On this episode of The Football Analytics Show, I’m joined by Kostya Medvedovsky, creator the DARKO NBA player projections. Among other topics, we discuss:

• How a lawyer created one of the most well respected NBA player predictions (3:22)
• From a lawyer who uses Python to a new career (11:34)
• Kalman filtering for basketball (20:00)
• Machine learning and Brandon Ingram (24:50)
• Steph Curry, Klay Thompson and predicting three point percentage (32:30)
• Plus minus for NBA players and Tim Duncan (39:03)
• NBA spreads (49:55)

I really enjoyed talking with Kostya about his methods, and I hope you get the same value out of this conversation.

To listen here, click on the right pointing triangle.

This episode of The Football Analytics Show is also available on Spotify and Apple Podcasts.

You want to bet a Masters outright. You check out a site like Data Golf and get their win probability for each golfer. Compare it to the markets and make a bet.

Unfortunately, most of these predictions won’t show any value. For example, Xander Schauffele has a 3.0% chance to win, which doesn’t suggest value at +2400 (4% break even win probability).

To find value, you have to go beyond the predictions. Data Golf uses data on each and every shot to make predictions. Perhaps some of these shots have more randomness than others, and this presents an advantage.

To make a comparison to a different sport, consider college basketball. As I discussed in March, three point shooting has a large component of randomness. Teams that have made a high rate of threes should expect regression to the mean.

Brandon Gdula of numberFire found a similar result for putting. To understand his results, consider strokes gained, a metric for how well a golfer performs compared to the tour average. For example, consider a putt within five feet, a shot made 97% of the time. If a golfer sinks this putt, then he gains 0.03 strokes (one minus the tour average of 0.97). Missing this putt results in -0.97 strokes gained.

Brandon found that strokes gained for short putts explains most of the variance for strokes gained on all putts. To be specific, let’s look at the R-squared value. Strokes gained for putts of 15 feet or less explained 80.5% of the strokes gained overall for putting.

This implies a huge random component for long putts. Brandon suggests looking at the 73% of putts from 15 feet or less in getting a more accurate assessment of putting.

To apply this idea to the Masters, consider Xander Schauffele. Through March 18th, Brandon pulled data for players with at least 100 rounds of strokes gained data. Overall, Schauffele was 15th in strokes gained putting.

However, when only using putts from within 15 feet, Schauffele ranks first overall. He is great on putts that are predictive, and he has had some bad luck on longer putts that pulls down his overall putting strokes gained.

This analysis will be far from perfect. In Brandon’s article, he mentioned that he didn’t find putting at any distance to be predictive over half year intervals. There is a ton of randomness in putting.

However, Schauffele is a great all around golfer who finished 3rd in last year’s Masters. The putting data suggests that the numbers might be low on him, and there is value in betting Schauffele to win at +2400.

Here are golf and Masters resources to check out:

• The putting study by Brandon Gdula
• Masters predictions by Brandon Gdula
• Data Golf. They put most of their predictions behind a pay wall but they also have many free resources.

Sports betting with a PhD edge

This article was sent to The Power Rank’s email newsletter, a free service that strives to be:

• Valuable
• Concise
• Entertaining

While the primary focus is NFL and college football, the content covers golf, the NFL draft and the NBA during the off season.

To get this free service delivered straight into your inbox, enter your best email address and click on “Sign up now!”

 

 

 

Your weekly dose of sports betting tips and news, every Saturday morning at 10am Eastern.

College basketball predictions – These predictions come from The Power Rank’s member numbers that combine a number of powerful predictors into an ensemble. These conference tournament games are all on neutral sites.

• Duke will beat Virginia Tech by 7.0 points – The markets have favored Duke more aggressively than my numbers. However, as of Saturday morning, they have Duke -5.5. Either the markets move strongly in Duke’s favor or they have lost some faith in this team after not looking sharp in their two ACC tournament games.
• Iowa will beat Indiana by 3.7 points – The Hoosiers have won both of their Big Ten tournament games by five or fewer points. They had a regular season Big Ten record of 2-5 in these type of close games. The markets have Iowa -6.
• Arizona will beat UCLA by 1.5 points – UCLA thumped Arizona by 16 at home, and then Arizona won by 10 over UCLA at home.

NFL futures after QB trades – One of the more foolhardy segments in sports talk is arguing which side of a major transaction won the trade. Often times, it is a stalemate in which no team will end up in a better place. Even if a team did win a trade, we will not know for years.

