With so many better metrics available, it’s hard to believe the mainstream sports media still uses per game statistics to evaluate player performance.
ESPN Radio’s Colin Cowherd recently compared Derrick Rose to Allen Iverson. The comparison makes sense on some levels. Both players are shoot first, pass second point guards. Both are incredibly quick and great finishers. Cowherd’s mistake was using per game statistics, which made the players appear closer in performance than they really are.
Cowherd started by saying Iverson had the edge in points per game over Rose in their third NBA seasons: 26.8 to 25.0. This brings up the biggest reason per game numbers fall short: starters vary tremendously in how many minutes they see per game. Iverson played 41.5 minutes per game versus 37.4 for Rose. Using points per 40 minutes to even the playing field, Rose (26.7) has actually scored more than Iverson (25.8).
Rose had a huge edge in assists per game (7.9) over Iverson (4.6) in their third seasons. That difference increases with the more revealing assists per 40 minutes figures: 8.4 to 4.5. Iverson did spend extensive time at shooting guard that year while Eric Snow played point for the Sixers, which impacted his assist numbers. Still, Iverson never came close to matching Rose’s assists per 40 minutes figure in any career season. Rose had also shot for the higher percentage from both two-point (47.2 to 44.0) and three-point range (33.2 to 29.1) in season number three.
Rose had the advantage in John Hollinger’s PER as well, 23.4 to 22.2 over Iverson. Both players have high usage rates – which estimates the number of their team’s plays they use while on the court – of nearly 33 percent. So while they both use a high percentage of their team’s possessions, Rose produces more in those opportunities.
Iverson did have a big edge in steals per 40 minutes in his third season. And while he reached the foul line more often than Rose, they made nearly the same number of free throws per minute due to Rose’s far superior free throw percentage.
Most importantly, Rose is younger than Iverson was in his third season by one year and four months. It makes more sense to compare Rose’s third season to Iverson’s second campaign, which would cause the gap between the players to widen even further. Finally, Rose stands three inches taller than Iverson and weighs 25 pounds more.
While they have some similarities, Rose holds a decisive edge over Iverson at the same stage of their careers. That becomes clear when taking a look beyond their per game statistics.
Iverson was a great player. But in both performance and from a branding perspective, Rose is on track to soar much higher than Iverson ever did.
Thursday, March 31, 2011
Wednesday, March 16, 2011
Sports Statistics as a Marketing Tool
Politicians discovered the power of numbers long ago. One might say “my administration created one million more jobs than any other in history.” Of course, it only takes a few minutes to pick that apart: How many jobs were lost? What was the net increase? What was the percentage increase? What was the average salary of these created jobs? But by the time his statement gets scrutinized, the politician moves on to the next talking point.
The same thing works in sports. In 2009, when Colt McCoy was a top Heisman Trophy candidate, the media repeated this statement over and over: “McCoy has won more games than any quarterback in college football history.” Wins are powerful, the true currency of sports. The stat spread everywhere and stuck in people’s minds, even though it was not a particular good statistic.
McCoy won more football games partly because he played in so many. Longer seasons and conference title games gave him more opportunity to record victories. Yes, he won the most games of any quarterback, but he was one of 22 starters. And many of those former teammates have joined him in the NFL.
Obviously it took a great quarterback to win that many football games. McCoy had to earn the starting job and keep it four years; no easy feat at a top program. He had a major role in 45 wins. Nonetheless, teams win games, not quarterbacks.
The McCoy stat still got extensive airtime on sports talk radio, highlight shows and game broadcasts. As with a smooth-talking politician, there was little opportunity in those settings to contradict it with objective evidence.
This demonstrates the power of numbers. Since few people effectively use sports statistics as a marketing tool, they present a blank canvas to work with. And the timing couldn’t be better with the rise of social media, when you may only get 140 characters to send a clear powerful message.
If even bad stats have value, can you imagine the impact from innovative statistical content? Finding this isn’t easy – as the best information lies beyond the core stats that dominate the mainstream sports media – but it is well worth the effort.
