Ensemble prediction for MLB 2017 win totals

I’m working on baseball predictions, and these predictions look pretty stupid in April unless you use preseason expectations.

These ensemble win total predictions combine the markets with 4 different computer models (Joe Peta, Clay Davenport, PECOTA and Fangraphs).

1. Los Angeles Dodgers, 94.9
2. Chicago Cubs, 94.6
3. Cleveland, 92.6
4. Houston, 92.1
5. Washington, 90.5
6. Boston, 90.2
7. San Francisco, 88.1
8. New York Mets, 86.8
9. Seattle, 84.8
10. Texas, 84.5
11. Toronto, 84.4
12. St. Louis, 83.1
13. Pittsburgh, 82.6
14. New York Yankees, 81.5
15. Detroit, 81.2
16. Tampa Bay, 80.5
17. Los Angeles Angels, 80.4
18. Baltimore, 78.8
19. Colorado, 78.2
20. Miami, 77.8
21. Arizona, 77.5
22. Atlanta, 75.4
23. Minnesota, 75.1
24. Oakland, 74.9
25. Kansas City, 74.8
26. Philadelphia, 71.6
27. Milwaukee, 71.1
28. Chicago White Sox, 70.5
29. Cincinnati, 70.2
30. San Diego, 66.8

Podcast: Chase Stuart on the analytics of NFL draft picks

On this pre-draft episode, Chase Stuart, founder of Football Perspective and 2017 panelist at the Sloan Sports Analytics Conference, joins me to talk about the NFL draft. Among other topics, we discuss the following:

  • The story behind his ubiquitous NFL draft trade value calculator
  • How to use Approximate Value to access every player, even offensive linemen and defensive backs
  • The reason an NFL GM might use the Jimmy Johnson draft value chart instead of his analytics based chart
  • The surprising admission of the Patriots on why they traded for Brandin Cooks
  • Why Chase writes every day at his site Football Perspective

After the interview, I talk about 3 highlights of my wisdom of crowds model for predicting the 2017 NFL Draft.

To listen on iTunes, click here.

To listen on the site, click on the right pointing triangle.

Predicting the 2017 NFL Draft by Wisdom of Crowds

You want know which player your team will pick in the 2017 NFL draft.

You’re not alone. Hope springs eternal in the off season, even in Cleveland. There’s no shortage of mock drafts on the internet to satisfy this interest.

As a football analytics site, The Power Rank wants to use this data to make even better predictions on the draft. I take a wisdom of crowds approach and combine the predictions of 15 different mock drafts.

Research suggests this wisdom of crowds will lead to better draft predictions. A study from 2014 found that this approach better predicted draft position than any individual mock draft.

The preseason AP college football poll also uses the wisdom of crowds. My work has found that the higher ranked team wins almost 60% of bowl games, a remarkable rate because this predictor uses no data from the regular season.

For the NFL draft, the wisdom of crowds approach has clear flaws. Teams use the draft to fill certain needs, so some of these predictions will not make sense. I’ll discuss these issues based on my NFL adjusted statistics below.

However, these predictions should capture the relative position of where each player gets picked in the draft. Let’s look at the wisdom of crowds prediction for the 2017 NFL draft.

1. Cleveland, Myles Garrett, DE

The Browns needs lots of help of both sides of the ball. Garrett, the pass rush phenom from Texas A&M, was first in every mock draft.

2. San Francisco, Solomon Thomas, DE

When you need as much help as the Niners on both sides of the ball, it makes sense to get a defensive lineman that can rush the passer. Stanford stud Thomas stays in the Bay Area to join the Niners.

3. Chicago, Jonathan Allen, DE/DT

The Bears had the 11th best pass offense by adjusted yards per attempt last season despite a rash of QB injuries. Hence, it makes sense to draft a defensive player like Allen, the defensive end from Alabama with 10.5 sacks and 16 tackles for loss last season.

4. Jacksonville, Marshon Lattimore, CB

The Jags had the 6th best pass defense by my adjusted yards per attempt, perhaps a surprise. Still, this is a strong draft at cornerback, one of the NFL’s toughest positions to fill. Lattimore, the cornerback from Ohio State, emerges from the wisdom of crowds as the top secondary player.

The pass offense was terrible by my adjusted yards per attempt (27th). However, it doesn’t seem like this team is ready to give up on young QB Blake Bortles.

5. Tennessee, Malik Hooker, S

The Titans got this pick by trading away the last year’s top pick, which the Rams used on QB Jared Goff. The Titans need help on defense, and either Hooker or Lattimore, Ohio State teammates from last season, makes sense here.

