-provoking. I think somebody earlier mentioned that it all comes down to points per possession.

10 possessions, 1 shot each:

3 pt shots (30%) = 10 x 0.3 x 3 = 9 points

2 pt shots (40%) = 10 x 0.4 x 2 = 8 points


Obviously there's much more to a game than simply choosing between 2 equally open shots, but it's interesting to think about. It's also no suprise, given the success the Raptors have had in games when their 3pt shots have been falling.

Really that's not even analytics...it's common sense and basic math.

Kyle Lowry shoots 40% from three (actually 39% but 40% work nicer) which is the equivalent of someone shooting 60% from the field. (calc: 3*0.4=1.2 1.2/2=0.6)

Conclusion: Kyle Lowry shooting threes is our better than anyone on our team shooting two's.

Bigger conclusion: John Lucas III shoots 42% from three. He should take all the shots ever.

Obviously this is an oversimplification, JLIII doesn't get open for threes is Gay, DD, and Amir arent drawing the defense in. This does make me re-evaluate the value of a guy like Terrance Ross though, who likes to jack threes. Maybe he's actually making good decisions: he shoots 33% from 3, which is equivalent to almost 50% from the floor.

[JimiCliff]

This is the fine balance of coaching. Do you make the kid ride the pine for his errors or do you let him learn on the fly? Finding the balance between the two is the whole art of developmental coaching. Too much of a hardass and you kill his confidence, but too much lee-way isn't any good either as it tends to create the primadona attitude that sabotaged many a NBA careers.

For me, the pinacle of NBA coaching can be summed up in a simple sentence, "What would Pop do?". I think Popovich is the best coach in the NBA (and has been for a long time) because he gets the most out of both the superstars and the role players. I have to think that Pop would sit JV on the pine if he missed a rotation too. Now I know the context is different since the Spurs are actually a contending team, but the philosophy should be the same.

+ 1,000,000 on this statement...

however...

Pops is known for putting his young, inexperienced players in crunch-time and letting them play:

http://espn.go.com/blog/truehoop/pos...gregg-popovich (well worth the read!)

He may have a quick hook sometimes, but that isn't his only way.

[phez]

the 3pt analysis is silly. if they are going entirely by logic, then they should be saying every team should just drive to the net: much higher %, increased chance for a foul on the opposing team, and thusly the chance for the third point from the free throw line, which is also very high %, all three of these metrics derozan is outstanding.

the ghost players also really dont show anything a video replay doesnt already show you. "thats you amir johnson, and thats you not defending david west" simple?

the only thing i guess this is useful for is the fact the software can recognize plays. so you could analyze a large batch of games using computer data, and let the software tell you "you suck against pick and rolls" or "always leaving wings undefended"

[JimiCliff]

While I like the overall use of advanced analytics, I have to agree with Casey in the sense that you don't want the numbers developing his "star". Players still need to learn and sometimes the old fashion way works best. I would love to see more JV and less Gray, especially at this point in the season, but how can JV learn if there is no ramifications to his actions?

Ideally, the Raps can use the ghost defence and project the ghost players on the court in practice to show the players how it's done. Heck, at this point I'd let the ghost team projections take the court in a real game, couldn't be worse

He can learn by being given more court time. The idea that taking away a player's minutes when they blow an assignment doesn't hold water to me, unless it's clear that they're willfully contradicting what they've been told to do. But if they just aren't able to 'get it' yet but are making an effort, I say let them play through their mistakes. Especially in Jonas' case, because the numbers show that he can make up for his mistakes.

[Nilanka]

I've stated before that I can appreciate a coach benching a rookie for not following orders on the court, but I disagree with the player being glued to the bench for the remainder of the game. I like the idea of making a rookie (or any other player really) earn his minutes and learn from his mistakes.

However, I think any player worth developing should be given the opportunity within the same game (ie: same opponents, same gameplan, same opportunities to run plays on both ends of the court) to demonstrate his ability to properly execute the gameplan and to prove both his desire and ability to learn from mistakes (and from the coaching that comes in-game while benched for making the first mistake). This is where I see Casey's approach lacking, in that he has the quick hook, without providing the opportunity for redemption.

It also raises the question of how/what we define as a "mistake". If Jonas has the athletic ability to correct his "mistakes" on the fly, and has a net positive impact on the floor, can his mistakes really be considered "mistakes"?

