Physicists have long been searching for the Theory of Everything, one equation that unifies the incomprehensibly big (general relativity) with the unimaginably small (quantum mechanics).

Fantasy baseball, too, has its Theory of Everything, one equation that can account for every conceivable facet of a player’s game to accurately analyze past performance and predict future returns. Like those genius physicists we haven’t discovered it yet, but bit by bit we’re getting there, and that’s due in large part to sabermetrics.

And that’s why sabermetrics are so appealing to me. I have an addiction to all of these crazy, obscure stats that most people have never heard of and the rest don’t know how to use. Because fantasy baseball is first and foremost a numbers game — unlike actual baseball where team chemistry is a huge factor — I’m a firm believer that we can use these obscure stats to turn the tables on our opponents.

One stat I’ve always been fascinated by because of its elegance is speed score.

Speed score was created by Bill James in the late 1980s to provide an on-the-surface method of evaluating a player’s speed. Fangraphs carries speed scores, and the version they use averages four separate components of a player’s speed:

1. Stolen base percentage
2. Frequency of stolen base attempts
3. Percentage of triples
4. Runs scored percentage

Note: If you’re wondering what these are, check out Fangraphs’ speed score explanation page. The comments are helpful.

So, does speed score have any practical use in fantasy baseball? In your standard 5×5 leagues, runs, batting average and steals are all influenced by a player’s speed. Faster players generally score more runs, post higher BABIPs (increasing their batting averages versus players with similar ball-in-play rates) and steal more bases. Runs scored is highly dependent on lineup so I’ll eliminate that from this discussion, but does speed score have any sort of predictive value for the other two stats?

Batting Average

There are six main outcomes every time a batter steps into the box:

1. Ground ball
2. Line drive
3. Fly ball
4. Strikeout
5. Error
6. Walk

Of those six, the first five affect a player’s batting average. Of those five, the first four happen with great frequency (errors happen but have a very small effect on a player’s average so I’ll eliminate them from consideration to keep things simple). Of those four, only the first one — ground ball rate — has a strong speed-dependent component.

Therefore, if speed score has any predictive value in determining a player’s batting average, it should show up in his batting average on ground balls with faster players posting higher averages.

To see if this was true, I analyzed all ground ball data from 2008 to 2010. Over that three-year span, the league average batting average on ground balls was .234 and the league average speed score was 4.0. To eliminate error caused by small sample sizes, my analysis only included players who hit at least 80 ground balls in a given year. This constraint narrowed down the data pool to just 382 player seasons (note that if a player hit at least 80 ground balls each year from 2008 through 2010 those are recorded as three separate player seasons).

Using analysis similar to that in our explanation of Javier Vazquez‘s recent success, I plotted each of the 382 points and generated the following graph:

There’s a clear positive trend here, showing that players with higher speed scores tend to have better batting averages on ground balls, but there was a great deal of variation which we see with an r-squared (coefficient of determination) value of just 0.144. This means that just 14.4 percent of the variation in batting average can be attributed to changes in a player’s speed score. The two are connected, but it’s a loose connection.

If we were to plug the league average speed score into the equation of the Excel-generated best-fit line, we see that a speed score of 4.0 corresponds to a batting average on ground balls of .244. While that’s a little higher than the actual league average over this three-year span (.234), it does give us a general margin of error which we can use to evaluate hitters.

Using Speed Score to Analyze/Predict Batting Average

This has been an unexpectedly down season for Ichiro Suzuki. He’s batting just .242 on ground balls, which the graph above shows corresponds to a speed score of around 4.0. Ichiro’s actual speed score this year is 6.0, 50 percent higher. According to the graph a speed score of 6.0 corresponds to a batting average on grounders of .261.

Since ground balls account for about 60 percent of all balls Ichiro puts into play, that 19-point difference between actual and expected average on grounders equates to about a 10-point difference in Ichiro’s overall average. Since Ichiro is batting .268 at the time of writing this, we could reasonably expect he’s been unlucky and his average should be pushing .280. That’s still not what we’d expect out of the best player Japan has ever produced, but .278 looks a lot better than .268.

As I said above, the data indicates there’s not a tremendous amount of correlation between speed score and average on grounders. For this reason I do believe speed score can be used as a means of evaluating a player’s batting average on ground balls (and subsequently their overall average), but I would hesitate before using speed score to predict a player’s expected batting average.

Stolen Bases

When evaluating batting average, a hitter’s speed was only a small part of the equation (since ground balls are only one of the four major outcomes of an at-bat). While there are other factors that influence a player’s stolen base totals — whether there are runners on base ahead of him, how capable he is of reading pitchers’ moves to the plate, his ability to get a good lead/jump, manager’s philosophy, etc. — a player’s speed certainly is central.

Using the same method as above, here’s how the graph for speed score versus stolen base rate (stolen bases per plate adjusted appearance…more on this in a minute) shook out:

The first thing you might notice about this graph is that the trend line is curved rather than linear. The r-squared value for the trend line you see above is 0.769 (whereas the linear trend line yielded an r-squared of just 0.701). This tells us that 76.9 percent of the variation in steal rate can be explained by the variation in speed score.

What does the exponential nature of this data tell us? Simply, the faster players are, the more and more likely they are to run.

The second thing you might notice is that steal rate was calculated as SB/(PA-3B-HR). When this article was originally published, I had just calculated steal rate as SB/PA, but loyal reader and commenter Tom suggested we adjust the plate appearances to see if that would improve our r-squared value. Sure enough, subtracting triples and home runs (events you cannot get steals off of save for the rare steal of home) improved the r-squared from 0.759 to 0.769. Thanks, Tom!

Of the 261 player seasons plotted above, 126 fell above the trend line (48.3%) and 135 fell below (51.7%). The average theoretical steal rate, derived by averaging the theoretical steal rates of every player in the league from 2008 to 2010, was 0.0188 steals per plate appearance. The average actual steal rate for that span was 0.0191 steals per plate appearance.

Clearly, there’s a high degree of correlation here.

Using Speed Score to Analyze/Predict Stolen Bases

Of course, stolen base total is highly dependent on opportunity, which is measured here in plate appearances. Batters with low steal rates but high speed scores should be good bets to increase their production. Batters that fit the opposite mold would obviously be good bets to decrease their production.

Here are some players who fit those descriptions where SS means speed score. Remember, this means increase or decrease production based on steal rate and not based on stolen base totals.

Players likely to increase stolen base production:

• Colby Rasmus – 5 SB, 6.5 SS
• Nyjer Morgan – 6 SB, 7.1 SS
• Denard Span – 4 SB, 5.4 SS

Players likely to decrease stolen base production:

• Bobby Abreu – 14 SB, 3.7 SS
• Mark Trumbo – 8 SB, 3.4 SS
• Orlando Hudson – 13 SB, 5.5 SS