6. New York Jets, Jamal Adams, S

By my numbers, the Jets need help everywhere except run defense. They should pick the LSU safety Adams unless they want to play dice with a quarterback at 6th.

7. Los Angeles Chargers, Leonard Fournette, RB

The Chargers can’t run the ball, as they ranked 27th in my adjusted yards per carry last season. Fournette could help an offense that can throw the ball with Phillip Rivers.

8. Carolina, Mitch Trubisky, QB

There’s no way the Panthers take Trubisky with Cam Newton locked in as their franchise QB.

But this begs the question of why any team would take Trubisky in the first round. He only started one season at North Carolina, which means he couldn’t beat Marquise Williams out for the job the prior two seasons. Williams isn’t currently signed by any NFL team.

Kevin Cole did a nice Bayesian analysis to compare Trubisky with Deshaun Watson. The small sample size on Trubisky makes it hard to be certain of his pass efficiency.

Back to Carolina, who had a surprisingly good pass defense (10th by adjusted yards per attempt) after letting Josh Norman go and playing young cornerbacks. A running back like Fournette or McCaffrey makes sense here to help out Newton.

9. Cincinnati, OJ Howard, TE

By last season’s numbers, the Bengals need help with pass rush and pass protection. However, Alabama TE Howard would add to an explosive offense that already has QB Andy Dalton and WR A.J. Green.

Also, Cincinnati ended up 11th in my member team rankings despite a 6-9 record. Get this pick right, and they can easily bounce up to the playoffs in 2017.

10. Buffalo, Mike Williams, WR

The Bills ranked 25th by my adjusted yards per pass attempt last year, so a receiver makes sense here. Williams outplayed CB Marlon Humphrey when his Clemson Tigers beat Alabama in the college football championship game.

11. New Orleans, Reuben Foster, LB

The Saints need help on defense, so a linebacker who had 13 tackles for loss for Alabama last season makes sense here.

Foster did have a positive drug test at the NFL combine for dilute urine. If Foster drops, pass rusher Derek Barnett is another good pick for the Saints.

12. Cleveland, Christian McCaffrey, RB

Rumors suggest that the Browns like QB Mitch Trubisky, and this model suggests they would have to trade up from 12th to draft him.

13. Arizona, Derek Barnett, DE

Carson Palmer was awful for the Cardinals last season, as they dropped from 1st in 2015 to 22nd in 2016 in my adjusted yards per pass attempt.

However, a lot of the mock drafts have Arizona drafting a defensive player. A pass rush specialist that Barnett who had 13 sacks and 19 tackles for loss will help any defense.

14. Philadelphia, Corey Davis, WR

The Eagles pass offense ranked only 28th in my pass offense numbers, as they didn’t let rookie QB Carson Wentz throw deep that much. A talent like Davis from Western Michigan should help the pass offense, although they did sign Alshon Jeffrey and Torrey Smith in free agency (ht G.Moore).

15. Indianapolis, Haason Reddick, LB

The Colts had all kinds of problems on defense last season. Reddick had 9.5 sacks and 22.5 tackles for loss at Temple last season.

16. Baltimore, John Ross, WR

The Ravens won the Super Bowl right before QB Joe Flacco’s contract ended. With this recency bias, they now overpay him and have the 26th ranked pass offense to show for it.

Still, the Ravens contend for the playoffs every year, and they’ve built this team through the draft. Ross, the speedy receiver from Washington, should help their pass offense.

17. Washington, Deshaun Watson, QB

The Redskins still have Kirk Cousins at QB, although his long term contract status is uncertain. They could draft Watson, the accurate QB from Clemson, with an eye on the future.

Washington also had an awful secondary last season despite having Josh Norman at cornerback. Drafting Gareon Conley, the cornerback from Ohio State, probably makes the most sense.

18. Tennessee, Gareon Conley, CB

With the 22nd ranked pass defense, drafting a cornerback here for the Titans is a no brainer.

19. Tampa Bay, Cam Robinson, OT

The Bucs had the 28th ranked rush offense of 32 teams in my adjusted yards per carry. Robinson, the tackle out of Alabama, might help this situation, although they could also draft a running back like Dalvin Cook here.

20. Denver, Taco Charlton, DE

Charlton makes no sense as the Broncos have an incredible pass rush but awful pass protection. Cam Robinson out of Alabama or Ryan Ramczyk from Wisconsin would help bolster the offensive line for whoever plays QB next season.

