Why Joakim Soria will not continue to pitch this well

Screen Shot 2015-05-20 at 9.46.21 AMOver at the Detroit News, I used defense independent pitching statistics to look at each pitcher in the Tigers’ bullpen.

I had to lead with the closer Joakim Soria. He’s having a fantastic season with a 1.06 ERA. However, an unsustainable level of hit suppression has fueled this performance.

To get all the details, click here.

Mailbag: What do numbers say about Tom Brady and deflated footballs?

Thank you for submitting these questions to my first ever mailbag post. I won’t be able to answer all of the questions in this article, so you can still submit a question for the follow up article.

What do numbers say about Tom Brady and deflated footballs?

What’s your take on the statistics of the deflated balls? Is Tom Brady proof that everything that Michigan touches becomes tainted?

— Brilliant scientist by day, Twitter troll by night.

Deflated footballs? Some people from Pittsburgh made a video about the cure to that problem. For a good laugh, click on this link. Don’t worry, I’ll be here when you get back.

On a more serious note, the Tom Brady deflated football scandal caused an interesting study on fumble rates from Walter Sharp. Over the last 5 seasons, the Patriots averaged 187 plays between fumbles lost, the best rate of holding onto the football of any NFL team over this time.

However, to understand ball security, it’s more important to look at all fumbles, not only those lost to the other team. The Patriots have also excelled in this category, going 73 plays between fumbles over the last 5 years. Only Atlanta and New Orleans have fumbled less.

If you believe the Patriots are cheaters (not unreasonable given Spygate and Deflategate), this data supports your case. The Patriots purposely played with a deflated ball, which allowed them to grip it better and fumble less.

The problems with the fumble rate data

However, there are many problems with this analysis. If you want to get it right, you should break down the Patriots fumbles based on who had the ball. This requires looking at play by play data to separate QB Tom Brady, ball carriers, receivers and special teams.

Deadspin looked at fumble rates on carries and receptions for the Patriots. Since 2007, their ball carriers have led the NFL, but not by a significant margin. For fumbles after catches, they’re third. For this data, check out the bottom part of this article.

This suggest the Patriots ball security boils down to Brady and special teams. Since Brady handles the ball far more than punters and kick returners, the Patriots low fumble rate is most likely due to him.

In addition, consider the other two teams with low fumble rates: Atlanta and New Orleans. Sharp suggested this was because they played in domes. However, dome teams don’t fumble less on carries and receptions according to the Deadspin results. In addition, Atlanta and New Orleans have elite quarterbacks (Matt Ryan and Drew Brees).

This data suggest that quarterbacks have some skill in protecting the ball, which would have interesting applications in college football.

Did Tom Brady deflate balls to help him with this ball security? The NFL thinks so, as they suspended him for the first 4 games of the 2015 season. The NFL should also check the balls of the Falcons and Saints.

Does Michigan taint everything it touches?

No, Michigan isn’t bad luck for all it’s former players. There’s a guy named Jim Harbaugh that has crushed his coaching stints with Stanford and the San Francisco 49ers. He’s back in Ann Arbor to take over the reigns at Michigan.

You want a program that taints their players? How about Duke basketball? For all their success in the NCAA tournament, their best professional player might be J.J. Redick. Kyrie Irving escaped the curse by getting hurt most of his only year at Duke.

Preseason college football rankings

Anyway to project college football predictions this early using analytics this early based on rosters. Everyone got Florida State wrong in the preseason last year?

— Ray Watson.

It’s certainly possible to make good college football predictions right now. I’m currently working on my preseason rankings for 2015. These rankings come from a regression model that considers the past 4 years of team performance, turnover margin and returning starters.

In these rankings, each team has a rating that gives a predicted point spread against an average FBS team. The difference in rating of two teams plus 3 points for home field gives a predicted point spread. In 2014, this point spread predicted the winner in 70.0% of games. It has done an even better 70.5% in all games since 2005.

These rankings overrated Florida State last season. They had the Seminoles as the top ranked team, about 1.3 points better than Alabama. However, no one expected the steep decline in their defense. After two seasons in the top 5, they dropped to 53rd in my defense rankings based on yards per play adjusted for schedule strength.

Members will see all of my preseason numbers in early June when the markets role out win totals. I’ll make the rankings available to everyone over the summer.

Predicting totals in football and basketball

Do you foresee a way in which your analytics will not only predict winning percentages in games, but also total points? I like looking at game totals as well as sides in basketball and football.

— Ken

Yes, totals are a huge goal for The Power Rank over the next year. I started posting totals for college football last season, and I plan to do the same for the NFL this season. Basketball will come after that. Stay tuned.

How to determine how often a team covers the point spread

When betting ATS, how do you calculate your advantage to a win probability? For example, if the Houston Rockets are +7.5 at the books but your numbers imply they are really only 5 point dogs, you would have a 2.5 point advantage. How would you turn that 2.5 points into the probability that Houston would then cover the spread?

— Josh Burton, member.

To get a rough estimate, you could use the relationship between point spread and win probability.

For example, a NBA team favored by 2.5 points has a 58.4% win probability. If we make some blind assumptions about your Rockets game, then a 2.5 point edge from the model equates with a 58.4% chance to cover. However, this assumes your model is very good. Handicapper Bob Stoll discusses this issue in this article under Determining Value.

