APBRmetrics

Hi all, relatively new here. With the NBA Draft coming up and Kevin Pelton writing a new article about the topic, I was inspired to do some work on NBA draft pick trade values. What I did was code a system that allows for inputting any number of trade value charts. You can then view them in detail - the raw numbers, normalized numbers, and corresponding line graph. From there I wrote some more code that lets you compare existing charts (again both with numbers and graphs). I also created a trade simulator so that you can simulate trading picks from within the same draft, and it will tell you if you come out ahead or behind. That trade simulator lets you run it against any number of trade value charts in the system as well. Here is the page:

http://nbasense.jayroman.com/draft-pick-trade-value/

I added both of Kevin Pelton's trade value charts from 2015 and 2017, and I also added Nick Restifo's chart from 2016. I looked for more charts to input but didn't really find any, as a lot of others were raw analysis of the value of specific picks over time rather than trade value. If anyone here has their own trade value chart let me know and I can work on adding it in.

I looked at the 2017 file. I checked to see what picks 5 and 10 would get. The only thing the model would accept was the #1 pick. That doesn't seem to reflect the market. Is the market "wrong" or is the model or both?

I traded 20, 30 and 40. I had to receive #10 for it to work. That might be realistic but it might be a little high on the receive end, at first look.

Thoughts?

BRef has expected value for this class here: http://www.basketball-reference.com/draft/preview.html

I have my own draft pick trade value that I use, you're welcome to add it if you'd like: Pick Trade Value

Crow wrote:I looked at the 2017 file. I checked to see what picks 5 and 10 would get. The only thing the model would accept was the #1 pick. That doesn't seem to reflect the market. Is the market "wrong" or is the model or both?

I traded 20, 30 and 40. I had to receive #10 for it to work. That might be realistic but it might be a little high on the receive end, at first look.

Thoughts?

I think the market overvalues high draft picks. I'd trade 3 for 5 and 10 as soon as it was offered in an average draft

Crow wrote:I looked at the 2017 file. I checked to see what picks 5 and 10 would get. The only thing the model would accept was the #1 pick. That doesn't seem to reflect the market. Is the market "wrong" or is the model or both?

I traded 20, 30 and 40. I had to receive #10 for it to work. That might be realistic but it might be a little high on the receive end, at first look.

Thoughts?

The 2017 model was from Kevin Pelton who readjusted his trade value chart this year. I'm not sure if it was meant to reflect the market, but I know it was not meant to reflect this draft specifically. It was just a way to signify that he made a new model this year. It could have also said Version 2 instead of 2017 Version.

jgoldstein34 wrote:I have my own draft pick trade value that I use, you're welcome to add it if you'd like: Pick Trade Value

Great! I added it in, you can view it here: http://nbasense.jayroman.com/draft-pick ... oldstein-4
I can also provide the direct link to your Google Doc or another URL if you'd like.

I multiplied your values by 100 since my database is set to use integer values for each pick. You can click on Compare Trade Value Charts and select the other ones to show a graphed comparison, like this:

Very cool, thanks.

Odd that my model views 2nd rounders as so much more valuable than the others, I'll have to dig a bit more to try and figure out why.

No problem. Yeah it's definitely interesting to see that; if you dig into it and find out where that comes from I'd be interested to know.

So...no disagreement at #1 , less at #10-12 than elsewhere but a fair amount everywhere else in the top 30?

Well by the data's nature I normalized everything, so everyone's #1 pick is going to have the maximum value. But yeah quite a disparity in the 1st round!

Found the issue and corrected it in the google doc I linked before. Had a weird outlier late in the 2nd ruining my curve fitting haha.

jgoldstein34 wrote:Found the issue and corrected it in the google doc I linked before. Had a weird outlier late in the 2nd ruining my curve fitting haha.

I updated your values. Check out the new chart - there's definitely a difference from your old one.

Yeah the 2nd round was getting pulled up because I typed a number in wrong lol, for sure better now. Thanks!

I've been intrigued by draft value charts for a while now. I've played around with creating my own, but I always come back to the same question. What are we trying to measure?

I think what I'm really looking for is two charts (or maybe the synthesis of two charts). One that measures expected minutes played. Obviously there's some bias to deal with here, but I think that we want to know how much we expect guys to be on the court. The second is a rate of talent, not the total "wins" created by that player over X years.

Am I making any sense?