21. Detroit, Dalvin Cook, RB

The Lions need help in the secondary to help CB Darius Slay. After Lattimore and Conley, the wisdom of crowds model like Marlon Humphrey from Alabama and Kevin King from Washington as the next best cornerbacks.

The Lions should also consider Taco Charlton, the defensive end from Michigan that blossomed with 10 sacks and 13 tackles for loss as a senior.

22. Miami, Ryan Ramczyk, OT

The Dolphins could use help with pass protection, as they’ve ranked 25th the past two years in my adjusted sack rate allowed. Ramczyk is a good choice.

23. New York Giants, Pat Mahomes, QB

The model predicts Mahomes as the third QB off the board, and 13 of 15 mock drafts had him as a first round selection. As Eli Manning is 36, the Giants would be wise to look for his replacement here.

24. Oakland, Marlon Humphrey, CB

The Raiders had a league worst pass defense by my adjusted yards per attempt. Humphrey out of Alabama makes perfect sense here.

25. Houston, Forrest Lamp, OG

The Texans need a quarterback, and they should hope that Watson or Mahomes drops to them at 25th. A offensive lineman like Lamp could help a run offense that ranked 24th by my adjusted yards per carry.

26. Seattle, Garett Bolles, OT

The Seahawks need lots of help on the offensive line, as they played an offensive tackle that didn’t start playing football until 2015. Bolles would help that offensive line protect Russell Wilson.

27. Kansas City, Kevin King, CB

Andy Reid likes to draft linemen, so don’t be surprised if the Chiefs draft an offensive line prospect like Bolles or Lamp. It could also make sense to draft a cornerback like King from Washington.

28. Dallas, Charles Harris, DE

The Cowboys would bolster their pass rush with Harris, the defensive end out of Missouri with 9 sacks and 12 tackles for loss last season. A cornerback choice like Kevin King also makes sense.

29. Green Bay, Malik McDowell, DT/DE

I’m not sure how accurate a wisdom of crowds model can predict the draft slot for McDowell. He has clear NFL talent but often looked disinterested at Michigan State last season. He made the first round in only 7 of the 15 mock drafts, but was selected as high as 17 in two of them.

If the Packers want to take a risk, they could draft McDowell to replace the departed Julius Peppers.

30. Pittsburgh, Tre’Davious White, CB

The Steelers could enhance a pass defense ranked 13th last season with White. A pass rusher like Charles Harris or Malik McDowell also makes sense here.

31. Atlanta, Jarrad Davis, LB

The Falcons should improve their pass rush, which ranked 23rd in my adjusted sack rate last season. Davis only had 6 tackles for loss in 9 games for Florida last season.

A defensive end like Takkarist McKinley from UCLA makes more sense here. McKinley was ranked 33rd by the wisdom of crowds model.

32. New Orleans, Jabrill Peppers, S

The mock drafts couldn’t make up their mind on Peppers. He was a first found pick in 8 of 15 drafts, but he went has high as 10th in one of them.

Peppers could really enhance the Saints defense with this explosiveness (13 tackles for loss with Michigan last season). Or he could struggle to see the field because of his inability to cover slot receivers in the NFL.

Podcast: Kevin Cole on predicting the 2017 NFL Draft

On this week’s episode of the Football Analytics Show, I talk with Kevin Cole, data scientist and writer with Roto Viz and Fantasy Labs. Among other topics, we discuss the following:

  • The unusual but straight forward path Kevin took into football analytics.
  • The four variables that matter in predicting running back performance in the NFL.
  • What small sample size says about Deshaun Watson versus Mitch Trubisky.
  • Why the combine doesn’t matter in predicting wide receivers.
  • Kevin’s bold prediction for the 2017 NFL draft, which involves a quarterback.

I also really enjoyed Kevin’s favorite book, as it showed his depth of thought on topics far from sports analytics.

To listen on iTunes, click here.

To listen on the site, click on the right pointing triangle.

Podcast: John Urschel, NFL lineman, on football analytics and math research

On this episode of the Football Analytics Show, John Urschel, lineman for the Baltimore Ravens and Ph.D. candidate in mathematics at MIT, joins me for a wide ranging conversation. Topics include:

  • The NFL’s progress in using video tracking data
  • The football analytics problem John is working on
  • The acoustic operator in those Bose commercials with J.J. Watt
  • The breadth of math research in which John is engaged
  • The hidden factor in his success as an NFL lineman

John is an amazing human being, and I hope you enjoy this episode.

You also might be interested in an interview I did with John 3 years ago before he got drafted by the NFL.

To listen on iTunes, click here.

To listen here, click on the triangle.