To get a better answer, you could develop a simulation of the game of basketball. This simulation would capture features like the rate at which teams take 3 pointers and foul at the end of games. Both of these factors affect the distribution of outcomes for the game.

A game simulation gives you a better estimate of the probability the team covers. However, this estimate is still at the mercy of the quality of the underlying model.

Do stars determine NBA champions?

Has star power generally been a factor in determining the NBA champion?

— Yoni Aharon

Star power generally has a bigger role in basketball than other sports. Only 5 players take the court at one time, which allows for a superstar like LeBron James to have a bigger impact than his counterpart in other sports.

I would argue that star power is a huge factor in winning NBA titles since I can only name two championship teams that didn’t have a star in his prime.

  • 2004 Detroit Pistons. Coached by Larry Brown and led by Chauncey Billups, this Piston team destroyed the Lakers with Shaq and Kobe in the Finals. They made Finals again the following year only to lose to Tim Duncan and San Antonio.
  • 2014 San Antonio Spurs. The Spurs won 5 titles with Tim Duncan, but he was no longer in his prime during this last championship. The Spurs destroyed the Heat and LeBron James in the finals.

The 2014 Spurs have changed the NBA for the better. Other teams have copied their unselfish brand of offensive basketball that features crisp passing and high percentage shots. This wasn’t the case as recently as 2010 when Kobe Bryant won the Finals MVP after a game 7 in which shot 6-24 from the field and 0-6 from three.

The college game used to be a better place to find teams that played this brand of beautiful basketball. Think of the 2005 Illinois team with Deron Williams that made the championship game, or the 2014 Michigan team led by Nick Stauskas. The Spurs have brought this to the NBA. Thank you, Gregg Popovich.

Sources of college sports data

As someone looking to get started in doing his own amateur sports analysis, what are some good sources for data that I may be overlooking? In particular, for football I’ve found a lot of good data on Football Outsiders and ProFootballFocus, for basketball I’ve been using Basketball Reference, and for baseball I’ve been using FanGraphs and Baseball Reference. Are there any sources I’m overlooking there? Particularly for college sports, I’ve found there is less widely available data out there (I realize a lot of the good predictive data may be proprietary).

— Rick Taylor

Rick, that’s awesome that you’re looking to crunch your own numbers.

You hit on most of the big ones. I’m always on Baseball Reference and FanGraphs for my raw baseball data.

For college football, look at cfbstats.com. Marty does a fantastic job breaking down the play by play data into all kinds of useful ways. However, please be careful, since his data counts sacks as rush plays. Although all major sites do this, plays that end in sacks started as passes.

My friend Bill Connelly at SB Nation also posts college football data. Here’s an old link to his data. He promises to post more. You can bug him on Twitter.

Last, kostats has a wealth of game and market data in a lot of sports for a reasonable price.

6 surprising MLB teams early this 2015 season

Screen Shot 2015-05-08 at 10.16.48 AMBaseball is difficult to predict early in the season. Even with a month in the books, small sample size makes it hard to come to firm conclusions on teams.

On Numberfire, Daniel Lindsey looks at 6 teams that have defied preseason expectations early this 2015 season. He does a great job diving into the numbers and highlighting important players.

Numberfire does their own rankings, and they look like they only use data from the current season. It’s an interesting comparison with my rankings and predictions that still weight the preseason heavily.

For example, consider Washington, a team many had as their preseason World Series Champion. Numberfire has the Nationals ranked 22nd, while I have them 4th. With more games, we’ll find out their true strength.

To check out the Numberfire article on 6 surprising MLB teams, click here.

The Tigers’ bullpen will struggle according to predictive pitching statistics

Screen Shot 2015-05-08 at 9.24.31 AMMy latest article with the Detroit News looks at the Tigers’ bullpen through modern pitching statistics. The take home message: the bullpen isn’t that good.

However, the more fun part is explaining fielding independent statistics to the average Tigers fan. Research has shown that pitchers have no control over balls hit into the field of play.

While the numbers have confirmed this over huge sample sizes, it’s still hard for the human mind to accept. I was very upfront about this in the article. However, it didn’t stop readers from posting 47 comments, and about 49 of these comments were negative.

However, we must keep up the fight against dinosaur statistics like ERA.

To read my article on modern pitching statistics and the Tigers’ bullpen, click here.

Finally!! Early season baseball predictions available at The Power Rank

The predictions page now has a win probability for every baseball game.

These predictions start with the MLB team rankings. Later in the season, these rankings are my team rankings that adjust run differential for strength of schedule. For more details, click here.

However, these team rankings don’t make much sense in April. The rankings that include games through April 23rd, 2015 had the Mets as the best team in baseball, three runs better than the average MLB team.

For the first two months of the season, I’ll combine preseason expectations with these team rankings based on the current season. You can always find the most current rankings at the MLB rankings page, which gets updated daily.

To account for starting pitching, I use the ZiPS projections from FanGraphs. Dan Szymborski developed these pitcher projection based on 4 years of data and the ideas of defense independent pitching statistics.

The projections didn’t do that well yesterday. The team with greater than 50% win probability won 4 of 12 games.

However, this small sample size doesn’t worry me too much. Highly touted teams like the Nationals and Dodgers both lost yesterday. We’ll see how the predictions do over a larger sample of games.

Thank you to everyone who asked me to get these predictions up. Check the prediction page daily for latest.