The same thing works in sports. In 2009, when Colt McCoy was a top Heisman Trophy candidate, the media repeated this statement over and over: “McCoy has won more games than any quarterback in college football history.” Wins are powerful, the true currency of sports. The stat spread everywhere and stuck in people’s minds, even though it was not a particular good statistic.
McCoy won more football games partly because he played in so many. Longer seasons and conference title games gave him more opportunity to record victories. Yes, he won the most games of any quarterback, but he was one of 22 starters. And many of those former teammates have joined him in the NFL.
Obviously it took a great quarterback to win that many football games. McCoy had to earn the starting job and keep it four years; no easy feat at a top program. He had a major role in 45 wins. Nonetheless, teams win games, not quarterbacks.
The McCoy stat still got extensive airtime on sports talk radio, highlight shows and game broadcasts. As with a smooth-talking politician, there was little opportunity in those settings to contradict it with objective evidence.
This demonstrates the power of numbers. Since few people effectively use sports statistics as a marketing tool, they present a blank canvas to work with. And the timing couldn’t be better with the rise of social media, when you may only get 140 characters to send a clear powerful message.
If even bad stats have value, can you imagine the impact from innovative statistical content? Finding this isn’t easy – as the best information lies beyond the core stats that dominate the mainstream sports media – but it is well worth the effort.
Monday, March 14, 2011
Summary of the MIT Sports Analytics Conference
Here’s a wrap-up of the MIT Sports Analytics Conference held earlier this month, with the focus on items of interest to sports agents.
This year’s conference drew 1,500 people to the Boston Convention and Exhibition Center. Representatives from 53 different professional teams attended, according to the organizers. Now in its fifth year, the event has evolved into as much a business conference as an analytics one, with topics about sponsorships and enhancing the game day experience.
Rockets GM Daryl Morey, one of the event’s organizers, pointed out that basketball is a sport that punishes mistakes during the opening panel on developing the modern athlete. He explained how people focus on all the dunks, but mistakes are costly and need to be minimized for success. While players that shoot high percentages and avoid turnovers get little media attention, teams clearly build such contributions into their statistical models and projections.
In the same panel, Morey said that during the NBA Draft process they’re often looking for flaws more than attributes. They identify what problems a player has that they think they can improve upon. This shows why it may be a good idea to address a player’s shortcomings in draft packages and then demonstrate how they will overcome them.
The Baseball Analytics panel also had some interesting exchanges. Tom Tippett, director of baseball information services for the Red Sox, talked about the Carl Crawford contract. Although he lacked the power of most well-paid outfielders, Tippett said that between triples and home runs, Crawford clears the bases about 30 times per year. Tippett said the team also researched how Fenway Park’s dimensions would impact Crawford’s defensive performance.
Both Major League Baseball and the NBA are moving toward having a complete digital record of each game. This creates tremendous opportunities for sports agents and their staffs to analyze and present this data on behalf of their clients.
This year’s conference drew 1,500 people to the Boston Convention and Exhibition Center. Representatives from 53 different professional teams attended, according to the organizers. Now in its fifth year, the event has evolved into as much a business conference as an analytics one, with topics about sponsorships and enhancing the game day experience.
Rockets GM Daryl Morey, one of the event’s organizers, pointed out that basketball is a sport that punishes mistakes during the opening panel on developing the modern athlete. He explained how people focus on all the dunks, but mistakes are costly and need to be minimized for success. While players that shoot high percentages and avoid turnovers get little media attention, teams clearly build such contributions into their statistical models and projections.
In the same panel, Morey said that during the NBA Draft process they’re often looking for flaws more than attributes. They identify what problems a player has that they think they can improve upon. This shows why it may be a good idea to address a player’s shortcomings in draft packages and then demonstrate how they will overcome them.
The Baseball Analytics panel also had some interesting exchanges. Tom Tippett, director of baseball information services for the Red Sox, talked about the Carl Crawford contract. Although he lacked the power of most well-paid outfielders, Tippett said that between triples and home runs, Crawford clears the bases about 30 times per year. Tippett said the team also researched how Fenway Park’s dimensions would impact Crawford’s defensive performance.
