Scoring speed: How players grade, perform

Attempts to measure speed go back millennia, when you think about it. One of the oldest competitions, perhaps the oldest competition, mankind has known is the standard race. Start here, end there, whoever gets there first wins and is definitively faster. 

In the days since Neolithic times, we've made speed a little more complicated by inventing other competitions and sports. In most of them, it's not good enough simply to outrun your opponent. You need to kick a ball into a net, put a ball into the hoop or hit a ball first in order to make your speed truly useful. Sure, there's still track and field, and one of the world's great true speed events is the 100-meter dash, but the skills of the fastest runners don't necessarily translate to other sports. (World's fastest man Usain Bolt has long made his wishes to play for Manchester United known, but the club doesn't seem like it will be picking up the phone anytime soon, at least not for a serious tryout.)

In sports like baseball, in which other objectives often get in the way, we try to look for other methods to measure just who is fastest. Scouts time hitters' run times from home to first and use those, along with other hard data as well as their eyes, to give players a 20-80 grade on their run tools. The fastest of the fast get a perfect 80. The slowest of the slow get a 20. Those grades can be subjective, though.

On the other side of the coin, interested parties not at the games might turn to stolen bases and triples to determine speed. Those numbers, of course, are more about performance than just straight skill and, on their own, don't tell the whole story. (Jason Varitek stole 10 bases as a 32-year-old catcher in 2004, but no one has ever made the mistake of calling him especially fast.)

There is a stat that does its best to show us just how fast -- perhaps it's more appropriate to say effectively fast -- a player is without using a stopwatch. It's Speed (Spd) Score, a 0-10 scaled stat created by Bill James in 1987 that combines components centered around steals, triples and runs scored to tell us just how good a player is on the basepaths.

Now, it'd be simple to just use these scores to tell you which of MLB.com's recently released top 100 prospects were fast in 2014 and which weren't. But we already have a good idea of that based on their run tool grades. Instead, what I thought would be more interesting would be to compare how prospects' speed actually played in games versus what's expected of them according to their run tool. 

But first, some necessary...

...Background

It's important that you understand how Spd works in order to grasp the whole basis for this here article. The version of Spd that I've used here takes the mean of four components: Stolen Base Percentage, Frequency of Stolen Base Attempts, Percentage of Triples and Runs Scored Percentage. You can find the exact formulas for each of these here (big hat tip to FanGraphs' David Appelman), but here's a brief explanation of each component. (Each of these components is placed on a 0-10 scale, which is explained within the link earlier in this paragraph.)

• Stolen Base Percentage: Pretty much exactly as it sounds. How good is the player at stealing a base once he decides to go?
• Frequency of Attempts: It's one thing to be perfect in stolen bases. But if perfect means 1-for-1, that doesn't tell us much. Speedy players are more likely to attempt to steal bags each time they get on base. This gives us an idea of just how often they do take off.
• Percentage of Triples: If you've watched the game, you know guys with speed can sometimes turn singles into doubles, doubles into triples, triples into inside-the-park homers. Triples are most closely associated with speed and more frequent than inside-the-parkers, so this measures how often a player turns balls in play into triples.
• Runs Scored Percentage: By the same token as above, speedy players are more likely to wheel around the bases and score runs. If you've seen a slow player get stopped at third for fear he'll be gunned down at home, you know from whence I speak.

And that is how you arrive at Spd. Although it is a 0-10 scale, average tends to be about 4.5 with an excellent score being in the 7.0 region or above and terrible being 2.0 or below. These were the top and bottom five Spds among MLB.com's top 100 prospects last season. The values for all top-100 prospects can be found in a handy table here. 

Those tables shouldn't be much of a surprise to anyone who followed the Minors in 2014. Pompey and Peraza used their legs to put together breakout campaigns and now find themselves in line for starting Major League gigs at some point in 2015, probably on Opening Day in the case of Pompey. Speed was just one reason why the A's wanted Barreto in the Josh Donaldson deal with the Blue Jays, and it's also a reason, along with defense and youth, why Mondesi is still so highly ranked after a difficult season (.211 average, .610 OPS) at the plate as a teenager for Class A Advanced Wilmington. We'll get to Correa later.

As for the slowest of the slow, it's no shock to find three catchers -- one known for his bat (Plawecki), two for their defense (Hedges, O'Conner) -- at the bottom, given how little speed is expected from those who squat behind the plate for a living. The powerful sluggers Olson and Judge, who have 40 and 50 run tool grades, also have spots after stealing only three combined bases in three combined attempts and hitting relatively few triples across a full season's worth of at-bats.

Like I said, that's fairly cut-and-dry, so let's take it to another level.

