It tells us that Jones was a +3.16 when you take into account the following factors: The base equation in this formula is a linear regression with the output being On Court Net Rating (ORTG-DRTG) for each player and the inputs being: TS% Opponents TS% TOV % Opponents TOV% OREB% DREB% Just because Jones STUNK in the playoffs does not mean he STUNK in the regular season. He was actually very, very good before the Faried knee to the ribs. He even played defence Check out many different ways of measuring on court production, and you will see that he was much better than one man with multiple accounts, who posts a lot on each one, would have you believe. When he was on the court, the team scored at 52.5% from the field (50.8% off), and our opponents only scored at 47.3% (49% when off). [ http://www.82games.com/1415/14HOU14.HTM ] His man scored at just 41.5% on the season, and a miserly 49% from in close (Dwight is good at 55%) I know the trendy thing to say around here is 'ditch Jones, yo' - and if he is a disappearing act in the playoffs, then YES, do it. But I'm all for giving him one last shot - his numbers dipped considerably after the rib injury last season - if he plays this season like he did prior to that injury, we got ourselves a keeper.
It tells me that the numbers that went into computing this stat probably overrate Terrence Jones. Which, of course, isn't to say that the stat is useless. Any stat relying on incomplete information to approximate a player's "impact" will overrate some players and underrate others. That's why, in the end, its just one thing to look at and perhaps consider, but it shouldn't be treated as an end-all, be-all stat. That goes for any Adjusted +/- metric, however sophisticated it might appear to be.
Players more impactful than James Harden: BOLD are the serious LOLZ Khris Middleton Michael Carter-Williams LeBron James Iman Shumpert Timofey Mozgov Kyrie Irving Jonas Jerebko Isaiah Thomas Chris Paul DeAndre Jordan Blake Griffin JJ Reddick Matt Barnes Tony Allen Kyle Korver Thabo Sefolosha Michael Kidd-Gilchrist Elijah Millsap Darren Collison Corey Jefferson George Hill Anthony Davis Reggie Jackson Tyler Hansbrough Dwight Howard Kawhi Leonard Tiago Splitter Danny Green Kevin Durant Ricky Rubio Wesley Matthews Steph Curry Draymond Green Klay Thompson Andrew Bogut Andre Iguodala
No stat will be end-all, be-all. But no matter what kind of stat, at least it should serve a purpose, which is to explain a point with some facts backing it up. What is the point here? Is this stat's sole purpose is to explain Dwight and Harden are better than Black, Dorsey and Papa? Is this the point? If not, then when we look closer for the middle of the group, and as you said : "the numbers that went into computing this stat probably overrate Terrence Jones", then how do we know the numbers are not overrating Josh Smith and under-rating Bev? This is called adjusted stat, it kind of give me people a perception that it is more advanced. But just from these number, what exactly you can learn? Comparing it to a more simple fact : Both Terry and Bev's assist number aren't that good in the overall ranking of the PG position in the league, hence we know the Rox are having play making issue. When Dwight was out, we are more than often getting out rebounded by out opp., hence we can know how impactful Dwight was for us on the board and how bad our other big guys are in this same area. Again, these kind of simple stats are nothing advanced, and of course it won't be en-all, be-all, but at least they serve a purpose.
Very loose imprecise statistical models such as this one, which do not account for a variety of variables (lineups and schedule are probably the biggest ones) need to be taken with a handful of salt. Statistics do not lie. It is after all just a branch of mathematics. However statisticians and others frame statistical analysis in inappropriate ways. For instance the metric cited by the OP probably does not have a confidence interval on the analysis because the variables around basketball performance have many difficult to quantify variables. But, honestly lineups and schedule are relatively easy to quantify.
That DMo/TJones stat makes sense to me. Jones played way more minutes with Dwight. Dwight makes the game seem easy.
DMo played most of season without Dwight & Jones / Smith. Also he decresead his intensity level during his last month (fatigued, summer), until the injury happened.
