Podcast: Cynthia Frelund on football analytics, NFL linemen

On this episode of The Football Analytics Show, I talk with Cynthia Frelund, the NFL Network’s analytics expert. She has also worked on analytics for the NFL league office and ESPN.

Among other topics, we discuss:

  • her large sample size video study of NFL linemen
  • the lineman from UTEP to look out for in the draft
  • the real problem with the Browns, and what they’ll do in the 2018 draft
  • Next Gen data, and the 2018 project Cynthia is working on
  • the type of NFL coaches least likely to adopt analytics

Cynthia not only does quality work but also brings energy and excitement to football analytics.

To listen on iTunes, click here.

To listen here, click on the right pointing triangle.

2018 NBA Championship Odds at the start of the playoffs

Predicting the 2018 NBA playoffs is a mess. The list of problems starts with these issues:

  • Elite players like Steph Curry and Joel Embiid will start the playoffs injured.
  • Golden State has not performed up to expectation this season, as their defensive efficiency dropped from 2nd to 11th from 2017 to 2018.
  • Kyrie Irving is out for the entire playoffs.

Usually, I take my team rankings that include data from all regular season games and calculate championship probabilities. However, that will not work this season.

Here’s how I approached predicting the 2018 NBA playoffs.

First, I took the game results from the season and kept only games in which teams had all their key players. For example, Golden State had 29 games in which Steph Curry, Klay Thompson, Kevin Durant and Draymond Green all played.

This reduces the set of games, but it gives a better picture of how a team might perform with its top players. These numbers assume Golden State and Philadelphia will have Curry and Embiid respectively. With this reduced set of games, I take margin of victory and adjust for schedule with my ranking algorithm.

Second, I take the closing point spreads in the markets since January 1st and perform the same filtering process with top players. I generate market rankings by adjusting these spreads for schedule.

I decided to focus on the last part of the season to get a more recent picture of team performance. The markets still believe in Golden State, but not as much as in earlier in the season.

Then I blended these two rankings to give the following rankings of playoff teams. The rating gives an expected margin of victory against an average NBA team on a neutral court.

1. Houston, 8.53
2. Golden State, 8.45
3. Toronto, 5.49
4. Oklahoma City, 3.05
5. Cleveland, 2.69
6. Boston, 2.47
7. Utah, 2.46
8. Minnesota, 2.42
9. Philadelphia, 2.05
10. Washington, 1.93
11. San Antonio, 1.79
12. Portland, 1.56
14. Indiana, 0.70
15. New Orleans, 0.57
16. Milwaukee, 0.41
18. Miami, -0.48

This leads to the following championship probabilities for 2018 NBA playoffs.

1. Houston, 39.4%
2. Golden State, 36.5%
3. Toronto, 14.9%
4. Boston, 2.2%
5. Cleveland, 1.9%
6. Philadelphia, 1.6%
7. Oklahoma City, 0.9%
8. Washington, 0.5%
9. Utah, 0.5%
10. Portland, 0.4%
11. Minnesota, 0.3%
12. San Antonio, 0.2%
13. Indiana, 0.2%
14. Milwaukee, 0.2%
15. New Orleans, 0.1%
16. Miami, 0.1%

You can compare the calculations with these implied odds from the markets, taken from Bookmaker on Friday morning, April 13th, 2018.

1. Golden State, 36.6%
2. Houston, 28.9%
3. Cleveland, 10.9%
4. Toronto, 9.1%
5. Philadelphia, 4.8%
6. Oklahoma City, 2.1%
7. Utah, 1.7%
8. Portland, 1.4%
9. San Antonio, 1.2%
10. Boston, 1.0%
11. Indiana, 0.5%
12. Washington, 0.5%
13. Milwaukee, 0.4%
14. Miami, 0.4%
15. Minnesota, 0.3%
16. New Orleans, 0.3%

Houston and Golden State should have the best odds to win the title, and both the calculations and markets agree on this.

My calculations are low on Cleveland. The conventional wisdom is that regular season results don’t matter in predicting how LeBron James will perform in the playoffs. This may or may not be true.

