Wednesday, 5 August 2015

Draft value of NBA players (improved analysis)

(Originally posted on Sunday, 24 July 2016; updated on Thursday, 15 June 2017)

I updated all the results after the end of the 2016-2017 season.

This is an altered version of my previous post on this topic. Now I calculate the draft value a little differently. I also eliminated a minor flaw in my formula for the draft value and now the draft value of a player who played in very few games will be close to zero, no matter how well he played in those games.

In this post:
How to compare NBA players (improved analysis)
I described a way to compare NBA players. To compare different NBA drafts I used the values of basketball statistics I had described there.

To compare NBA drafts I needed to gather data for over 2000 players, so at this point I would like to thank Mike from the site
http://www.basketball-reference.com/,
who answered my plea for more precise NBA draft data by explaining to me how I could gather the required data in a very easy way. Thank You!

Mike explained to me that I can “filter by draft year for searches in the player season finder and can set the search for career stats”, with this example:
http://bkref.com/tiny/ifje3.

Now to the point of my post. I decided to gather the data from the last 41 NBA drafts (1976-2016), because the year 1976 was the year when NBA merged with the ABA and the players from that draft were the first ones to play their whole NBA career against all the best players in the world. Obviously the most recent drafts will be properly judged in about 20 years.

The main problem I had to tackle was the definition of the “draft value”. People can understand it in many different ways. One way of calculating the draft value is to judge the number of awards earned by the players from a particular draft, but not every MVP (and not every all-star) was equally good.

The only value that can be easily understood on its own is a per-game average, so I decided that I will calculate the draft value in a similar way.

The issue I had to address was the fact that some players played for only 10 years and some other played for over 20 years. The players who played for only 10 years were on average much younger when they ended their careers, so they had an easier time to post high career averages, comparing to the players who played over 20 years.

It's clear that the “star-players” who played more seasons were more valuable to their teams and overall to the league. Wouldn't it be better if Michael Jordan played 20 full seasons instead of taking two breaks from playing? Let's analyse an example. Below there are 6 different players that, just for the sake of the example, I analyse only as scorers.

1) 28.0 PPG in 1500 games = 42.000 PTS,
2) 30.0 PPG in 1000 games = 30.000 PTS,
3) 25.0 PPG in 1000 games = 25.000 PTS,
4) 20.0 PPG in 1300 games = 26.000 PTS,
5) 25.0 PPG in 500 games = 12.500 PTS,

Obviously player-1 is better (as a scorer) than player-2. It can actually be the same player who simply played more games and his PPG average dropped down a little when he was older. Imagine Michael Jordan who played 19 seasons without major injuries.

Player-4 may have a better career PTS total than player-3, but after 1000 games his PPG career average was definitely lower than the PPG career average of payer-4. So the player-3 was better than player-4.

Player-5 should be approaching the peak of his career, but we don't know how well and how long he will be playing in the future. He should not become better than player-4, but may become better than player-3. Right now he stands out from the crowd, but he is not among the best players ever, yet.

There were two main problems connected with the precise formula for the draft value:
1. What should be the reference point for a “valuable draft pick”?
2. How should be calculated the draft value for players who are OVER the reference point?

1. Reference point for a “valuable draft pick”.
Common sense points out that star-players (with high per-game averages) should be preferred over non-star players (with lower per-game averages) who were more durable and played more seasons. On the other hand a star-player who played more seasons should be preferred over a star-player who played fewer seasons. These two aspects of a player (per-game average and numer of games played) are combined in his career totals, so the reference point should be based on career total value-points.

I decided that the reference point will be 20.000 career total value-points (overall value) – it's as if a player played in 800 games (10 seasons with only 20 games missed) and averaged 25.0 value-points per-game. In the analysed drafts there were only 35 NBA players with the overall value over 20.000, at least so far.

My formula for the draft value of players with the overall value of LESS than 20.000:

Draft value = overall value / games played * (1 + (overall value – 20000) * 1 / 20000)

The formula means that the draft value multiplier will be exactly 1 at the reference point, so the draft value will be equal to the value-points per-game average of the player. The multiplier for the players with lower overall value will be between 1 and 0.

2. Bonus for durable star-players.
Some of the star-players played in over 1400 games (Karl Malone) accumulating much more career total value-points than some other star-players who played in only 900 games (Larry Bird and Magic Johnson). Karl Malone was awesome on his own, but he should not be rewarded for his durability too much, otherwise his draft value will be unnaturally high. However he does deserve some bonus for his long and reliable career. I decided that the draft value multiplier above the reference point should be less sensitive to the overall value than below the reference point. I decided that the sensitivity above the reference point will be only 25% of the sensitivity below the reference point.

My formula for the draft value of players with the overall value of MORE than 20.000:

Draft value = overall value / games played * (1 + 0.25 * (overall value – 20000) * 1 / 20000)

The formula means that every player gets +5% of his value-points per-game average for every 4000 career total value-points above the reference point.

Let's see the draft values of the players from the example (assuming that overall value = PTS):
1) 35.7
2) 33.8
3) 26.6
4) 21.5
5) 15.6

Now, let's see the best players (according to my values of basketball statistics and my draft value formula) drafted in the last 40 years:

1. Michael Jordan: 32.3
2. Karl Malone: 31.2
3. LeBron James: 30.7
4. Hakeem Olajuwon: 28.8
5. Larry Bird: 28.3
6. Magic Johnson: 28.1
7. Charles Barkley: 27.5
8. Shaquille O'Neal: 27.4
9. David Robinson: 27.0
10. Kobe Bryant: 26.0
11. Tim Duncan: 25.8
12. Allen Iverson: 25.2
13. Kevin Garnett: 24.7
14. Chris Paul: 24.0
15. Dirk Nowitzki: 23.7
16. Patrick Ewing: 23.7
17. Clyde Drexler: 23.3
18. Dwyane Wade: 23.1
19. Dominique Wilkins: 22.8
20. Carmelo Anthony: 22.2
21. Kevin Durant: 22.1
22. John Stockton: 22.0
23. Isiah Thomas: 22.0
24. Chris Webber: 21.8
25. Dwight Howard: 21.7
26. Adrian Dantley: 21.4
27. Pau Gasol: 21.3
28. Jason Kidd: 21.2
29. Paul Pierce: 20.5
30. Russell Westbrook: 20.2

These lists make even more sense than before. Now Michael Jordan is at the top of the list, instead of Karl Malone, but Kobe Bryant and Tim Duncan are behind David Robinson. Well, any way of calculating the draft value would be more or less controversial. I believe that the case for Karl Malone as the most valuable draft pick ever is still viable – it seems to me that he is usually criminally underrated:



This is the box-score for the game:
http://www.basketball-reference.com/boxscores/199001270UTA.html

No comments:

Post a Comment