Touted as one of the most revolutionary concepts in the history of baseball, the Moneyball method paved the way for relatively smaller teams to level the playing field against financial giants of the game.

Having first hit the mainstream in 2003 with the release of Michael Lewis’s book Moneyball: The Art of Winning an Unfair Game, the idea gained even more traction following the success of Oscar nominated movie Moneyball (released in 2011) starring Brad Pitt and Jonah Hill.

Lewis’s book centers around Oakland Athletics, a baseball team which ran on a shoestring budget compared to some of the big-market teams of the time such as the New York Yankees, Boston Red Sox, San Francisco Giants and Philadelphia Phillies. To put things into perspective, the Yankees had a payroll of $140 million in 2002 whereas Oakland was surviving on a mere $40 million.

To overcome this gargantuan financial disparity, Oakland’s manager Billy Beane (played by Brad Pitt in the movie) sought the help of data and statistics to find underrated players at a lower price and created a unique team of world beaters. The entire strategy essentially revolved around finding hidden gems. This concept came to be known as “Moneyball” as described in the book.

Today, every major sport team adopts the use of data analytics to refine the team’s diet, tactics, training, medical treatment and recovery to gain an advantage over the rest. With the ability to collect and analyze data almost in real-time, such an approach has become quite lucrative.

Historically, managers and scouts in sport teams picked players based on the “eye-test” and gut feeling which stemmed from years of experience and intuition from either playing the game or reporting on it. The advent of data analytics made this approach less desirable.

As described in Moneyball, a lot of in-game statistics are used to create a model which optimize for metrics such as runs which in turn guarantees you more wins. Some of these in-game variables could be hits, batting average, etc.

However, where this statistical model falls short is its inability to consider key intangible variables such as focus, anxiety, aggression, mood, form, positional sense, anticipation, decision-making etc. In my opinion, the intangibles are far more important than the empirical variables such as hits and batting average.

Yes, a few of the intangibles can be accounted for today by observing other variables such as amount of sleep, diet, exercise and recovery. But mental attributes that affect in-game performance are largely misunderstood.

That’s why so many big teams still fail in important competitions while smaller teams are able to cause upsets. It’s the randomness of unaccounted mental attributes that causes anomalous results on such a regular basis in essentially every sport.

Moneyball statistical model

Billy Beane adopted statistical concepts from pioneering baseball analyst Billy James, who was responsible for coining the term SABRmetrics which laid the foundation for the Moneyball approach. SABR stands for Society for American Baseball Research.

Beane was responsible for putting the concept into play in a practical sense with the Oakland Athletics.

In the book, Lewis argues that traditional indicators of success in baseball such as hits and batting average were not useful. Instead, he made a case for other variables such as OBP (on-base percentage), slug and OBS (on-base slugging percentage) which tended to be better indicators of success.

I analyzed this data using a Moneyball dataset and a linear regression model. The data consisted of 12 metrics across 30 MLB (Major League Baseball) teams. The goal was to determine which variables best predicted the “runs” variable. Runs was picked by Beane to be the best indicator of success for Oakland and was therefore the target variable for prediction.

The analysis can be found on my Github.

Based on the analysis, the newer variables referenced by Lewis in his book clearly outperformed the more traditional ones. The bar chart below shows the metrics with the corresponding R-squared values as a result of separate linear regression models for each variable. On-base plus slug, slugging, and on-base percentage all recorded values greater than 0.8 while the others had values less than 0.66.

Why were only the above six variables analyzed? This is because some of the other initial variables in the dataset had a low correlation to the “runs” target variable as shown in the correlation matrix below. Hence, they were excluded from the analysis altogether.

Furthermore, it didn’t make sense to include variables such as strikeouts or stolen bases in the analysis as they were mostly pitcher attributes and inherently had no correlation to runs.

To provide some context regarding these metrics, here are the definitions of the variables and supporting variables used in the models:

• hits: Occurs when a batter makes contact with a pitch and the batter reaches base while the hit stays in fair territory. The different types of hits are singles, doubles, triples and home runs. This does not take into account fielder’s choice or errors.
• at_bats: Recorded when a batter reaches base via fielder’s choice, hit or an error, or when a batter is put out on a non-sacrifice.
• bat_avg: Calculated by dividing a player’s hits by total at_bats. The bat_avg is typically between 0 and 1 with the historical average value being around 0.25.
• homeruns: When a batter hits the ball over the fence and is able to run to all four bases at their leisure as there is no threat of being thrown out. All batters on the field also receive the runs.
• new_onbase: Refers to how frequently a batter reaches base per plate appearance. Times on base include hits, walks and hit-by-pitches, but do not include errors, times reached on a fielder’s choice or a dropped third strike.
• new_slug: Total number of bases a player records per at bat. Unlike on-base percentage, slugging percentage deals only with hits and does not include walks and hit-by-pitches in its equation.
• new_obs: OPS adds on-base percentage and slugging percentage to get one number that unites the two. It’s meant to combine how well a hitter can reach base, with how well he can hit for average and for power.

If you are an avid baseball fan, the meaning of the above variables will be second nature to you. However, it took me a while to wrap my head around these concepts.

For a comprehensive list of definitions for the various metrics used in baseball, see this link.