All I know is that McHale better not play guys so many heavy minutes and make sure we are healthy for the playoffs.
The point of these kind of stats is to help predict whether adding certain players into a team's rotation might make the team better or worse. I can't comment on the accuracy of this stat, in that regard. But that would be the main goal.
I definitely don't think this stat is some great stat that blows other things like RAPM out of the water, but instead it is just another piece of information to look at in conjunction with other stats. The best thing about my stat imo is its ability to find underrated role players. Zaza Pachulia is the best example that comes to mind right now. Even RAPM only had him at +.6, but when asking Bucks fans what they think of him, they didn't blink twice at his +5.42 AIR because they can tell from watching him that he is a big difference maker even if the usual stats don't acknowledge it. Thabo Sefolosha is another player who scored really highly (+7.58) without receiving a ton of credit in the media for his play, although RAPM does like him. Comparing across teams probably isn't a great idea unless players play a similar role on similarly effective teams. Things to look for are players who scored much higher than the players they played a lot of minutes with. Robert Covington only scored a +1.11, which wouldn't be all that impressive on the Rockets, but still a positive impact player. On the Sixers though, he was able to score much higher than most of the team while shouldering a huge burden offensively, which is extremely impressive. Its not a sure thing, but I'd wager that on a good team with a smaller role, he would rate much better than the 1.11 he scored. Tony Allen crushed the rest of the Grizzlies in AIR, but I don't think he would be all that successful in a different system because most offenses aren't equipped to be successful with a non-shooting 2-guard. Speaking of Allen, this is a regular season stat, and playoff basketball is much different that regular season basketball. As for what you can learn from this stat, it tells you about how well the team does with that player on the court. Looking at DeMarre Carroll, I'm not sure how well he will be playing in Toronto. He is likely gonna have a much bigger role, and he was already one of the weaker starters for Toronto by AIR (+3.03, lowest starter, 6th on team) and giving him a bigger role probably will reduce his effectiveness. I could be wrong and he might have a great season there blossoming in his new, bigger role, or it could turn out that a good 3D player is about as good as he gets. When looking at free agents or trade targets, you can identify which role you need to find, and then look for the player who filled that type of role the best (you could look at raw AIR or see which players scored highest relative to the rest of their team in the case of finding a valuable role player on a subpar team)
Those stats seem about right, couple of oddities towards the middle but it's generally pretty accurate IMO.
Not necessarily, you need them but you have to be able to see and know basketball. Doesn't tell the entire story....like science:grin:
When you try to simplify something as complex as a player's productivity on both sides of the court with one number, you are bound to miss out on things. These ratings aren't for GM's and coaches; they are for casual fans that don't have the time, skill or access to mine the data. Do you think Berkshire Hathaway looks only at revenue and profit when determining acquisitions? The only aggregate numbers that matter in the NBA are scores, season records, and playoff series records. For anything else usefule, you need to dig deeper than people are willing or able to.
I thought about separating it into offense and defense, but that would only leave 3 X inputs for each, which is a little bit low for my liking. Also a GM could definitely use this as another piece of information when evaluating numbers. Do you know of any other metrics that rate Zaza so well?
...what's the real value in this stat? If it doesn't allow us to compare players and instead is derived solely on how well the team does when that player plays, what are you suppose to use this for - to determine optimum lineups? Actually I guess you couldn't use it to determine an optimum lineup because it says that a person's rating is influenced by whether they play with another really good player - so James harden's backup would not get the benefit of the Harden bump as this is based on team metrics....so what is this used for again? I don't get it.
Identifying undervalued role players is the most important part of it. This stat will expose which players actually have the team play better with them on the court, and not just because they played with the other starters all the time. If a team is looking for a 3&D player off the bench, they can be pretty certain they will get good returns if they have Thabo Sefolosha in that role. Its not unlike RAPM in a good amount of results, but it isn't a system of equations like RAPM is.
