rand wrote:Blame Rasho wrote:rand wrote:Bonner led the Spurs in On/Off just once, in 2009, when he had his career best season across the board. His total On/Off with the Spurs in 10 seasons was +1.9
However, if a player of Bonner's stature really was leading his team in On/Off for several years, I would think it an indication that said player was underrated and underutilized rather than the stat being a horrible metric. Think about it: if for several consecutive years across all sorts of roster and rotation changes, a given team has its highest net rating when one particular player is on the court, it really should say great things about how effective that player was (given an adequate volume of minutes).
In small samples or with uncommonly unbalanced rotations, I agree On/Off can throw wild results such that in any given year it can have a lot of noise but over the long run I think it gives very good indications of effectiveness.
I will copy and paste a discussion on this…Blame Rasho wrote:
I remember Popovich going on an amusing semi-rant about this many years ago. Basically how all the nerds in the organization were badgering him to give Matt Bonner as much PT as possible because their lineup data was off the charts when he was on the court. And yet from his perspective, all he could see was him getting his ass kicked on defense while making some 3s here and there and never passing so he said he eventually threw all that stuff in the trash and told them to stay out of his office. And then they'd try to point out that Manu also rated incredibly high in order to bolster their argument, and he'd be like yeah, but I don't need some guy with 10 degrees from Harvard to tell me that.
You have to put context in any stat esp this one.
Let’s put this in the big picture, any stat that says Matt Bonner is more important than Tim Duncan should immediately be thrown out.
I don't know of any all-in-one stat which doesn't routinely return some results which are at variance with common sense. EPM is justly respected but this year it says Zubac was better than AD, Garland was better than Ant, Pritchard was better than Brunson, Cam Johnson was better than LeBron, DFS was better than Banchero, Quickley was better than Booker, Kris Dunn was better than Trae, etc, etc, etc. This basis can be used to discredit all all-in-one stats.
On/Off needs to be used with more caution and context than most all-in-one stats because it's totally raw (RAPM is obviously a better +/-) but that doesn't make it horrible or useless. Again, if for years as you initially mistakenly claimed, the Spurs had a higher net rating when Bonner was on the court than off, with all of the roster and rotation changes the team went through, it should indicate that Bonner's true value was beyond his stature.
The biggest issue with RAPM based metrics is that they're misleading.
If a player is +1 and another player is +1.1, we have a tendency to say well player B is better than A per this stat. But...that's not exactly true. There's a margin of error. These metrics are looking for a best fit but there are countless possible best fits. They're just all within a reasonable range. So +1 and +1.1 are basically the same thing.
That's before we get into roles and minutes and all that.