It's like saying that when performing a layup from the left side, one should go up with one's left hand. But if the player goes up using his right hand instead (again from the left side)....and still scores....can it really be considered a mistake? Sure it's a mistake from a textbook POV, but the net result is the same (that is, a made bucket).

Regardless, I would assume that winning games is the ultimate goal of any franchise. That's hard to do when one of your net positive players is sitting on the bench because he achieved his net positive status in a way the coach disapproves of....

[hateslosing]

the 3pt analysis is silly. if they are going entirely by logic, then they should be saying every team should just drive to the net: much higher %, increased chance for a foul on the opposing team, and thusly the chance for the third point from the free throw line, which is also very high %, all three of these metrics derozan is outstanding.

the ghost players also really dont show anything a video replay doesnt already show you. "thats you amir johnson, and thats you not defending david west" simple?

the only thing i guess this is useful for is the fact the software can recognize plays. so you could analyze a large batch of games using computer data, and let the software tell you "you suck against pick and rolls" or "always leaving wings undefended"

Not neccesarily. A Let's say someone makes 69% of the shots at the rim, which is pretty good (it's what Amir makes and he leads all PF who have played at least 40 games this year http://www.hoopdata.com/shotstats.as...&gp2=40&mins=0). Using the math presented earlier, if you had a player who shot at least 46% from three, and there are several players in the league who do (Calderon has since the trade), you are actually doing better than taking a shot at the rim. Now if we look at the league average for shots at the rim, 64%, that is the equivalent of about 42%, which is what JLIII and a bunch of other guys shoot (I count 33, but some of those are guys who don't take threes regurly).

Point is, there are some players that you would rather have shoot threes than take a shot at the rim. Now this doesn't account for free throws which is a big benefit to getting to the rim, but then it is also easier to get a good shot from three than it is to get to the rim.

Edit: Interestingly, Terrance Ross has the highest at rim shooting % among all SG's and is 6th overall.

[iblastoff]

It also raises the question of how/what we define as a "mistake". If Jonas has the athletic ability to correct his "mistakes" on the fly, and has a net positive impact on the floor, can his mistakes really be considered "mistakes"?

It's like saying that when performing a layup from the left side, one should go up with one's left hand. But if the player goes up using his right hand instead (again from the left side)....and still scores....can it really be considered a mistake? Sure it's a mistake from a textbook POV, but the net result is the same (that is, a made bucket).

Regardless, I would assume that winning games is the ultimate goal of any franchise. That's hard to do when one of your net positive players is sitting on the bench because he achieved his net positive status in a way the coach disapproves of....

i would assume the right hand basket from the left side would be considered a 'mistake' because it has a lower chance of success? so even if the basket did go in, you don't want your player to continuously attempt it in the long run.

[Nilanka]

He can learn by being given more court time. The idea that taking away a player's minutes when they blow an assignment doesn't hold water to me, unless it's clear that they're willfully contradicting what they been told to do. But if they just aren't able to 'get it' yet but are making an effort, I say let them play through their mistakes. Especially in Jonas' case, because the numbers show that he can make up for his mistakes.

+1

Jonas doesn't exactly seem like the type to be like, "Screw you coach. I only care about minutes".

I'm sure he's absorbing everything Casey's telling him, regardless of how many minutes he plays.

[CalgaryRapsFan]

I respect that you have a different philosophical approach to developing a rookie learning from mistakes, but that`s very different than pointing to analytics saying that the D is better with JV to paint the coach as a dumbass.

I never said anything about the bolded part, I was just continuing the conversation from your post.

Personally, I think coaching experience and analytics could be integrated to improve the overall management of a team, but I certainly wouldn't want coaches replaced by analytics.

[Nilanka]

i would assume the right hand basket from the left side would be considered a 'mistake' because it has a lower chance of success? so even if the basket did go in, you don't want your player to continuously attempt it in the long run.

I understand that....but what if a player has a knack for seeing seams in the defense, can jump higher/quicker than most defenders, and actually scores consistently from the left side using his right hand? If the statistical analysis proves this to be true, can it be considered a mistake?

