Speed Score

We’ve added Bill James’ Speed Score to the player pages, leaderboards, and team pages.

We’re using the 4 component version, which is an average of Stolen Base Percentage, Frequency of Stolen Base Attempts, Percentage of Triples, and Runs Scored Percentage.

Print This Post

David Appelman is the creator of FanGraphs.

24 Responses to “Speed Score”

You can follow any responses to this entry through the RSS 2.0 feed.
  1. Decatur says:

    Apologies if you’ve answered this before, but a) are Speed Score’s incorporated into players win values and/or wOBA? and b) how much can adding speed scores to the site alter the valuation of a player, both in extreme cases and on average?

    Vote -1 Vote +1

  2. SB and CS are included in wOBA. When it comes to just Win Values, Speed Score isn’t something you’d want to add in there. You’d want just a base running component, which Speed Score isn’t.

    Overall on base running/stolen bases/caught stealing, you’re looking at a maximum of about a 1 win difference, but we already include SB and CS, so I’d guess you could potentially add/subtract about .5 wins with an additional base running component.

    Vote -1 Vote +1

  3. Also, Speed Score in this case is supposed to be more of a “skill” stat. I consider it similar to something like ISO, where it’s not something you’d convert to runs and add to a player’s Win Value because most of what it measures is already taken into account.

    But it can be useful to see which way it is trending to see if a player has “slowed down” in recent years.

    Vote -1 Vote +1

  4. Jerry says:

    I heart FanGraphs. Seriously.

    Vote -1 Vote +1

  5. David:

    Another cool thing added. I actually put together my own 4-component SS SQL, but the scores came out lower that I expected, although so far they seem to be about the same as yours. Perhaps, as in the “standard” version, the lowest number needs to be dropped before averaging?

    That’s meant as a question, not a criticism.

    Vote -1 Vote +1

    • I’m really not sure if you’re supposed to drop the lowest component in the 4 component version. From what I’ve read, I think the answer is no, but I could be wrong.

      I’m also not really trying to duplicate the 6 component version, so maybe they come out low compared to that, but the average is still around 5, which makes perfect sense to me on a 0-10 scale.

      Vote -1 Vote +1

      • D’oh. I think I thought the original was the 5-component version… Have to look back and see what I used/got, etc.

        Ah, you rightly leave out GIDP (which, I imagine, doesn’t make sense without opps.).

        Vote -1 Vote +1

  6. Toffer Peak says:

    Could you explain how you measured each of the four components (Percentage of Triples = 3Bs/ ABs? 2Bs? Hs?) and how they are weighed (z-scores?). Thanks.

    Vote -1 Vote +1

    • Stole Base Percentage = (((SB + 3)/(SB + CS + 7)) – .4) x 20

      Frequency of Attempts = Square root of ((SB + CS)/(1B + BB)) / .07

      Percentage of Triples = 3B/(AB-HR-SO) and then put on a scale based on this chart:

      < .001 = 0 .001 - .00231 = 1 .00231 - .00392 = 2 .00392 - .00583 = 3 .00583 - .00804 = 4 .00804 - .01055 = 5 .01055 - .0136 = 6 .0136 - .01587 = 7 .01587 - .01898 = 8 .01898 - .02239 = 9 > .02239 = 10

      Runs Scored Percentage = (((R – HR)/(H + BB – HR)) – .1) / .04

      Then anything one of the components that is below 0 you make 0, anything about 10 you make 10. Then take the mean of the 4 components.

      Vote -1 Vote +1

      • Toffer Peak says:

        Wow, that’s a lot different and more complicated than I would have imagined. I would have figured taking the SD above or below the mean for each of these measurements (possibly giving different weights to each of them) and adding them up would be a lot easier. Not sure if it would be any more accurate though of course. I wonder how he came to these weights/calculations. Is there a source online that explains his reasoning?

        Since lots of people claim that triples are simply doubles hit by fast players I would think 3B/2B would be a better measurement of speed. 3B/(AB-HR-SO) would seem to give too much weight to doubles hitters with speed as opposed to singles hitters with speed.

        Also wouldn’t you want to include HBPs in Frequency of Attempts?

        Thanks so much for the quick response.

        Vote -1 Vote +1

  7. David, great addition, but: Why is the runs scored percentage included? Obviously the hitter’s lineup will also influence this number, so it doesn’t really tell you what you want. If Jacoby Ellsbury was on the Giants (fewest runs scored this year), then his runs scored percentage would be lower, but of course his speed hasn’t actually changed.

    Vote -1 Vote +1

  8. dan says:

    Is it just me, or are the top speed scores way down from last years? I’m pretty much eye-balling the top 15, and they’re significantly lower than they were in 2008. And 2008 they’re slightly lower than in 2007.

    Vote -1 Vote +1

    • Ron says:

      I think at least part of the difference between this year and last has to do with the small sample size so far since Stolen Base Percentage is given as a counting stat. Extrapolating Michael Bourn’s stolen base production over a 162 game schedule, for example, would boost his SBP score from 4.9 to 7.2 and his Speed Score from 7.8 to 8.4.

      It’s small, I know, but if you add just a few tenths to everybody’s score I think you’ll find this year’s leaderboard looking a lot like last year’s, or you could give it a few months and it’ll probably catch up on its own.

      Vote -1 Vote +1

  9. dan says:

    I also think it would make sense to adjust the triples % and runs scored % for ballpark and lineup. Some parks are tailored for triples and obviously some lineups score more than others. The difference probably wouldn’t be huge.

    Scaling runs scored % to team runs per game would work, and using this article below (used by statcorner) would do the trick for triples.

    Vote -1 Vote +1

  10. I don’t disagree with any of the methodology comments and I’ll look into implementing what I can, but as to why certain things are the way they are, I didn’t create this stat, but in general I’ve found it useful as a speed gauge as is.

    Can it be made better, sure, but it’ll do the trick in a pinch until I (or someone else) wants to put in the heavy lifting.

    Dan, at least on average, speed score is almost exactly the same as it was last year. successful stolen bases are down about 1 percent to the lowest in 3 years. That could have something to do with it.

    Vote -1 Vote +1

  11. pm says:

    I would also add an Infield hit and GIDP component to this. That would make it more accurate.

    Vote -1 Vote +1

  12. Hey David,

    Can we get ISO against on this site? I know that doubles and triples have more luck (due to factors like fielder positioning) involved in their affect on ERA and the likes, but a guy who doesnt give up a lot of HRs but does consistently give up doubles will have a better FIP than a guy who gives up HRs but weak contact otherwise.

    I think it would be a great additional stat for the site!

    Vote -1 Vote +1

    • dan says:

      “a guy who gives up HRs but weak contact otherwise.”

      Such a player doesn’t really exist. Well ok that might not be 100% true. Players who may exhibit this trait are likely the result of noise (random chance) than underlying skill. What you’re proposing would probably fluctuate even more wildly than BABIP.

      Vote -1 Vote +1

  13. Bodhizefa says:

    I’d be curious to know if there was a good way to add bunt hits into the score. After watching Trout beat out what looked like a mediocre run-of-the-mill bunt last night, it got me wondering why bunts aren’t a part of the speed score calculation.

    Vote -1 Vote +1