Crow wrote:
How would you say your Wins Added metric varies most significantly from other major media metrics? Which one is it most like?
I feel like it kills high turnover guys more than most metrics do. At some point I'll compare it to others just to see which it varies most with, but I'd wager it's closer to BPM than RPM.
Crow wrote:
Do you use weights on your comp system inputs? If so, what gets the most weight? Do use height as a category? Do you use an aging curve? Your own or a published one?
I do use weights with most weight going to TS%, Minutes, Usage, and Position. Don't use height or weight, though I'm planning on adding those in the future. Yes to an aging curve, I use the one the Basketball Reference Simple Projection System outlines.
Look forward to more comps from your system of pros or draftees or young pros / recent draftees. Are players actually turning out similar to their draft comps after a few years? If similar players are changing from comp system and / or conventional wisdom in similar ways that would be news to ponder & try to use. Which comp change path unleashed the biggest change in value over draft expectations? Aggregating data above single player results. And / or aggregating by selecting team. If a player doesn't change, you'd hope most of the insiders would guess decently on player outcome. If they do change similar / type, I'd expect the outcome will vary more by quality of organization or part of it.
Which rookie of year or top 3 finalist in last 10 years is Saric most similar too? Limited this way is he closer to Zingis, Mirotic, Faried, B Lopez, Scola, Bargnani or Rudy Gay (or other of different height)? The likely vote of him as at least a top 3 ROY finalist (and presumably using eye test / subjectively selective stats more than pure roll-up analytic metrics) would give him a different set of comps. Roughly equal or higher than the similarity systems? Probably higher. Which will look more right in retrospect in a few years? And is the answer for this for Saric the norm or the exception? If you did this for all such top 3 guys is the average ws/48 for their top 7 comps more different by the two approaches for Saric than typical for the others? Could be worthwhile to check. Who (and what types) did the systems do better and worse on than their own other work and compared to the eyes & selective stats / non-roll-up guys? Who did those guys do better & worse on? Random or uncovering of potential bias / error patterns to try to learn from?
Top 5:
Player SimScore Sim Percentile
------ ------ ------
Andrea Bargnani 0.256 .978
Rudy Gay 0.258 .976
Nikola Mirotic 0.268 .950
Kristaps Porzingis 0.274 .933
Blake Griffin 0.277 .925
Others of Note:
Player SimScore Sim Percentile
------ ------ ------
Brook Lopez 0.328 .744
Kenneth Faried 0.345 .682
Luis Scola 0.362 .620
These numbers seem to all make sense to me, maybe outside of Blake being so similar. I think if I projected forward based on this listing Saric would certainly be better, but I'm not sure it would be by a crazy amount or anything. This may be a way of thinking about a realistic ceiling for him more than expectations.
If any object to my use of the phrase subjectively selective users of discrete stats along with eye test because you feel that one can as comprehensive and accurate (or more so) than system metrics you are imo arguing for an invisible mental analytic & summation system superior to the enumerated ones. Can you be more sophisticated and successful with a system that isn't written down and that you can't prove has been consistently and fully applied? Maybe if you believe in expert intuition. Me? I have higher doubts of the invisible over the visible. At least until shown results suggesting otherwise.
So Saric's best comps in your system from my list drawn from the ROY voter selections (and considering size) are Faried and Scola (best)? That might be considered disappointing by some or many. I am guessing (but don't know) that the average ws/48 of all top 3 ROY candidates in the period is higher than for these 7 or for those 2. But maybe I biased it by excluding a few name bigs that appeared vastly better Cousins, Davis, Griffin) and thus not good comps to me. So fwiw, how good were those comps? Really lousy or better than that?
The main reasons for keeping a system invisible instead enumerated would seem to be to me:
1) it is not complete
2) it is not clearly weighted
3) it is not fully understood
4) it is being shielded from critique
5) it is not being consistently or fully applied
6) the owner reserves the right to change any of it for any reason or momentarily more prominent factor or whim
7) the owner believes in non-systematic analytics or summation over the opposite because of "experience" (unmeasured systematically probably) or just belief.
But I should remind myself that it doesn't have to be either / or. It can be both. It probably should be both enumerated and interior / invisible. Both may offer insights. Both could be tricky. More so one to me, maybe the opposite for you. The opposite of both is total rejection of one. I might sound that way and maybe sometimes I am that way but on my better days I am an advocate of both (but not necessarily simply even).
Apologies for not being clear, best comps have lowest sim score in my system. So Bargani, Gay, Mirotic, Zingis are closest out of the recent ROY top 3 finishers.
Besides Griffin being a decent comp, Davis and Cousins rated as pretty poor comps. Cousins was in the 87th percentile for similarity and Davis was in the 71st, so neither would seem to be a good indicator of where Saric is going to go from here.
I think expectations for him need to be lowered a bit. He may win ROY, but I think it would ultimately be similar to the Michael Carter-Williams ROY award where they just kind of had to give it to someone and he had the most counting stats. Saric is probably a better player than MCW, but I'd say he's a good role player in all likelihood and not a future starter.