Regularization isn't there to specifically deal with 2 players playing at the same time. It will prevent the overfit you'll get with the limited input of one-playing-one-benched situations, sure, but it doesn't magically help you estimate correctly the impact of one player in the duo.
My comment is still valid, this model is bad at assessing one player's particular contribution if he's used consistently within a group of players that play with each other. Think of it like this, if a coach decided to use only either his starting 5, or platoon swap 5 other guys, every game, then you wouldn't be able to know which of the 5 starting guys is doing the heavy lifting.
A concrete example of this is DJ and Chris Paul, they played 29 minutes per game together last year. Each played around 31.5 minutes total (sharing 92% of their playtime). Both had very high DRPM (the defensive component of ESPN's version of RAPM). This year DJ is down by a lot in this metric (and CP is now a negative according to it, but he only has played 5 games). Did DJ become a worse defender? Or was the metric evaluating the duo rather than the individual parts like it claims (or rather, JimmyJ claims) it is able to do?
It's way easier for this model to assess Lowry and Derozan (pretty bad example actually): last year they played 37 and 35 minutes each, but shared only 26 minutes (70-74% of shared playtime).
On November 28 2017 02:04 ZenithM wrote: My comment is still valid, this model is bad at assessing one player's particular contribution
i agree that co linearity is a tough problem. labelling this model "bad" ... you'd have to compare to something else. xRAPM is better than APM and Vanilla plus-minus at cracking the co linearity problem. it does not solve it 100%. and it requires a lot of data to untangle 2 players that play together a lot.
more work will be done on this issue in the future as Plus/Minus, APM, RAPM, and xRAPM continue to be worked on by analytics guys.
On November 28 2017 02:04 ZenithM wrote: It's way easier for this model to assess Lowry and Derozan (pretty bad example actually): last year they played 37 and 35 minutes each, but shared only 26 minutes (70-74% of shared playtime).
good point. the model has been around 5 years now. in previous years they were on the court together a lot. the model always correctly identified Lowry as the better defender. and it always identified Derozan as very bad. Which he was. And Lowry as very good. Which he was.
co-linearity is still a problem. the player grouping thing is still a problem. it is a much bigger problem with APM and vanilla Plus Minus. xRAPM or "ESPN Real Plus Minus" goes a long way to untangling players that play together and figuring out who is responsible for what.
the model has always worked great for Raptors players since it was published by ESPN and given the consumer label "ESPN Real Plus Minus" 5 years ago. its why i started paying more attention to it 2 years ago.
Because Plus/Minus is such a valuable and unique slice of reality look for plenty of top minds in NBA analytics to continue to evolve and refine the stat. xRAPM is just the latest publicly visible refinement. Development isn't stopping. And its moving forward in both hockey and basketball.
wasn't Gasol angry about getting benched last night or something? maybe it was Fizdale's final shot at him knowing his time was up?
76ers v. Cavs Tonight there are many ways to measure "strength of schedule". whatever measurement you use the 76ers have had a tough schedule so far this year. And they are 11-7. this only adds to the intrigue of tonight's game with Cleveland.
On November 28 2017 11:52 JimmiC wrote: When your the best player for your entire life because you are an abolute athletic freak. Everyone loves you, many worship you. Then in your mid 20s that goes away and now your getting so much hate. Look how it fucked with durrant and he can still dominate on the court!
Is Durant already over the hill of his athletic peak?
On November 28 2017 11:33 urmomdresslikafloozy wrote: Derrick Rose reminds me of Andrew Bynum, a player ahead of his time and not disciplined enough to maintain long term.
Derrick Rose is the Bobby Orr of the 21st Century. Both players played so hard and rode their bodies so hard that long term success was not physically possible. No one criticized Orr for "not being disciplined enough". They marvelled at his willingness to "put it all on the line". It's all a matter of perspective.
Obviously, Orr was better.. but you get the idea.