Both Major League Baseball and the NBA are moving toward having a complete digital record of each game. This creates tremendous opportunities for sports agents and their staffs to analyze and present this data on behalf of their clients.
Wednesday, January 26, 2011
Relievers and Consistency
Are relief pitchers more volatile than starters? While that seems to be the case, relievers also get evaluated in much smaller sample sizes.
Statistics usually get expressed in terms of seasons. That works great for starters, but not so well for relievers. Starters throw approximately three times as many innings as bullpen pitchers in a typical season, which gives them far more time to work through their struggles.
For this reason, we compared starters in one-third of a season timeframes to relievers in full seasons. This evens the playing field to help determine which group maintains more consistency.
We identified all relief pitchers that pitched 150 innings or more from 2008 through 2010 and posted an earned run average between 3.00 and 3.50. The starters had to pitch 150 or more innings in 2010 alone with an ERA in that same range. Nineteen relievers and 17 starters met the criteria.
The starters group included CC Sabathia, Cliff Lee, Chris Carpenter and Tim Lincecum. The relievers had pitchers like Francisco Cordero, Huston Street, Brian Fuentes and Jonathan Broxton. The two groups accumulated a comparable number of innings: the bullpen group totaled 3510.2 innings versus 3602.1 for the starters. Both groups had an identical ERA of 3.25.
Based on inspection, the starters appeared slightly more volatile. None of the relievers posted an ERA over 5.00 in any of their annual timeframes. However, two starters did so in their two-month periods: Max Scherzer (6.42) in April/May and Jon Garland (5.16) in June/July. The starters failed to post a single ERA below 2.00, while the bullpen group had three: Jeremy Affeldt (1.73) and Ryan Franklin (1.92) in 2009 and Chris Perez (1.71) in 2010.
Statistics usually get expressed in terms of seasons. That works great for starters, but not so well for relievers. Starters throw approximately three times as many innings as bullpen pitchers in a typical season, which gives them far more time to work through their struggles.
For this reason, we compared starters in one-third of a season timeframes to relievers in full seasons. This evens the playing field to help determine which group maintains more consistency.
We identified all relief pitchers that pitched 150 innings or more from 2008 through 2010 and posted an earned run average between 3.00 and 3.50. The starters had to pitch 150 or more innings in 2010 alone with an ERA in that same range. Nineteen relievers and 17 starters met the criteria.
The starters group included CC Sabathia, Cliff Lee, Chris Carpenter and Tim Lincecum. The relievers had pitchers like Francisco Cordero, Huston Street, Brian Fuentes and Jonathan Broxton. The two groups accumulated a comparable number of innings: the bullpen group totaled 3510.2 innings versus 3602.1 for the starters. Both groups had an identical ERA of 3.25.
Based on inspection, the starters appeared slightly more volatile. None of the relievers posted an ERA over 5.00 in any of their annual timeframes. However, two starters did so in their two-month periods: Max Scherzer (6.42) in April/May and Jon Garland (5.16) in June/July. The starters failed to post a single ERA below 2.00, while the bullpen group had three: Jeremy Affeldt (1.73) and Ryan Franklin (1.92) in 2009 and Chris Perez (1.71) in 2010.
The best way to measure consistency is through standard deviation, which simply measures how much figures differ from the average. The standard deviation for the starters (.635) was slightly lower than the relievers (.677). Basically, there was very little difference in consistency between these groups.
The starters also had one huge advantage. While the relievers had an entire offseason in between their seasons, the starters flowed directly from one two-month period into the next. In several cases, the relievers changed teams and/or leagues during the offseason, making it even harder to sustain consistent performance.
Before making any conclusions, more data should be examined beyond this rather small study. It would also be valuable to examine statistics like OPS allowed, since ERA does not account for how relievers pitch with inherited runners.
Nonetheless, it appears that what some perceive as lack of consistency has more to do with the limited number of innings relievers pitch per season.