[themasao]

One thing that was interesting, but a bit puzzling for me (although I haven't put THAT much thought into it) was the way the analysis of what the 'ghost players' should be doing on defense seemed to belie what it recommends players do on offence.

Specifically, they were criticizing the Raps for not rotating hard to close out on a long 2-pointer from West in that Indiana possession, but what's the point in scrambling off of the 3 point line and off of Hibbert in the post (what it claims, on offence, to be the most valued shots) in order to contest a shot that, ideally, you want the other team to be taking.

In other words, why does it call for over-helping on D when it on O it emphasizes the pivotal importance of 3 pointers (which is exactly what you give up when you over-rotate).

Need some more time to digest this, and I really love the analytical approach, but I'm scratching my head a bit right now.

[JimiCliff]

One thing that was interesting, but a bit puzzling for me (although I haven't put THAT much thought into it) was the way the analysis of what the 'ghost players' should be doing on defense seemed to belie what it recommends players do on offence.

Specifically, they were criticizing the Raps for not rotating hard to close out on a long 2-pointer from West in that Indiana possession, but what's the point in scrambling off of the 3 point line and off of Hibbert in the post (what it claims, on offence, to be the most valued shots) in order to contest a shot that, ideally, you want the other team to be taking.

In other words, why does it call for over-helping on D when it on O it emphasizes the pivotal importance of 3 pointers (which is exactly what you give up when you over-rotate).

Need some more time to digest this, and I really love the analytical approach, but I'm scratching my head a bit right now.

The ghosts represent how the coaching staff wants the team to play. Apparently, they want a close-out on that shot.

[themasao]

The ghosts represent how the coaching staff wants the team to play. Apparently, they want a close-out on that shot.

Are you sure? I could have sworn the article was suggesting that the ghost players are doing what would yield the most effective decision from a analytical PPP perspective, and that the coaches aren't totally on board with replicating the 'ghost movements'.

Specifically, Jonas over-rotates, and Casey benches him as a result, but the ghost projection suggests that this is actually the better thing to be doing on the whole.

[Puffer]

Yeah that three point shot part was really crazy. So Demar is better off shooting threes than long two's. Doesn't mean I won't cringe every time he shoots a three.

Especially Demar. How many of his shots are just 18" inside the line? He might as well step back the half p[ace for the 50% boost in scoring if he hits it.

[Puffer]

+1

Jonas doesn't exactly seem like the type to be like, "Screw you coach. I only care about minutes".

I'm sure he's absorbing everything Casey's telling him, regardless of how many minutes he plays.

And there is a big problem with this if Casey is telling him "Don't leave your man to Help" when the analytics shows that leaving your man to help is the correct play 95% of the time.

[Puffer]

Are you sure? I could have sworn the article was suggesting that the ghost players are doing what would yield the most effective decision from a analytical PPP perspective....

This is correct. The ghost players are reacting in a manner that is in accord with giving the opponents the lowest percentage shots, based on analysis. The article specifically said that what many would call "over helping" is the better move most often.

[RapthoseLeafs]

.

This analytical approach throws a few curves to the typical coaching approach, that spices its' thinking with traditional stats - some simple; some more complicated.
.

Our results suggest that the integration of spatial approaches and player tracking data promise to improve the status quo of defensive analytics but also reveal some important challenges associated with evaluating defense. [Kirk Goldsberry and Eric Weiss - Harvard University]

I posted in a previous thread State of the Nation - The Long Division Version, where I mentioned this Harvard study based on data gleamed from the SportVu camera systems, currently used by 15 teams.

The Dwight Effect: A New Ensemble of Interior Defense Analytics for the NBA

The argument in my thread related to Bargnani, and his posting the 2nd best number. The study summarized fg% of opponents when each Defender was within 5 feet. Of course this analysis only applied to 15 teams in the study, so possibly there are 225 players (give or take a few) + the current #1 who are better then Andrea.

When I hear of stats like these, and the study by Harvard, I reason that fans are tainted more by AB's display of indifference (and shaky rebounding), then by the myth that his defense is the biggest contibutor to this team's lack of success.

Never-the-less, we could debate till the cows come home, without realizing that there may have been a reason for Casey's obsession with putting Bargnani out there (in place of another Big), and at the 3 point line. The former seems to be for defensive purposes, while the latter is because even at .311 - his worse 3 pt shooting average - Andrea still nabs more points there (per possesion), then Rudy, or Demar, or Kyle do, when inside the perimeter.