Success during early 2000s

Despite the financial disadvantages of the Athletics, Beane and his team of misfits managed to pull off brilliant performances season after season, most of which can be attributed to the SABRmetrics approach.

The years spanning 2000 to 2006 came to be known as the “Moneyball era”, a period of time when the A’s reached the post-season playoffs five out of seven times. Unfortunately, in four of those post-season appearances, the A’s lost in the Division Series itself which is the first stage of the post-season.

In 2006 however, Oakland did make it past the Division Series by beating the Minnesota Twins 3-0 only to get swept by the Detroit Tigers 4-0 in the next round. The win over Minnesota proved to be the only playoff series victory for the A’s during Beane’s 20-year tenure.

Flaws of Moneyball

So why were the A’s not able to progress past the early stages of post-season on a regular basis despite performing well during the regular season?

There could be several reasons for this.

One of the reasons is that over a long period of time (regular season consisting of 162 games) a tried and tested method generally wins out. It’s the same reason why a statistical approach gets more and more accurate as you increase the sample size. However, if you simply take a smaller sample size (playoff series consisting of 3-4 games) any result is possible due to randomness.

As described earlier, randomness in sport can be attributed to many things such as:

• Nerves
• Bad weather
• Lack of experience
• Discipline
• Crowd pressure
• Focus
• Preparation
• Hunger
• Lack of sleep
• Discipline
• Personal life

These are essentially things that cannot be quantified easily, or at least back then in the 2000s.

Therefore, the Moneyball method always needs to be complimented by controlling other factors such as nerves, focus, pressure, sleep, etc., attributes that were not considered in the SABRmetrics models. These play a much greater role in the playoffs due to the high stakes nature of this stage as compared to the regular season.

Even in the movie, a parallel story runs which shows a younger Billy Beane rejecting a scholarship opportunity at Stanford to pursue a baseball career but never being able to live up to his true potential which was evident at a younger age. While he was clearly one of the most promising players during his teens, he couldn’t make it in the big leagues due to mental attributes which were not quantifiable.

Mental attributes are critical in every sport and are the hardest to quantify.

In soccer, for example, one of the best club teams in the world Manchester City have won the last four out of five English Premier League titles (arguably the toughest domestic league in the world). However, during the same period, they have always seemed to falter in the knockout stages of the continental competition, the UEFA Champions League (UCL).

Their European exits in the last five years include a final, a semi-final and three quarter-finals. If we go back a further four years, they were knocked out in the round of 16 on three occasions and in the semifinals on another occasion. During these four years, they managed to win the domestic league title twice, where each team plays 38 games (larger sample size).

Each knockout stage of the UCL is played over two legs so a team needs to be sharp and focused during both legs to progress. The final, on the other hand, consists of a single game which makes the effects of randomness even stronger. Similar to the A’s, City performed brilliantly during the group stages of the UCL, sweeping away their opponents with relative ease. A lot of that can be attributed to a lack of pressure and nerves during this stage of the competition.

Another reason why the A’s performed better in regular season is that they were potentially playing less skilled opponents at that stage. Having a larger gap in quality between two teams can reduce the effects of randomness. This allowed the A’s to finish first on multiple occasions. Even then, randomness still existed which possibly owed to Oakland’s two second place finishes during the Moneyball period.

The ability to find hidden gems means that it must be done discretely to avoid the big teams becoming aware of the situation and creating more demand whereby raising the asking price of the player. Due to widespread scouting networks and a ton of information available today, it is more difficult to stay discrete.

These hidden gems are not necessarily the cream of the crop. Yes, they are great business deals, but they won’t guarantee the best performance on the pitch. They are likely to perform significantly better than the price they demanded but they have a relatively low ceiling when it comes to raw talent and mentality.

Truly great players are still going to cost a fortune which only the big market teams can afford. These are the kinds of players that win you championships.

It’s difficult to keep these superstars of the game discrete and as a result are picked up by the big teams with ease. Once in a while one might be able to find a hidden gem who is truly world class, but the rest can only perform to a certain level.

Combining analytics with intuition

How can the key intangible variables be better quantified? 

It’s all through better analytics which is possible today. Every detail of a player’s workout, diet, sleep, recovery, health, etc. is taken into account nowadays which allows them to perform better on the field.

Even then, some intangibles remain mostly misunderstood such as nerves, focus, determination, work ethic, mental toughness. Since these are inherently difficult to quantify, they need to be judged through intuition and gut feeling. This is where former players and coaches come into play who can provide their expertise based on the “eye-test”.

In the Moneyball movie, both Brad Pitt and Jonah Hill’s characters laughed off the intuitions of the existing scouts in picking new players. The truth is that this intuition stems from years of experience in the game and they probably know factors that cannot be quantified by a statistical model. Their intuition is based on subconscious data science, the kind all humans develop through experience.

At the end of the day, data needs to be combined with intuition to have a recipe for success, or something close to it, simply because we either don’t have the ability to quantify all the key factors or we don’t have enough data for our models to make the most informed decisions.

Hope this sheds some light on the merits of incorporating analytics into sport and its drawbacks that managers need to be aware of. Make sure to comment whether you think the Moneyball method can be used for sustainably winning championships or if you think intuition is always going to be more important for success in sports. Cheers!