Berkshire continuously makes the same mistake in his analyses -- he does not know what a covariance is. When using statistics, you cannot just linearly combine 100 different stats as he did, as many statistics vary together.
I have known that Berkshire does this for years, and sure enough I found it in his article.
For example, he is including both corsi and shot attempts as separate inputs into his model. You cannot do that as you end up including the sane information twice. He does this over and over.
It's a cute little exercise that meets the standards of a blog needing hits, but it would be rejected if he submitted it to a statistics-heavy journal. It would also lose points as an undergraduate homework problem -- covariances are basic stuff.
But is he equally applying this to each defenseman? So in effect, whatever approach he takes, it's applied equally? Personally, I would like to see constructive criticism. Do you have the time to do your homework and provide your top 20 guys with your own system? I wonder how much it would vary from his rankings?
It's not like he has garbage in his top 20. Doughty, Subban, Karlsson are the top 3. What guys don't belong? I know I would have Hedman as my #1 guy but that is just the eye test and watching games with no calculated measure.
I actually have no problem with his evaluation. Why 3 years? Not sure. But he clearly breaks down his approach at evaluating Offense, Transition, Defence, and Degree of Difficulty. Seems like a reasonable approach to me. If anyone wants to contest it, stick your neck out, do your own homework, and provide your own ranking and how you came up with it.
Tracking some quick stats over the last 3 years.... I tracked a few stats on Karlsson, Petry, Giordano, and Keith to compare a few. Not sure how he is ranked higher than Keith for example but Petry's stats hold their own. Don't have time to do each player.
1. Karlsson
- 230 games, 0.93 pt/game, 0.31 PP pt/game
- 2.88 shots per game, 0.91 hits per game, 2.16 blocks per game
- 1.25 giveaways per game, 0.76 takeaways per game
- 27.51 TOI, 3.88 PP-TOI, 1.64 SH-TOI
- (-) 17
2. Doughty
3. Subban
4. Giordano
- 245 games, 0.54 pt/game, 0.18 PP pt/game
- 2.36 shots per game, 0.96 hits per game, 2.20 blocks per game
- 0.64 giveaways per game, 0.57 takeaways per game
- 24.39 TOI, 2.97 PP-TOI, 2.91 SH-TOI
- (+) 17
5. Klingberg
6. Fowler
7. Burns
8. Jones
9. Josi
10. OEL
11. Pietrangelo
12. Hedman
13. Petry
- 213 games, 0.40 pt/game, 0.16 PP pt/game
- 2.10 shots per game, 1.99 hits per game, 1.67 blocks per game
- 1.19 giveaways per game, 0.36 takeaways per game
- 22.71 TOI, 2.68 PP-TOI, 2.21 SH-TOI
- (-) 23
14. Letang
15. Keith
- 229 games, 0.56 pt/game, 0.19 PP pt/game
- 2.18 shots per game, 0.35 hits per game, 1.69 blocks per game
- 0.84 giveaways per game, 0.38 takeaways per game
- 24.89 TOI, 2.86 PP-TOI, 2.52 SH-TOI
- (+) 6
16. Werenski
17. Byfuglien
18. Lindholm
19. Barrie
20. Parayko