Minnesota Twins News & Rumors Forum
  • Twins 8, Brewers 6 and Stealing Bases

    The Minnesota Twins won last night, swiping (and sweeping) the season series from the Milwaukee Brewers. The victory raised their record to 23-28 which translates to a 73 win pace over the course of a 162-game season.

    But I want to talk about a different type of swiping that happened, or more accurately, did not happen. In the bottom of the sixth inning, Aaron Hicks was caught trying to steal second base. What struck me as strange at the time was that there were runners on the corners and one out. It seemed like a stupid decision to attempt that at the time, as it can short-circuit a good scoring opportunity. However, one always looks stupid when one is caught. So I wondered, was it as risky as it seemed?

    It turns out – I studied this exact situation last year on 5/11, only it was Denard Span that was thrown out. Here’s how that went:

    Generally, one studies something like this using Palmer & Thorn's Run Expectancy Matrix. It's a neat grid that shows, given a certain number of outs and people on base, the average number of runs that should score that inning, based on 75 years of major league games. It was published in The Hidden Game of Baseball by Pete Palmer and John Thorn. You can find it here.

    Here's the numbers we care about:
    1. 1.088 - That's how many runs a team on average would score with runners on 1st and 3rd and one out.
    2. 1.371 - If Span would've stolen the base, that's how many runs the average team would've scored.
    3. 0.382 - If he was caught, that's how many runs the average team would score.

    So Span risked a gain of .283 runs if he stole that base, but a loss of .706 if he was caught. Converting those to percentages, if he steals that base 71% of the time, the team breaks even. That's not especially different than the 2/3 view that is the case for most base stealers. This wasn't especially risky.
    Tonight, though, I want to take it a step further. How does it compare to other situations in which one might try to steal a base?

    There are 12 situations where a runner might try to steal a base without coordination with the other runner. Here they are:

    Runner on 1st, 0 outs
    Runner on 1st, one out
    Runner on 1st, 2 outs
    Runner on 2nd, 0 outs
    Runner on 2nd, 1 out
    Runner on 2nd, 2 outs
    Runners on corners, 0 out
    Runners on the corners, 1 out
    Runners on corners, 2 outs
    Runners on 1st and 2nd, trying to steal 3rd, 0 outs
    Runners on 1st and 2nd, trying to steal 3rd, 1 out
    Runners on 1st and 2nd, trying to steal 3rd, 2 outs

    And here they are again, ranked by just how risky they are. The percentage indicates how often one needs to succeed for it to be a good decision.

    92.51% - Runners on 1st and 2nd, trying to steal 3rd, 2 outs
    91.10% - Runner on 2nd, 2 outs
    79.67% - Runner on 2nd, 0 outs
    77.69% - Runners on 1st and 2nd, trying to steal 3rd, 0 outs
    77.25% - Runners on 1st and 2nd, trying to steal 3rd, 1 out
    75.31% - Runner on 2nd, 1 out
    74.74% - Runners on corners, 2 outs
    71.39% - Runners on the corners, 1 out
    70.73% - Runners on corners, 0 out
    65.20% - Runner on 1st, 0 outs
    63.41% - Runner on 1st, one out
    60.06% - Runner on 1st, 2 outs

    So it wasn’t patently stupid, like trying to steal 3rd when you’re already in scoring position. But it was the 2nd riskiest situation in which to try to steal 2nd base. The only thing that would have been worse was if it could have ended the inning.
    This article was originally published in blog: Twins 8, Brewers 6 and Stealing Bases started by John Bonnes
    Comments 20 Comments
    1. Seth Stohs's Avatar
      Seth Stohs -
      I didn't mind it at all. Hicks has good speed. I would rather have runners on 2nd and 3rd with one out than leave the door open for a double play to end the inning. I agree with and understand the numbers you're sharing. Getting thrown out in conjunction with the batter getting out is essentially replacing the double play ground ball. But, I do like 2 runners in scoring position, so I would never complain about a good base runner trying to steal there... especially when you already have a lead.
    1. jorgenswest's Avatar
      jorgenswest -
      John,

      Are the success percentages tied to the run scoring environment? Should they be? Are the percentages you presented tied to the current run scoring environment? We certainly are in a lower run scoring environment now than a handful of years ago.