However, in the short term, Denver became drastically better by acquiring Seattle QB Russell Wilson. Per Next Gen Stats, Wilson finished 15th among qualifying quarterbacks in Completion Percentage Over Expected last season.

By adding Wilson, Denver now has the 6th lowest odds to win the Super Bowl (+1300 per FanDuel Sportsbook). It is worth noting they are in the same division as Kansas City at +850, and only one team in the AFC gets homefield advantage throughout the playoffs.

In addition, Green Bay kept Aaron Rodgers, and they have the second best Super Bowl odds (+750) behind Buffalo.

Model vs subjective analysis – On the Circles Off podcast, The Power Rank founder Ed Feng talked with Rob and Johnny about the balance between predictive analytics and subjective adjustments.

Ed talked about how his NFL model has very little contribution from rushing statistics. As a consequence, he made some bets against the Colts and Eagles this past season that didn’t get closing line value.

In addition to football betting, Ed also discussed content creation and March Madness bracket strategies. Circles Off has a ton of great episodes worth checking out.

Three point shooting in college basketball – As an analytics fella, I want to look at three point shooting offense and defense when projecting the NCAA Tournament. However, in an episode of Bracket Wisdom, Ed Feng explained the randomness in these statistics.

Among projected tournament teams, you might consider fading South Dakota State, Purdue and Michigan State because of a high three point field goal percentage. In contrast, LSU, San Francisco and Iowa State have unsustainable rates of three point shooting defense.

These projected teams were obtained from Bracketmatrix, a wisdom of crowds estimator of the teams that will make the tournament.

Baylor – Scott Drew’s team got bumped from the Big 12 tournament in the quarterfinals against Oklahoma. However, don’t make too much of one game. Oklahoma shot a blistering 52.4% from 3 while Baylor went cold at 13.6%.

As discussed on a Bracket Wisdom episode, Baylor thrives on shots at the rim and offensive rebounding. On defense, they rank a very respectable 12th in The Power Rank’s adjusted points per possession.

In addition, athletic freshman Kendall Brown and Jeremy Sochan are potential 1st round picks in the next NBA draft. This gives Baylor an upside that few other tournament contenders can match.

You can get all of the daily Bracket Wisdom episodes on Apple Podcasts and Spotify.

Baseball is Back – If fireworks can be legally popped when a home team wins World Series games, why can’t we do the same across the country knowing we are getting a full 162-game season? Fangraphs went over what the new Collective Bargaining Agreement will look like and how it will affect our enjoyment of the game.

This weekend could mark a free agent frenzy, given how soon Spring Training will begin. The Dodgers are clear favorites to win the World Series, but for value plays, it may be worth forecasting which free agents will go where, or pouncing the moment news is released on a new signing. Spotrac.com may be worth hitting the refresh button to keep track of free agency.

Contrarian strategies – In 2019, Duke came into the NCAA tournament as a big favorite. Zion Williamson came back from an injury, and RJ Barrett was one of college basketball’s most dominant players on the drive in the last decade.

The Power Rank gave Duke a 34% chance to win this tournament, and picking Duke as champion gave you the best chance to win a small pool. However, 36% of the public brackets on ESPN had Duke as champion. This suggests a better strategy for intermediate sized pool.

In The Power Rank 2022 Bracket Advice and Course, there is a 22 minute audio course that explains this strategy.

Data driven betting information

These nuggets were sent to The Power Rank’s email newsletter. This is a free service that strives to be:

• Valuable
• Concise
• Entertaining

You also get The Power Rank’s March Madness cheat sheet that makes it drop dead easy to fill out your bracket.

To get this free service delivered straight into your inbox, enter your best email address and click on “Sign up now!”

 

 

 

This content is for members only

On this episode of The Football Analytics Show, I’m joined by Colin Davy, former director of data science at The Action Network and Jeopardy champion through analytics. In talking about how to find sports betting value, we discuss:

• The foundational idea behind Betscope (5:29)
• How to use correlations to find sports betting value (14:41)
• The analytics (underlying distributions) behind player props (17:16)
• The underlying distributions for 100 yard rushers (24:10)
• Attracting smart sports bettors (38:00)

To listen to here, click on the right pointing triangle:

This episode of The Football Analytics Show is also available on Apple Podcasts and Spotify.