This topic contains 8 replies, has 6 voices, and was last updated by Levy2725 12 years, 10 months ago.
- AuthorPosts
- Posted on: Mon, 05/16/2011 - 3:55pm #29018
Levy2725ParticipantHey Everyone
I write for Hickory-High and The Two Man Game. I wanted to share a new feature I have going at Hickory-High. I’ve built a system for making statistical comparisons between draft prospects and players drafted from other seasons. The system produces similarity scores, something Basketball-Reference and others, have used in the past. The players are compared across a number of categories and a score is generated on a scale of 0-1000, 1000 being a perfect match. For example, Kemba Walker’s closest comparables were Troy Bell-892, Eric Maynor-884, and Devin Harris-882.
It’s similar to systems that Kevin Pelton and John Hollinger use, although mine is obviously a lot less refined. I believe Hollinger’s system is a straight projection of NBA performance. Pelton’s system compares college players to NBA players to try and predict a career arc. Mine is simply a snapshot of a moment in time: Player A’s college production most closely resembles Player B’s. It doesn’t account for things like potential, personality, athleticism, etc.; other than how they manifest in a player’s statistical production. For that reason it’s a little limited in projecting how successful a player may be. I also haven’t figured out how to convert Euroleague stats, so at this time it just covers college players.
I have a table of contents page set-up with an explanation of the whole system, link: http://www.hickory-high.com/?page_id=1740 All the results are finished and can be accessed in a spreadsheet from that page. I’ve also built individual pages for all of the 40 prospects I included this season. Each page has links to a few profiles and then a table showing the 10 closest comparisons. Hopefully this can be an interesting tool and discussion starter. I’d love to hear feedback if anyone has thoughts!
0 - Posted on: Mon, 05/16/2011 - 4:05pm #532418
mikeyvthedonParticipantI mean, it is kind of cool, but I do not know whether it necessarily translates into what to expect from these prospects at the NBA level. Hollinger’s tells you next to nothing about possible NBA production, and I do not know about the other guy, but it was cool to see whose statistics these prospects come close too. Nonetheless, good work, and it is definitely something different. If only your system could use other intangibles, than it could be a thing of beauty, but you can only get so much off of raw statistical analysis. Also, do you use pace adjustment? Just wondering.
0 - Posted on: Mon, 05/16/2011 - 4:12pm #532425
providencefriars1ParticipantWhy did you snub Marshon Brooks from your site? You have guys who will be picked all around him like Cole, Thomas, and Lighty.
0- Posted on: Wed, 05/18/2011 - 3:10am #532761
Levy2725ParticipantMarshon Brooks page is up. Closest comparison was James Anderson from Oklahoma State, drafted by the Spurs last year.
0
- Posted on: Mon, 05/16/2011 - 5:16pm #532451
JJeff6ParticipantI actually like that, keep up the good work, its entertaining.
0 - Posted on: Tue, 05/17/2011 - 1:27am #532491
Levy2725ParticipantThanks for the positive comments everyone. All the stats I use are pace adjusted. Again, there are lots of limitations to the system and certainly shouldn’t be used independently of subjective evaluations. But it seems like there are plenty of holes in primarily subjective evaluations as well. Perhaps the combination of the two can make for a more powerful tool. Of cource like you pointed out, only so much can be gleaned from statistical analysis. Statistical production at the college level is not always, or maybe even usually, the best indicator of future success.
As far as Marshon Brooks, this may not be the best forum to admit this in, but I use Draftexpress a little bit more frequently. I started with the top 40 collegiate prospects as listed by Draftexpress. Marshon Brooks would have been the 41st. I’ll be happy to add him though. Check back in a day or two and I’ll build a page for him!
0 - Posted on: Tue, 05/17/2011 - 3:20am #532518
M-DYMESParticipantVery cool, however it does not account for the way in which players get those stats. For example Alec Burks could average the same amount of points and assists as a guy like Ray Allen, but his game is completely different. He is a driving force and Allen was a straight shooter.
Still it is really interesting and awesome you took the time to make something like this. Mad props.
0- Posted on: Tue, 05/17/2011 - 4:49am #532529
Levy2725ParticipantYeah, unfortunately with the stats that are available it’s difficult to capture too much of a player’s tendencies. The system contains the "how much" but not the "how." That’s why you get Derrick Williams’ first comparable being Ike Diogu. I think in some cases though the strange comparisons add some information. If a power forward’s comparables are mostly shooting guards it should give you at least some information about their style of play.
The other thing I’ve found in looking through it is how much NBA success colors our perceptions of a player’s past achievements. Kareem Rush and Ike Diogu weren’t tremendous prospects and have had disappointing NBA careers. So it’s easy to forget how incredibly productive they were in college. Derrick Williams being compared to Diogu may not bode well for his NBA future, but is should speak highly of what he’s accomplished in college.
0
- Posted on: Tue, 05/17/2011 - 5:39am #532535
paradigmnParticipantGREAT POST by the way… Since these guys have been compared to each other thier whole lives rightfully so, its interesting to see the different players that each one is compared to…notable names markieff has that marcus doesnt is charlie villanueva and al horford while marcus has kevin love and nick collison…they both have bosh but he compares to markieff more…with projecting players careers being such an un-exact science…any tool such as this one can atleast provide information that can possibly be helpful for GM’s making borderline decisions.
0 - AuthorPosts
You must be logged in to reply to this topic. | Login |