The starters also had one huge advantage. While the relievers had an entire offseason in between their seasons, the starters flowed directly from one two-month period into the next. In several cases, the relievers changed teams and/or leagues during the offseason, making it even harder to sustain consistent performance.
Before making any conclusions, more data should be examined beyond this rather small study. It would also be valuable to examine statistics like OPS allowed, since ERA does not account for how relievers pitch with inherited runners.
Nonetheless, it appears that what some perceive as lack of consistency has more to do with the limited number of innings relievers pitch per season.
Monday, September 20, 2010
The Truth about Strikeouts
Managers hate when hitters strike out. The mainstream media often criticizes high-strikeout players like Mark Reynolds, much more than it should.
When it comes to winning and losing, strikeouts by hitters aren’t much more costly than other types of outs. While pitchers’ strikeouts have a major effect on run scoring, the same doesn’t hold true for hitters.
This insight comes from research using an advanced statistical technique called regression analysis. Without getting into the details, regression analysis determines how well statistics correlate with each other. Pitchers’ strikeouts have a much greater correlation with run prevention than hitters’ strikeouts have on run scoring.
How can this be? In general, hitters who strike out a lot also hit home runs and draw walks. On the other hand, strikeout pitchers limit offense better on average than pitchers who miss bats less often. They are also less dependent on their defense to make plays behind them.
Whether in arbitration or free agency, baseball agents can emphasize the value of high strikeout pitchers. And if you represent a high-strikeout batter, exhibits with this information provide hard evidence in his favor.
When it comes to winning and losing, strikeouts by hitters aren’t much more costly than other types of outs. While pitchers’ strikeouts have a major effect on run scoring, the same doesn’t hold true for hitters.
This insight comes from research using an advanced statistical technique called regression analysis. Without getting into the details, regression analysis determines how well statistics correlate with each other. Pitchers’ strikeouts have a much greater correlation with run prevention than hitters’ strikeouts have on run scoring.
How can this be? In general, hitters who strike out a lot also hit home runs and draw walks. On the other hand, strikeout pitchers limit offense better on average than pitchers who miss bats less often. They are also less dependent on their defense to make plays behind them.
Whether in arbitration or free agency, baseball agents can emphasize the value of high strikeout pitchers. And if you represent a high-strikeout batter, exhibits with this information provide hard evidence in his favor.
Thursday, September 16, 2010
Opportunity and Statistics
Most sports statistics – especially the ones that get attention in the mainstream media – are opportunity based. Other metrics filter out opportunity, and they carry tremendous comparative value.
Many still fixate on per game numbers, and they don’t begin to tell the story for players like DeJuan Blair. His 7.8 points per game and 6.4 rebounds per game in 2009-10 look pedestrian. However, Blair posted these numbers in limited opportunities – playing just 18.2 minutes per contest.
Rebounds per 48 minutes is not impacted by how much players see action. Among NBA players with at least 750 minutes played, Blair ranked sixth in rebounds per 48 minutes (16.9). He topped all NBA players in offensive rebounds per 48 minutes (6.43), and remember that he was a 20-year-old rookie!
Even the offensive rebounds per 48 minutes statistic gets impacted by opportunity. Some teams play at a faster pace than others, and some miss more shots. Their players have more opportunities to grab offensive boards. The Spurs played at a slower pace than most teams and had the NBA’s sixth-highest shooting percentage. So these factors hurt Blair, yet he still out-rebounded everybody at the offensive end.
The best metric to show Blair’s rebounding excellence is rebound rate, John Hollinger’s measurement for the percentage of missed shots that a player rebounds when he’s on the court. Blair had a 16.0 offensive rebound rate last season. To put that in perspective, NBA teams grab 26-27 percent of available offensive boards on average. The Golden State Warriors had an offensive rebound rate of 20.9. Blair fell just 4.9 short of that figure, by himself.
While playing his final season at Pittsburgh, Blair put up unbelievable stats in this category. Despite playing in the rugged Big East, his 23.6 offensive rebound rate topped the nation’s next closest player by 5.0. Blair even surpassed the team figure for six Division I colleges.
So how did a player – who can out-rebound an entire team – last until the 37th pick of the 2009 NBA Draft? It’s hard to say. Blair’s 2008-09 rebounds per game figure (12.3) looked good but unspectacular, which may have been a factor. Of course, he played only 27.3 minutes per game on a very slow-paced team. Only adjusting his numbers for opportunity made Blair stand out.
Many still fixate on per game numbers, and they don’t begin to tell the story for players like DeJuan Blair. His 7.8 points per game and 6.4 rebounds per game in 2009-10 look pedestrian. However, Blair posted these numbers in limited opportunities – playing just 18.2 minutes per contest.
Rebounds per 48 minutes is not impacted by how much players see action. Among NBA players with at least 750 minutes played, Blair ranked sixth in rebounds per 48 minutes (16.9). He topped all NBA players in offensive rebounds per 48 minutes (6.43), and remember that he was a 20-year-old rookie!
Even the offensive rebounds per 48 minutes statistic gets impacted by opportunity. Some teams play at a faster pace than others, and some miss more shots. Their players have more opportunities to grab offensive boards. The Spurs played at a slower pace than most teams and had the NBA’s sixth-highest shooting percentage. So these factors hurt Blair, yet he still out-rebounded everybody at the offensive end.
The best metric to show Blair’s rebounding excellence is rebound rate, John Hollinger’s measurement for the percentage of missed shots that a player rebounds when he’s on the court. Blair had a 16.0 offensive rebound rate last season. To put that in perspective, NBA teams grab 26-27 percent of available offensive boards on average. The Golden State Warriors had an offensive rebound rate of 20.9. Blair fell just 4.9 short of that figure, by himself.
While playing his final season at Pittsburgh, Blair put up unbelievable stats in this category. Despite playing in the rugged Big East, his 23.6 offensive rebound rate topped the nation’s next closest player by 5.0. Blair even surpassed the team figure for six Division I colleges.
So how did a player – who can out-rebound an entire team – last until the 37th pick of the 2009 NBA Draft? It’s hard to say. Blair’s 2008-09 rebounds per game figure (12.3) looked good but unspectacular, which may have been a factor. Of course, he played only 27.3 minutes per game on a very slow-paced team. Only adjusting his numbers for opportunity made Blair stand out.
Tuesday, September 14, 2010
Caution: Falling Offense
Remember 1992? That was the last time National League offense had gone lower than the current level of 4.36 runs per game. The same goes for the American League, which has seen an even sharper scoring drop-off since last season. AL teams averaged 4.82 runs per game in 2009. That figure had plunged to 4.45 through September 14. The NL had a more gradual decline from 4.43 runs per game last year to 4.36.
This presents a challenge for agents with arbitration-eligible and free agent position players this offseason. Clubs will no doubt pull out comparables from recent seasons when the run context was substantially higher.
Fortunately, there is a solution. Agents can adjust for the decreased offense in the same way economists do so for inflation. The Sports Resource has built a statistical model that adjusts for run context, which helps your position player clients when scoring drops.
You can even turn the scoring trend into a positive for hitters: some of this season’s individual achievements will stand out even more at contract time. For example, should Jose Bautista reach 50 home runs, he will match a feat last accomplished in 1990. Look for another post on this topic in the weeks ahead.
This presents a challenge for agents with arbitration-eligible and free agent position players this offseason. Clubs will no doubt pull out comparables from recent seasons when the run context was substantially higher.
Fortunately, there is a solution. Agents can adjust for the decreased offense in the same way economists do so for inflation. The Sports Resource has built a statistical model that adjusts for run context, which helps your position player clients when scoring drops.
You can even turn the scoring trend into a positive for hitters: some of this season’s individual achievements will stand out even more at contract time. For example, should Jose Bautista reach 50 home runs, he will match a feat last accomplished in 1990. Look for another post on this topic in the weeks ahead.
Subscribe to:
Posts (Atom)