Spd vs. Run tool grades

With the exception of Correa above, those players performed about as expected on the basepaths, given their scouting profile, but there's little fun in that. So what about those who over- or underperformed our expectations for them? How can we figure that out?

Three words: "mean" and "standard deviation."

Yes, let's go back to stats class -- personal shoutout to Professor Ginovyan from sophomore year -- to compare our data here.

Standard deviation is one unit sized, both positively and negatively, away from the mean of our set of data that encapsulates most of the data points. It's taken by getting the square root of the variance. (This is all shop talk, so it's understandable if you skip on to the results below.)

Put another way, the mean of the Speed Scores is 5.4 -- you'll notice that it's higher than a typical Spd average, which shouldn't be a surprise given our data includes top prospects here -- and the standard deviation is calculated to be 1.75. That means about 67 percent of our Spd data falls between 3.65 and 7.15, both of which are 1 standard deviation away from the mean. The mean and standard deviation for run tool grades were 50.3 and 11.48, respectively.

By focusing on how many standard deviations each player's Spds and run tool grades are from their means, we put both Spd and scouting grades into the same unit of measure, making them easier to compare. (Admittedly, I tried making Spd into a 20-80 scouting scale, but the relationship between Spd and tool grades is not as linear as I'd have hoped. C'est la vie.) The closer the the players' two standard deviations are to each other, the more a player performed to expectations. The wider the gap, the less they performed to expectations with positive numbers showing outperfomance and negatives underperformance.

With that, the numbers...

In order to keep the tables less cluttered, I didn't include all of the data that went into each Spd here, so time for a little context. As a catcher, not much is expected of Realmuto speedwise. But he did a good job of putting that perception aside, swiping a career-high 18 bases in 23 attempts at Double-A Jacksonville and adding six triples in 375 at-bats. According to Spd, he was much closer to a 60-grade runner on the field, a 20-point swing when compared to his MLB.com run grade. 

Correa, meanwhile, might be the most interesting of the bunch. The Astros shortstop is considered to be great at many things, namely hitting, power and arm strength, but his speed tool grades out as average. It's one reason why several think the 6-foot-4 20-year-old could move to third base down the line. But before a broken fibula ended Correa's season in June, Houston's top prospect was proving to be a handful for opposing California League clubs on the bases. He was 20-for-24 in stolen bases and added six triples in only 62 games.This can be taken with a grain of salt, since it came at the Class A Advanced level, but if Correa can exceed his believed speed limitations at the upper levels, his already high ceiling could move even higher.

If we're going to use grains of salt for Correa in the Cal League, we might as well use the whole shaker for Devers, who played at the Dominican Summer and Gulf Coast Leagues in 2014. Barreto has plus speed but performed more like a 70-grade runner as an 18-year-old in the Northwest League. As a 30-grade runner according to scouts, Franco's expectations were so low that even a 3-for-4 stolen-base rate and a career-high four triples at Triple-A Lehigh Valley were enough to overperform.

Two themes of the set of five above: high expectations and injuries. 

Meadows was limited by hamstring injuries to only 45 games in his first full season in the Pirates system, so when he did make his 2014 debut on June 30, testing his legs again on the basepaths wasn't a primary objective. As such, he attemped to steal only five bases and was successful in only two of those attempts. With his health hopefully fully restored in 2015, his above-average speed should showcase better in 2015.

Hamstring injuries also held back Soler, who has had chronic problems with those muscles and caused the Cubs to develop a new workout system for him. The Cuban outfielder played in only 62 games for Cubs affiliates in 2014 and was unsuccessful in his only stolen base attempt. 

Turner and Buxton are two of the three prospects to receive a 75 or better scouting grade for speed, so even when their performance was indeed above-average -- Buxton, in particular, had a Spd equivalent to that of a 65-grade runner -- it still looks underwhelming on paper. Like Meadows, Buxton missed most of 2014 due to injuries, including a reaggravated wrist injury that occured on a slide into a base in May, that limited his sample significantly. 

Quick notes
 

• Forty-four of the 52 position players (about 85 percent) in MLB.com's ranking of top 100 prospects fell within 1.0 standard deviation difference of their expected Spd when compared to their run tool. What does that tell us? These scouting grades are pretty good -- that's what.
• Five players (Nomar Mazara, Michael Taylor, Christian Bethancourt, Alex Jackson and Stephen Piscotty) performed exactly as expected in terms of speed. That is to say, the difference between their Spd standard deviation and run grade standard deviation rounded to 0.0.
• Included in the table along with this story are positional breakdowns. As expected, shortstops proved to be the fastest of the bunch, putting together an average 6.3 Spd. Catchers (4.2) and first basemen (4.0, only two in the sample) were slowest.