Hitter’s Luck as a Rate Stat

On Monday, I introduced my Luck stat for pitchers and hitters. Today, I will look into some improvements to the hitter portion of the stat.

Moving Luck to a Rate Stat

The main focus for the stat was to try to find players under- or over-performing. The Luck stat I previously released was the accumulation of the luck a player had for the entire season. The problem was that two players with similar BABIPs and HR/FB ratios would be rated differently by the stat because of their PAs. I just divided the previous Luck value by PA and then readjusted the multiplier to get near -10 to +10 scale. Here is a comparison of old and new Luck leaders and laggards (min 150 PA):

Old Luck Leaders and Laggards:

Name Old Luck New Luck
Gonzalez Adrian 8.7 8.3
Bautista Jose 8.4 8.9
Stanton Mike 7.6 8.6
Morse Michael 7.2 9.0
Kemp Matt 6.7 7.2
Escobar Alcides -8.1 -5.3
Suzuki Ichiro -8.1 -4.4
Andrus Elvis -8.2 -4.9
Bartlett Jason -8.6 -5.6
Revere Ben -9.8 -9.6

New Luck Leaders and Laggards:

Name New Luck Old Luck
Jones Andruw 14.8 4.4
Jennings Desmond 11.3 3.2
Lillibridge Brent 10.9 3.4
Downs Matt 10.0 2.5
Johnson Reed 9.0 2.9
Revere Ben -9.6 -9.8
Stewart Chris -9.7 -4.6
Counsell Craig -9.8 -4.9
Forsythe Logan -11.3 -5.3
Young Jr. Eric -13.1 -5.7

The differences between the charts can be seen fairly quickly. With the old Luck values, players that are somewhat lucky get pushed to the top of the list because of their playing time. The second table has some players with extreme values because of a smaller number of PAs.

The one name that makes both lists is Ben Revere for being unlucky. His unluckiness is from having a 0.260 BABIP with an xBABIP of 0.349. Also, he has hit no home runs on 39 fly balls. After a 0.326 average in the minors, he has only able to hit for a 0.246 average in the majors. He may be looking to improve in 2012.

Home Run Rate Regression Changed

With the old Luck, I regressed the player’s home run rate with 300 PAs of the league average value. Power hitter dominated the top of the leader board. I wanted use only this season stats for ease of computation, so I change the regression value from 300 PA to 150 PA.

The change takes a little less emphasis off HR/FB%, but still shows players that are under or over performing.

Last Year’s Results

Taking the top 10 luckiest hitters from 2010, I compared to how their wOBA did then in 2011:

Name Luck 2010 wOBA 2011 wOBA Difference
Stanton Mike 12.3 0.355 0.379 0.024
Thome Jim 12.3 0.437 0.344 -0.093
Morneau Justin 11.5 0.447 0.274 -0.173
Thames Marcus 10.5 0.365 0.254 -0.111
Alvarez Pedro 10.4 0.343 0.247 -0.096
Gonzalez Carlos 10.0 0.416 0.387 -0.029
Betemit Wilson 9.8 0.385 0.329 -0.056
Hamilton Josh 9.4 0.447 0.367 -0.080
Cust Jack 8.9 0.371 0.307 -0.064
Votto Joey 8.8 0.439 0.415 -0.024
Avg. Diff. = -0.070

Of the 10 players, only one (Mike Stanton) improved in 2011 and all the rest saw their wOBA drop. As a group, their wOBA dropped by an average of 70 points.

Final Thoughts

I like where Luck stands right now as a rate stat to find players who were helped by an over inflated BABIP and HR/FB%. Here is the complete list of players from 2010 and 2011 in a Google Doc to examine. Let me know if you have any comments and I will look at pitcher’s Luck on Monday.




Print This Post

Jeff writes for FanGraphs, The Hardball Times and Royals Review, as well as his own website, Baseball Heat Maps with his brother Darrell. In tandem with Bill Petti, he won the 2013 SABR Analytics Research Award for Contemporary Analysis. Follow him on Twitter @jeffwzimmerman.


16 Responses to “Hitter’s Luck as a Rate Stat”

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

    It’s a lot harder to stay “lucky” for a long time, because regression will inevitably cut into your “luck”. So shouldn’t players who stay “lucky” for a long time get more credit in a stat that attempts to measure “luck”?

    Vote -1 Vote +1

  2. Scott says:

    Reading the “unluckiest” for 2011 is mostly a who’s who of powerless slap hitters. I’d like to see unluckiest 2010 vs. what they did in 2011. My guess would be the gains in wOBA for the unlucky aren’t as great as the regressions in wOBA for the “lucky”

    It seems that one might look for the “unlucky” w/ the highest ISO’s as a list of players that still have an underlying skill that screams “bounce back candidate”

    Overall this is fascinating work though.

    Vote -1 Vote +1

    • Scott says:

      BTW, I just tried to do this but the player names are formatted in reverse of the FG exportable leaderboards…. :(

      It’d take more time than I have to map the Luck spreadsheet back to player names w/ ISO from the leaderboards with the google doc formatted as it is currently.

      Anybody out there got a quick and dirty way to make “Bautista Jose” into “Jose Bautista” in excel?

      Vote -1 Vote +1

      • Will says:

        – Clear several columns past your list of names.
        – Data => Text to Columns
        – “Delimited” using “space”

        Now you have “Bautista” in column A and “Jose” in column B.