However, LeBron does play almost 4 more minutes per game in the playoffs than in the regular season. This should make Cleveland better in the playoffs. My numbers don’t account for different minute distribution and shows how better models could be built with play by play data.

My calculations are also too high on Boston. I did not exclude games in which Kyrie Irving played in my analysis, so some adjustment should be made for this injury.

Get the game by game predictions each day on the main predictions page.

Prediction for 2018 college hockey championship game

Minn-Duluth vs Notre Dame: Minn-Duluth has a 45.6% chance to win. Notre Dame has a 34.1% chance to win. There is a 20.3% chance for overtime.

This prediction is based on the college hockey rankings (see below), which originate from offense and defense rankings for 60 Division I teams. These offense and defense rankings imply a goal rate for an offense against an opposing defense.

Based on these goal rates, I use a Poisson process model to calculate the probability for a win, loss or overtime for each game in regulation. If a game is tied, it goes to an overtime in which the first team to score wins.

Here are the team rankings based on offense and defense rankings. The rating after each team gives an expected goals against an average Division I hockey team.

1. St Cloud State, 1.70
2. Denver, 1.61
3. Ohio State, 1.48
4. Minn-Duluth, 1.46
5. Cornell, 1.37
6. Minnesota State, 1.36
7. Clarkson, 1.24
8. Northeastern, 1.19
9. Notre Dame, 1.17
10. Providence, 0.97
11. North Dakota, 0.97
12. Boston College, 0.90
13. Penn State, 0.87
14. Minnesota, 0.80
15. Michigan, 0.79
16. Boston University, 0.75
17. Harvard, 0.63
18. Princeton, 0.61
19. Union, 0.44
20. Wisconsin, 0.32

Podcast: Colin Davy on predicting the Masters

On this episode of The Football Analytics Show, I talk with Colin Davy, director of data science at The Action Network. In discussing the Masters (yes, the first appearance of golf on this site), we discuss the following:

  • The algorithm he developed based on Markov chains to rank tennis players
  • Whether a thick or thin playbook leads to better offense in football
  • How to apply the tennis algorithm to golf
  • The difficulty in predicting Tigers Woods at the Masters
  • The one golfer to keep an eye on this weekend

Colin is an exceptional data scientist, and I’ll know you’ll discussion of the Masters. You can check out all of the Masters content at the Action Network by clicking here.

To listen to the show on iTunes, click here.

To listen here, click on the right pointing triangle.

Predictions for 2018 Frozen Four, College Hockey

These predictions are based on the college hockey rankings (see below), which originate from offense and defense rankings for 60 Division I teams. These offense and defense rankings imply a goal rate for an offense against an opposing defense.

Based on these goal rates, I use a Poisson process model to calculate the probability for a win, loss or draw for each game in regulation. If a game is tied, it goes to an overtime in which the first team to score wins.

Minn-Duluth vs Ohio State: Ohio State has a 40.2% chance to win. Minn-Duluth has a 39.9% chance to win. There is a 19.9% chance for overtime.

Michigan vs Notre Dame: Notre Dame has a 47.9% chance to win. Michigan has a 35.3% chance to win. There is a 16.8% chance for overtime.

Here are the team rankings based on offense and defense rankings. The rating after each team gives an expected goals against an average Division I hockey team.

1. St Cloud State, 1.7
2. Denver, 1.6
3. Ohio State, 1.5
4. Minn-Duluth, 1.5
5. Cornell, 1.4
6. Minnesota State, 1.4
7. Clarkson, 1.2
8. Northeastern, 1.2
9. Notre Dame, 1.2
10. Providence, 1.0
11. North Dakota, 1.0
12. Boston College, 0.9
13. Penn State, 0.9
14. Minnesota, 0.8
15. Michigan, 0.8
16. Boston University, 0.8
17. Harvard, 0.6
18. Princeton, 0.6
19. Union, 0.4
20. Wisconsin, 0.3
21. Omaha, 0.3
22. Maine, 0.2
23. Western Michigan, 0.2
24. Miami, 0.2
25. Connecticut, 0.1

The semi-final between Ohio State and Minnesota Duluth features the two top teams, and the winner of this game would be the favorite in the final.