In his role of 2nd banana (to Bosh), Bargnani averaged .375 from the 3 point line over a 4 year period. That translates into an equivalent .563 - the current average this year for our top shooter .... Amir Johnson.

.

Judging these numbers - like Bargnani's - against certain members of the team, should not deflect from the reality that Rudy's numbers don't make for a good comparison. Or measure of success. But it's important to realize the relevance for which certain roles evolved. This analysis should put some pressure on Demar, as his 3 point shooting needs a boost in skill level.

.

[JimiCliff]

Are you sure? I could have sworn the article was suggesting that the ghost players are doing what would yield the most effective decision from a analytical PPP perspective, and that the coaches aren't totally on board with replicating the 'ghost movements'.

Specifically, Jonas over-rotates, and Casey benches him as a result, but the ghost projection suggests that this is actually the better thing to be doing on the whole.

With respect to defence, it's actually unclear. The article says:

"Those are ghost players, and they are doing what Toronto's coaching staff and analytics team believe the players should have done on this play — and on every other Toronto play the cameras have recorded."

But I'm going to go out on a limb and suggest that the analytics team have been listening to the coaches more than the other way around, which would mean the defender ghosts are more reflective of the ideas of the coaching staff.

With respect to offence, the article makes it clear that there are some philosophical divides between coaching and analytics. But that had nothing to do with the ghosts really.

[Craiger]

.

This analytical approach throws a few curves to the typical coaching approach, that spices its' thinking with traditional stats - some simple; some more complicated.
.



I posted in a previous thread State of the Nation - The Long Division Version, where I mentioned this Harvard study based on data gleamed from the SportVu camera systems, currently used by 15 teams.

The Dwight Effect: A New Ensemble of Interior Defense Analytics for the NBA

The argument in my thread related to Bargnani, and his posting the 2nd best number. The study summarized fg% of opponents when each Defender was within 5 feet. Of course this analysis only applied to 15 teams in the study, so possibly there are 225 players (give or take a few) + the current #1 who are better then Andrea.

When I hear of stats like these, and the study by Harvard, I reason that fans are tainted more by AB's display of indifference (and shaky rebounding), then by the myth that his defense is the biggest contibutor to this team's lack of success.

Never-the-less, we could debate till the cows come home, without realizing that there may have been a reason for Casey's obsession with putting Bargnani out there (in place of another Big), and at the 3 point line. The former seems to be for defensive purposes, while the latter is because even at .311 - his worse 3 pt shooting average - Andrea still nabs more points there (per possesion), then Rudy, or Demar, or Kyle do, when inside the perimeter.

In his role of 2nd banana (to Bosh), Bargnani averaged .375 from the 3 point line over a 4 year period. That translates into an equivalent .563 - the current average this year for our top shooter .... Amir Johnson.

.

Judging these numbers - like Bargnani's - against certain members of the team, should not deflect from the reality that Rudy's numbers don't make for a good comparison. Or measure of success. But it's important to realize the relevance for which certain roles evolved. This analysis should put some pressure on Demar, as his 3 point shooting needs a boost in skill level.

.

You should perhaps do a little more research on that study. Andrea's #s are so 'good' because he rarely gets within 5 ft which skewed the #s (explained and detailed at sloan and by the author). When that is accounted for Bargnani is actually one of the worst in the leage. That basically supports that Bargnani lack of defense is not a myth, rather very well statistically quantified.

[phez]

With respect to defence, it's actually unclear. The article says:

"Those are ghost players, and they are doing what Toronto's coaching staff and analytics team believe the players should have done on this play — and on every other Toronto play the cameras have recorded."

But I'm going to go out on a limb and suggest that the analytics team have been listening to the coaches more than the other way around, which would mean the defender ghosts are more reflective of the ideas of the coaching staff.

With respect to offence, the article makes it clear that there are some philosophical divides between coaching and analytics. But that had nothing to do with the ghosts really.

in the pick and roll video the players pretty much did the same play what the ghost players did. the ghost players were just rotated into slightly different spots. the amir johnson one didnt show anything different either.

again saying it doesnt really add anything over a regular video replay, aside from being able to crunch the data from a vast amount of games.