        – in cell C1 type =A1&” “&B1 which will giveyou Jose(space)Bautista.

        You’ll probably actually need 4 columns (think Jorge de la Rosa)

        Vote -1 Vote +1

    • Scott says:

      Cross referencing some of the worst luck scores w/ ISO’s above .130 for 2011 (i.e. guys that have still demonstrated ability to make solid contact despite their poor luck):

      Tyler Colvin: -6.2, .157
      Mike Fontenot: -6, .136
      Scott Rolen: -4.6, .155
      Chris Coghlan: -3.4, .138
      Hank Conger: -3.3, .136
      Bryan Petersen: -3.2, .135
      Dioner Navarro: -2.7, .131
      Ian Kinsler: -2.6, .225
      Erick Aybar: -2.3, .130
      David DeJesus: -2.1, .133
      Jimmy Rollins: -1.9, .134
      Ben Francisco: -1.6, .133
      Domonic Brown: -1.5, .148
      Will Venable: -1.4, .141
      Hideki Matsui: -1.2, .132
      Lonnie Chisenhall: -1.2, .151
      Evan Longoria: -1.1, .242
      Carl Crawford: -1.1, .150

      Especially Kinsler and Longo are outliers on this list that have way more power than the average “unlucky” players from the 2011 list.

      Vote -1 Vote +1

  3. Scott says:

    Also, Jeff: Any theories on Carlos Gonzalez being in the top 10 for luck 2010 and 2011? Does he have a sustainable skill that looks lucky? Does the crazy amount of outfield space in Coors inflate his BABIP consistently? Can we park adjust these xBABIP calculations?

    Vote -1 Vote +1

    • Jeff Zimmerman says:

      Probably. I have not gotten access to the park factors from the power that be yet. Besides BABIP corrections, I need to run some HR corrections.

      Vote -1 Vote +1

  4. Scott says:

    Anybody else notice that players we all think of as “powerful” hitters (Ortiz, Cruz, M. Cabrera, Stanton etc.) seem to carry their positive “luck” score season to season? There seems to be a power/ISO/xBH bias in the method.

    Vote -1 Vote +1

    • Brad Johnson says:

      Yea, the above leaderboards made me wonder if the specifications are “correct.” It seems like power equals positive luck and punchiness equals negative luck. We’re talking about 5 names so I wasn’t going to say anything.

      Nevertheless, Jeff’s method doesn’t quite pass the smell test for me. We know that BABIP and HR/FB are the main luck stats, but they aren’t the only ones. Just as a guess, how much of “luck” do you think is encompassed by those two stats? Say 70 or 80%? Furthermore, we have the problem of sorting out the difference between “luck” and temporary/permanent change in skill.

      Take Justin Morneau for example. Sure his concussion problems are unlucky, but that’s not what we’re talking about here. Because his concussions have changed his skill level, I would argue that he has probably NOT been unlucky, that his injury simply makes him appear unlucky.

      Vote -1 Vote +1

      • Jeff Zimmerman says:

        If you look through the 2010 list, BABIP – xBABIP is more of the issue than HR/FB differences.

        I due agree on the light hitters (players that can’t hit a ball out of the park). I am trying to think of a way to get them to zero.

        The 2010 list passes the smell test fine IMO.

        Vote -1 Vote +1

      • Brad Johnson says:

        I guess I’m just picturing the standard regression formula Y=B0+B1X+B2X+… and finding things awry. But we’re not technically doing regression analysis here, so perhaps I just need to put that out of my mind.

        Vote -1 Vote +1

  5. Tim says:

    Would a hitter with more power (e.g., Stanton) tend to have a healtheir HR/FB ratio since he will just hit the ball further? If that is the case then should it be considered luck or just hitting ability?

    Vote -1 Vote +1

  6. jkb says:

    Do text to columns in Excel so you have a first name and last name column then recombine them using =firstname&” “&lastname

    Vote -1 Vote +1

    • Scott says:

      thanks for the tip,that’s much simpler.

      I ended up doing it a much less elegant way:

      1. making a second column of names
      2. Delete the first names in the first column w/ a find and replace for text followed by a space
      3. deleted the last names in the second column w/ a find and replace for text preceded by a space
      4.Concatenated those columns together w/ Last, First

      I really overthought that one….LOL….but at least my Vlookup for the ISO values worked. See my reply to my comment above for notable high ISO low luck players from 2011.

      Vote -1 Vote +1

  7. Jeffrey Gross says:

    You can’t just regress HR/FB rates for hitters. Each has a unique number per year. Its too variable to justifiably say “this is luck” or not.

    For hitters, just use the following formula:
    1. xBABIP-BABIP converted into luck hits (+/-), or xH
    2. Add xH to actual hits (aH), divide by AB for xAVG
    3. Use (xH+aH) in the OBP formula
    4. Subtract SLG-AVG (ISO), and add to xAVG for xSLG.

    That will give you the best “luck reduced” measured triple slash line, in my opinion.

    Converting luck hits into home runs, though, is foolish. That presumed robbed home runs, which are very rare. “bad luck” and “good luck” on hits tends to be on non-home run plays, meaning that the luck would not have likely translated into home runs. I personally prefer to keep ISO constant, to presume steady power, but alternatively, you could just pessimistically presume +/- xH all singles…

    I can probably explain in better detail if you want to email me. I presume you have my email still, but if not ask Eno or mail to the name listed here if you can see it

    Vote -1 Vote +1