On November 28 2017 09:44 Twinkle Toes wrote: JJ im intrigued now, say prime Wade and LBJ play together almost together at the same time, how do you cimpute RAPM for them?
ESPN Real PLus Minus isn't good at predicting how well players mesh together. The stat is good at assessing who is responsible for the good things and bad things that a team's offense/defense does. I'll provide an example currently relevent to this year.
What Is Wrong With Minnesota's Defense? Currently, Minnesota is 25th in Team Defense. They are 25th in points surrendered per game and 25th in points given up per 100 possessions. No matter how you measure it.. Minnesota's team defense ain't good. So the question is... WHO IS TO BLAME? Minnie's starting line up is Teague, Butler, Wiggins, Gibson. and Towns. Thibs plays his starters a lot. So who is fucking up here?
Let's mosey on over to the ESPN Real Plus Minus Stats page and check it out. Teague is ~15th amongst starting point guards in Defensive Real PLus Minus. He is not the problem. He is average. As usual. Butler is 3rd amongst starting shooting guards in Defensive Real Plus Minus. He is very good. As usual. Wiggins is ~20th amongst starting small forward in Defensive Real Plus Minus. He is somewhat below average. Being somewhat below average can't destroy a whole team's defense. Gibson is 7th amongst starting power forwards in ESPN Defensive Real Plus Minus. He is a very good defender this year.
Towns is dead last amongst starting NBA Centers in ESPN Defensive Real plus minus. Towns is THE WORST CENTER IN THE LEAGUE ON DEFENSE BY A SUBSTANTIAL MARGIN
Towns is fucking the entire Timberwolves defense into the ground; if you listen carefully to hardcore Timberwolf fans he is quickly becoming the whipping boy for all that is wrong with Minnesota's defense. This is based on many subjective/anecdotal observations. Some of these observations are by smart people who knwo what they are talking about. i am not diminishing their perspective. What i am saying is ESPN Defensive Real Plus Minus provides some objective statistical evidence to the theory that Towns is letting the team down on defense.
Wiggins is not the problem this year. His effort level is much higher than previous years. Wiggins ain't great... but he is no longer the total dog fucker he was in past years. Just watch a few Minnesota games and you'll notice the difference.
I wonder if Jimmy Butler will begin to call out Towns by name. Butler has said "our defense is bad" without naming names a couple of times. I wonder how long it will be before Butler points the finger directly at Towns. The Minnesota hard core fan base is already pointing their collective fingers at Towns.
What is wrong with Minnesota's Defense? Karl Anthony Towns is what is wrong.
In other news, Blake Griffin's knee being bent sideways has LCL tear/sprain written all over it. No official word yet. I guess they'll know for certain after an MRI is done. However, if its a complete tear a DR can diagnose that without an MRI being done.
Wow it really doesnt seem like that at all based on how Towns plays. Can you prove it? Can it be proven using RAPM or other advanced stats that you know of?
On November 28 2017 22:50 JimmiC wrote: I struggle with your observations (and how you always think effort is what makes people improve).
effort is a necessary but not sufficient condition for improvement. zero effort and you can not improve no matter what you do. in Wiggins particular case it has helped his defense.
On November 28 2017 22:50 JimmiC wrote: Wiggins effort is not noticibly better this year
i disagree... and i'm not the only one noticing the difference in Wiggins' effort level.
On November 28 2017 23:13 JimmiC wrote: That data is not effort, that you believe that makes discussion impossible. I tried to slowly walk you through that with kyrie, zenith tried with cp3/jordan. Many have tried many have failed. You burnt me out time to go back to ignoring you.
that is a good move because you don't have anything to back your points.
i'm not posting the other stuff plenty of others like Paul Pierce and Jalen Rose have noticed about his lack of effort on defense in previous years. there is a mountain of that stuff out there.
What i've seen when i watch him play in previous years is congruent with what knowledgable people have said about Wiggins' lack of effort. It then shows itself statistically as he was sleep walking on defense in the 2016-2017 season.
EPT
