Researching Robustly for a Post Takes A Long Time

Sorry about the lack of posting.But my twitter account got suspended.

I’ve been working on the twitter thing instead of some of my ideas for posts. But the twitter thing is taxing. Twitter automates the suspension of accounts but uses REAL people to resolve suspension disputes.

That means I’m probably in a giant queue of DMV-like doom. I couldn’t read basketball tweets about the NBA Finals, and I’ll probably miss the NBA draft too.

I could make a new account but I’m really tired. And angry. But mostly tired. Instead, I’m gonna build up some long range articles and wait for the twitter account to get unfrozen. Also, my RSS Reader (Reeder) got entirely messed up due to a Google Reader change. Linking to articles is 10x more difficult without RSS searching out the name for me. I’ll link more when it comes back online. You’ll just have to take my word, for now.

Here’s some articles I’m working on. Some of these are long term projects.

1. Durant and Harden. Who’s better?

This will actually be a positional comparison. Using the Wins Produced formula (See Research), I’ll assign a percentage value. For instance, Durant is great at scoring so what percentage of scoring efficiency affects Durant’s wp48 score.

Let’s say Durant averages 2x as good as Harden in scoring, and let’s say scoring is 20% of wins produced. And let’s say Durant hypothetically does everything else average. Harden scores 10 on avg while Durant scores 20. So 20 is 20% of 100.

Durant 20/80

Harden 10/40

He would have to be twice as good in rebounds over Durant to catch up. Etc Etc. Or a mixture thereof. I’m still fiddling with the math. Most stats have a limiter. Scoring Per Possession. Steals and Rebounds Over Personal Fouls (What % of Personal Fouls come from Defense? Probably a a majority. So research that coefficient or just do an average of the personal fouls collected by the top 30 ball thieves. I’m thinking taking charges or fouling on purpose is a minor part of a person’s game. Should not be a significant coefficient. .)

Hopefully by the end, I’ll have some thresholds. How much scoring per possession is enough to counteract being deficient in everything else? Technically, you don’t even need Harden and Durant for these questions/answers but the idea originated from that quandary. (Wagesofwin Podcast Link)

2. On The Different Statistical Models

Basketball Prospectus’ Kevin Pelton uses WARP.

Wagesofwins’ dberri uses Wins Produced.

Basketball Reference uses Win Shares.

What is the difference between these 3? Have they attacked their own models to see what their prediction rates are?

I was reading Ed Weiland the other day and he has a theory that high 2PP% and a high STL count should be the main factors in determining a PG draft prospect. I think Ian Levy did a thing where he implies that Scoring skills don’t translate well at all from the NCAA to the NBA. That fits with common perception but I’d like Levy to actually pump out the data. I’d LIKE to see his NBA Skill Transfer Analysis done on a per position basis. Maybe shooting guards have a higher percentage of their FT% or 3PP% transfer over to the NBA. My own small dataset shows that most good FT% shooters have good 3PP% in the NBA/NCAA. It’ll be wild if there was a correlation between NCAA FT% and NBA 3PP%. (But I’m not gonna say it because I’d like to get the code crunched through at least 12 years of Data (3 4-yr draft cycles. It’s important not to include the rookie year. Arturo did a data crunch showing that 9 out of 200 top NBAers have had a good rookie year. Or some ridiculously low number. Don’t quote me.) .

I really don’t think Ed Weiland has tested against his own theory robustly. I mean, he got Lin right but then again, so did Wins Produced. People sometimes get so in love with their idea they don’t want to do the hard work of breaking it down. Like PER. Oh god, I hate PER.

3. What are the biggest gambles in the draft?

I’d like to do this one first. Hopefully before the draft. Drummond, Eli Holman, Hummel etc.

The methodology would be to use Ian Levy’s ( Similarity Score method to see who these people are similar to. I will be taking fg% and all shooting % out of the equation. This is a pure leap of faith that shooting doesn’t translate at all in the NBA with the notable exception of FT%. This work has also been done by Ian Levy.

4. How good is your team’s Head Draft Guy?

This one is the furthest along as I’m only doing the Lottery Teams. What is the average win score or wins produced per draft pick they’ve chosen? I’ve already done some of this in the Research section. Note to all: stats use winshares. The results are a bit wonky with winshares. I want to do one with the more familiar Wins Produced but that requires more work as I’d have to calc it myself.

The purpose would be to predict the player a GM would most likely pick. A GM who normally picks Bench-level players would probably pick another Bench-level player. The problem with this approach is that the sample sizes for many of these lottery teams are small. Neil Olshey has 0 draft history with the Blazers but he has a few with the Clippers. There’s a lot of wikipediaing for “Was Olshey responsible for 200X’s draft for the X team?” Hard to code for that kind of logic. Time intensive, which is why I plan to do it for just the Lottery teams. Will probably do a top tier team like the spurs and the thunder for control purposes.

Basically, the mission of this blog is to expand on other people’s work.

***P.S. How sick was that Heat/Thunder game!

***PPS How sick was Mark Cuban’s interview on First Take! I wonder if Thunder would have won if they played zone. Lebron’s outside the paint jumper was 16% or some such. Then again, Lebron barely took any of those. He either shot 3s or went to the hoop (rmdr: Link to Ken’s shot chart). Dude is a monster.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: