Big ideas in sports analytics and statistical tools for their investigation

Probably not a substantial claim: may not pass S3 (load-bearing).

Fernandez et al. (2021) implement deep learning methods to examine the instantaneous expected value of soccer possessions.

“Fernandez et al. (2021) implement deep learning methods to examine the instantaneous expected value of soccer possessions.”

• cites Fernandez et al. (2021) sourceSupported

May combine two claims.

Yurko et al. (2019)'s win probability model is fully reproducible and uses publicly available data.

“A vital highlight of Yurko et al. (2019)'s win probability model is that it is fully reproducible and uses publicly available data.”

• cites Yurko et al. (2019) sourceSupported

May combine two claims.

In the example with runners on first and third, tied bottom ninth, no one out, the expected number of runs scored in the remainder of the inning is 1.94 and the probability of scoring zero runs is 0.13.

“Information derived from Table 1 reveals that the expected number of runs scored in the remainder of the inning is 1.94 runs, while the probability of scoring zero runs is 0.13.”

• AssumptionRests on an assumption: For a state with runners on first and third and no outs in a tied bottom ninth, the probability of scoring zero runs is 0.13.

Probably not a substantial claim: may not pass S3 (load-bearing).

The EPV model estimate of the probability of Chalmers making the three-point shot in the example is 42.5%.

“Chalmers' shot is worth 3 points, so the EPV of 1.276 points implies that the model estimate of the probability of Chalmers making this shot is 42.5%.”

May combine two claims.

In that example, by walking the hitter to load the bases, the defense reduces the chances of scoring to 82%, even though they raise the expected number of runs scored to 2.22.

“However, by walking the hitter to load the bases, they create the opportunity to force the lead runner at home and thus reduce the chances of scoring to 82%, even though they raise the expected number of runs scored to 2.22.”

• AssumptionRests on an assumption: For a state with bases loaded and no outs in a tied bottom ninth, the probability of scoring at least one run is 82%.

Probably not a substantial claim: may not pass S3 (load-bearing).

In the illustrative chess game, the author was rated 1732 and his opponent was rated 1683 prior to the game.

“Prior to the game, the author was rated 1732, whereas his opponent was rated 1683.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The post-game Elo rating for the author in the illustrative chess game is 1740.

“The post-game Elo rating for the author is thus R Ã a ¼ 1732 þ 20 1 À 0:591 ð Þ¼1740:”

Probably not a substantial claim: may not pass S3 (load-bearing).

The probability of the author defeating their opponent in the illustrative chess game was 0.591.

“The probability of the author (a) defeating their opponent (b) was Π ab ¼ 1 1 þ 20 1683À1732 ð Þ =400 ¼ 0:591:”

May combine two claims.

The sum of implied probabilities from moneylines is greater than one, which is why the sportsbook makes money regardless of who wins the championship.

“The sum of those probabilities is greater than one-this is why the sportsbook makes money regardless of who wins the championship.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims. The quote only partly states this claim.

In the FanDuel futures market on January 9, 2023, the Boston Celtics have a 16.3% chance of winning the 2023 NBA championship, and the Milwaukee Bucks have a 13.3% chance.

“In this case, the FanDuel futures market suggests that the Celtics have a 16.3% chance of winning the championship, while the Milwaukee Bucks have the second best chance, at 13.3%.”

May combine two claims.

In American football, teams are going for it on fourth down far more often than in the past, a direct result of statistical analysis showing that most teams were previously overly conservative.

“In American football, teams are going for it on fourth down far more often than in the past, a direct result of statistical analysis showing that most teams were previously overly conservative (Lopez, 2020).”

• cites Lopez, 2020Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Four big ideas are common across multiple sports: the expected value of a game state, win probability, measures of team strength, and the use of sports betting market data.

“In this article, we connect four big ideas that are common across multiple sports: the expected value of a game state, win probability, measures of team strength, and the use of sports betting market data.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Similar insights are now being made in less obvious settings, including esports.

“Similar insights are now being made in less obvious settings, including esports (Clark et al., 2020;Maymin, 2021).”

• cites Clark et al., 2020Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Maymin, 2021Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Sports analytics has expanded its footprint both in the professional sports industry and in academia over the past 30 years.

“Sports analytics-broadly defined as the pursuit of improvement in athletic performance through the analysis of data-has expanded its footprint both in the professional sports industry and in academia over the past 30 years.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

In baseball, teams are using defensive shifts to maximize the probability of recording an out, encouraging hitters to improve their launch angles, and optimizing pitcher repertoires to minimize contact.

“And, of course, in baseball, teams are using defensive shifts to maximize the probability of recording an out, encouraging hitters to improve their launch angles, and optimizing pitcher repertoires to minimize contact (Healey, 2017).”

• cites Healey, 2017Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

In basketball, analytics has driven a shift in conventional wisdom about shot selection, with most teams shooting more three-pointers, settling for fewer long two-point shots, deploying more versatile defenders, and relying less on pounding the ball into the paint.

“In basketball, analytics has driven a shift in the conventional wisdom about shot selection.Most teams are shooting more three-pointers, settling for fewer long two-point shots, deploying more versatile defenders, and relying less on the strategy of pounding the ball into the paint in an attempt to get a highpercentage shot (Schuhmann, 2021).”

• cites Schuhmann, 2021Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

High-resolution tracking data allows researchers to produce advanced analyses of captured spatiotemporal information and better understand the game.

“This high-resolution data allows researchers to produce advanced analyses of the captured spatiotemporal information and better understand the game.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The strategy of maximizing expected points in basketball has led directly to the proliferation of three-point shooting in the NBA.

“The strategy of maximizing expected points in basketball has led directly to the proliferation of three-point shooting in the NBA.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Generally, the goal of any sport is to score more points than the other team, which most often means trying to score as many points as possible, leading to a general strategy of maximizing expected points.

“Generally, the goal of any sport is to score more points than the other team, which most often means trying to score as many points as possible, leading to a general strategy of maximizing expected points.”

May combine two claims.

Maximizing the probability of scoring is optimal in any sudden-death situation, which has (but currently does not) included overtime in American football.

“Maximizing the probability of scoring is optimal in any sudden-death situation, which has (but currently does not) included overtime in American football (Martin et al., 2018).”

• cites Martin et al., 2018Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing).

Carroll et al. (1988) determine that every extra 25 yards is associated with 2 more points scored on average for a football team.

“They determine that every extra 25 yards is associated with 2 more points scored on average for a football team.”

• cites Carroll et al. (1988)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Baseball and cricket differ in strategies such as batting order or sacrifice plays.

“For example, despite belonging to the same bat-and-ball family, baseball and cricket differ in strategies such as batting order or sacrifice plays.”

May combine two claims.

Tango et al. (2007) conclude that the sacrifice bunt is rarely worth it, because most teams would be expected to score more runs with a runner on first and no outs than they would with a runner on second and one out.

“Tango et al. (2007) (and many subsequent analyses) conclude that the sacrifice bunt is rarely worth it, because most teams would be expected to score more runs with a runner on first and no outs than they would with a runner on second and one out.”

• cites Tango et al. (2007)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

In the bottom of the ninth inning of a tied baseball game, the optimal strategy for winning the game is maximizing the probability of scoring at least one run, which may differ from the strategy of maximizing expected runs.

“For example, in the bottom of the ninth inning of a tied baseball game, the optimal strategy for winning the game is maximizing the probability of scoring at least one run, which may differ from the strategy of maximizing expected runs.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Cervone et al. (2014, 2016) introduce expected possession value (EPV) as a continuous-time estimate of the expected number of points for the offensive team on a given possession using player and ball locations.

“In basketball, Cervone et al. (2014) and Cervone et al. (2016) introduce expected possession value (EPV) as a means toward an assessment of a player's on-court performance.This metric is a continuous-time estimate of the expected number of points for the offensive team on a given possession using player and ball locations.”

• cites Cervone et al. (2014)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Cervone et al. (2016) sourceNot checked yet

  Checked before — its earlier verdict is held back until a new check runs.

May combine two claims.

Beaudoin and Swartz (2010) show that NHL coaches do not always employ the optimal strategies, usually by waiting too long to pull their goalies.

“Beaudoin and Swartz (2010) show that NHL coaches do not always employ the optimal strategies, usually by waiting too long to pull their goalies.”

• cites Beaudoin and Swartz (2010)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Classified as probably not a substantial claim.

Romer (2006) concluded that teams do not go for it enough if teams are trying to maximize their probability of winning the game.

“They concluded that teams do not go for it enough if teams are trying to maximize their probability of winning the game.”

• cites Romer (2006) sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The expected run matrix is foundational in baseball analytics.

“The collection of estimates  Xjs ½ for all 24 states is called the expected run matrix, 2 and it is foundational in baseball analytics.”

Probably not a substantial claim: may not pass S4 (based).

Over the past decade or so, professional sports leagues have collected tracking data which record the locations of all players and the ball (or puck) throughout a game.

“Over the past decade or so, professional sports leagues have collected tracking data which record the locations of all players and the ball (or puck) throughout a game.”

Classified as probably not a substantial claim. May combine two claims.

Lopez (2020) concludes that coaches are too conservative on 4th down decision making, but notes further that past estimates of the magnitude of how conservative coaches are on 4th down may be overstated due to the way in which to go yardage was recorded only approximately in the past.

“Lopez (2020) also concludes that coaches are too conservative on 4th down decision making, but notes further that past estimates of the magnitude of how conservative coaches are on 4th down may be overstated due to the way in which to go yardage was recorded only approximately in the past.”

• cites Lopez (2020)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing).

Alamar (2010) determines that teams are under-utilizing passing plays in some situations.

“Based on play-by-play data from 2005 through 2008, they determine that teams are under-utilizing passing plays in some situations.”

• cites Alamar (2010)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing).

Urschel and Zhuang (2011) conclude that coaches in the NFL tend to make conservative decisions.

“Using data from the 2009 NFL season, they conclude, as many have, that coaches in the NFL tend to make conservative decisions.”

• cites Urschel and Zhuang (2011) sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

In a low run-scoring environment where hits are hard to come by, stolen bases and sacrifice bunts may be comparatively more valuable.

“Conversely, in a low run-scoring environment where hits are hard to come by, stolen bases and sacrifice bunts may be comparatively more valuable.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Macdonald (2012) uses expected goals to evaluate ice hockey players, but does not have access to player tracking data necessary to evaluate possessions.

“Macdonald (2012) uses expected goals to evaluate ice hockey players, but does not have access to player tracking data necessary to evaluate possessions.”

• cites Macdonald (2012)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing).

Deshpande and Jensen (2016) extend Stern (1994)'s framework by applying it in a Bayesian setting.

“On a related note, Deshpande and Jensen (2016) extend Stern (1994)'s framework by applying it in a Bayesian setting.”

• cites Deshpande and Jensen (2016)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Yurko et al. (2019)'s work inspired nflfastR (Carl & Baldwin, 2022), now considered the successor to nflscrapR.

“One of Yurko et al. (2019)'s goals was also to encourage researchers to “use, explore, and improve upon our work,” which ultimately inspired nflfastR (Carl & Baldwin, 2022), now considered the successor to nflscrapR.”

• cites Carl & Baldwin, 2022Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing).

Brenzel et al. (2019) use three-dimensional Markov models to estimate win probability throughout a curling match.

“Brenzel et al. (2019) use three-dimensional Markov models to estimate win probability throughout a curling match.”

• cites Brenzel et al. (2019)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing).

The Bradley-Terry model is perhaps the most widely-used probability model for predicting the outcome of a paired comparison.

“Perhaps the most widely-used probability model for predicting the outcome of a paired comparison is the Bradley-Terry model (BTM) (Bradley & Terry, 1952).”

• cites Bradley & Terry, 1952 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

The Elo rating system was originally developed for chess (Elo, 1978).

“Another widely known tool for measuring team strength is the Elo rating system (Elo, 1978), which was originally developed for chess.”

• cites Elo, 1978Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing).

Expected winning percentage (Pythagorean expectation) was developed by James (2003).

“A simple method for estimating team strength that has become popular in sports analytics is expected winning percentage-often called Pythagorean expectation-developed by James (2003).”

• cites James (2003)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

Wins and losses carry some signal of team strength, but we can do better than win-loss records.

“As binary outcomes, and with all sports (except perhaps chess) involving some element of chance, wins and losses carry some signal of team strength, but we can do better.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The International Chess Federation (FIDE) uses K = 10 for players with any previously achieved rating of at least 2400.

“The International Chess Federation (FIDE) uses K ¼ 10 for players with any previously achieved rating of at least 2400.”

• cites FIDE, 2022Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

FIDE gives K = 40 to new players with under 30 games played, and players under the age of 18 with rating less than 2300.

“Finally, K ¼ 40 is given to new players with under 30 games played, and players under the age of 18 with rating less than 2300 (FIDE, 2022).”

• cites FIDE, 2022Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Some researchers have tried to demonstrate a positive return on betting strategies, with very limited success.

“Some researchers have then tried to demonstrate a positive return on some of these betting strategies, with very limited success.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Gandar et al. (1988) concludes that statistical tests fail to reject the hypothesis of rationality of NFL betting markets.

“Early work by Gandar et al. (1988) examines the rationality of NFL betting markets and concludes that statistical tests fail to reject the hypothesis of rationality.”

• cites Gandar et al. (1988) sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S4 (based).

The efficiency of betting market data in estimating true probabilities has been demonstrated time and time again.

“The efficiency of betting market data in this respect has been demonstrated time and time again.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The teamcolors R package provides color palettes and logos for men's and women's professional and collegiate sports teams, as well as color and fill scale functions compatible with ggplot2.

“The teamcolors R package (Baumer & Matthews, 2020) provides color palettes and logos for men's and women's professional and collegiate sports teams, as well as color and fill scale functions compatible with ggplot2 (Wickham et al., 2022).”

• cites Baumer & Matthews, 2020Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Wickham et al., 2022Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

Normalized implied probabilities are accurate, unbiased, and efficient estimates of the true unknowable probabilities.

“Many different researchers have shown that these normalized implied probabilities are accurate, unbiased, and efficient estimates of the true unknowable probabilities (see Lopez et al., 2018 for discussion and an extensive list of references).”

• cites Lopez et al., 2018 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

Metrick (1996) finds that bettors overback the heaviest favorites.

“Metrick (1996) finds that bettors overback the heaviest favorites.”

• cites Metrick (1996)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing).

The use of team colors to identify lines makes it possible to follow the trajectory of most teams over the course of the season in Figure 5.

“However, the use of team colors to identify the lines makes it possible to follow the trajectory of most teams over the course of the season.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Soebbing and Humphreys (2013) find evidence that sports bettors think tanking in the NBA is occurring, although the evidence for whether it actually is remains mixed.

“Soebbing and Humphreys (2013) find evidence that sports bettors think tanking in the NBA is occurring, although the evidence for whether it actually is remains mixed.”

• cites Soebbing and Humphreys (2013)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing).

The sportyR package contains generic playing surfaces for baseball, basketball, curling, American football, ice hockey, soccer, and tennis that can be added to ggplot graphics with a single function call.

“Thankfully, the sportyR package (Drucker, 2022) contains generic playing surfaces for baseball, basketball, curling, American football, ice hockey, soccer, and tennis that can be added to ggplot graphics with a single function call.”

• cites Drucker, 2022Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

Catcher framing is a concept that scouts talked about for decades, but that could not be quantified by analysts until data of the appropriate resolution became available.

“Catcher framing is a notable example of a concept that scouts talked about for decades, but that could not be quantified by analysts until data of the appropriate resolution became available.”

• cites Deshpande & Wyner, 2017Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

As a maturing discipline, sports analytics has been able to address fundamental questions common to many sports.

“However, as a maturing discipline, sports analytics has been able to address fundamental questions common to many sports.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

Play-by-play data allows us to make valuations between plays, while player tracking data allows us to make valuations within plays.

“While play-by-play data allows us to make valuations between plays, player tracking data allows us to make valuations within plays.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Over time, the resolution of baseball data has improved to include play-by-play data, pitch-by-pitch data, and now player tracking data.

“Over time, the resolution of baseball data has improved to include play-by-play data, pitch-by-pitch data, and now player tracking data.”

Probably not a substantial claim: may not pass S3 (load-bearing).

These R packages enable research by making data more easily available.

“Thus, these R packages enable research by making data more easily available.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

As the granularity of baseball data has evolved over time, so too have the statistical methodologies for modeling that data, and the tools for working with it.

“As the granularity of baseball data has evolved over time, so too have the statistical methodologies for modeling that data, and the tools for working with it.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based). May combine two claims.

These venues have been an important source of new ideas and have contributed to the diversification of the field by breaking down barriers to entry.

“These venues have been an important source of new ideas and have contributed to the diversification of the field by breaking down barriers to entry.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

The dynamic of increased data resolution catalyzing new research directions is playing out in all sports.

“this same dynamic is playing out in all sports”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

Increased resolution of available data has catalyzed new research directions in baseball.

“the increased resolution of available data has catalyzed new research directions in baseball”

Probably not a substantial claim: may not pass S3 (load-bearing).

Because R is scriptable, these packages make it easier to share research that is reproducible.

“Moreover, because R is scriptable, they make it easier to share research that is reproducible.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing), S4 (based).

As an applied science, sports analytics may lack a grand unified theory that succinctly characterizes game play across all sports.

“As an applied science, sports analytics may lack a grand unified theory that succinctly characterizes game play across all sports.”