      I would think that the steal becomes less risky as the chance of driving in the run becomes lower. If true, then the success percentages not only change by era but also game to game, pitcher to pitcher, batter to batter and park to park.

      If Florimon, Carroll and Dozier are the next batters due up it seems like a less risky proposition than if it were Mauer, Willingham and Morneau.
    1. IdahoPilgrim's Avatar
      IdahoPilgrim -
      I think what needs to be done is have this analysis taken farther, taking into account game situation, perhaps using WPA as was discussed in an earlier article.

      If I remember correctly, the Brewers had just put 3 runs on the board, cut the lead to 6-3, and made it an interesting game. In that case, perhaps the risk/reward equation shifts, making it more acceptable to run than when the game is not close.

      Just a thought.
    1. tjsyam921's Avatar
      tjsyam921 -
      Looked like Hicks had a bad jump as the throw was off-line and he was still caught rather easily.
    1. Don't Feed the Greed Guy's Avatar
      Don't Feed the Greed Guy -
      There are so many other factors to consider: the pitcher's ability to hold runners--PD (time to home plate), who's catching, the count... These are not factored into the run expectancy matrix. Like you said, we are second-guessing this because Hicks got caught. I like pushing the defense, making the Brewers and future opponents account for an aggressive running game. That's a strategic advantage too, that doesn't show up in a computer matrix.
    1. boylan's Avatar
      boylan -
      Very cool set of numbers.

      In this case I'm not sure they are relevant however. Don't know whether he was sent or went but it appeared to me to be a case of stealing a base simply as part of transitioning to the majors. Hicks has been on base so little he really hasn't had a chance to get "use to" or get any exposure to stealing bases in the majors. With that lead I would have been shocked he wasn't sent/went in practically any situation.
    1. Willihammer's Avatar
      Willihammer -
      If you want to split hairs, all of those break even values should be weighted for a team's home run hitting proficiency, where obviously the Twins are lacking. Which makes the steal a more viable option for them in all situations.

      By the same logic, to get even more precise, the break even calculation should be weighted against the home run hitting proficiency of the batter at the plate - Florimon in this case. Obviously that lowers Hicks' break even point even further.
    1. mike wants wins's Avatar
      mike wants wins -
      John is right. It is bad strategy,
    1. LaBombo's Avatar
      LaBombo -
      Hicks is fast, no question, but he didn't look exceptionally quick in the first three or four strides. Maybe it was more the late jump.

      At any rate, I'm ok with him getting more chances to run, even if some of those chances exceed what's best for the optimal run-scoring strategy. If it costs them runs, tough hop. It's the cost of conducting on-the-job training at the major league level.
    1. DeepFriedTwinkie's Avatar
      DeepFriedTwinkie -
      We can account for the situation (bottom 6th, one out, up four) and calculate the breakeven rate by comparing win expectancies (The Win Probability Inquirer):
      WE before steal: 97.12%
      WE if successful steal: 97.38%
      WE if caught stealing: 95.86%
      Implied breakeven rate: 83%

      This shows two things:
      1. You should believe that Hicks has an 83% chance of stealing 2nd in this situation before sending him.
      2. Even in the unfavorable outcome (caught stealing), you still have a 95% chance of winning.
    1. LookHere's Avatar
      LookHere -
      Nice write-up. There are so many amazing statistics out there that undoubtedly would help a team win more games over the course of a season if the manager put them to use. I can't help but wonder if Gardy even cares about stuff like this. I hope so, but I doubt it.
    1. Seth Stohs's Avatar
      Seth Stohs -
      Quote Originally Posted by DeepFriedTwinkie View Post
      We can account for the situation (bottom 6th, one out, up four) and calculate the breakeven rate by comparing win expectancies (The Win Probability Inquirer):
      WE before steal: 97.12%
      WE if successful steal: 97.38%
      WE if caught stealing: 95.86%
      Implied breakeven rate: 83%

      This shows two things:
      1. You should believe that Hicks has an 83% chance of stealing 2nd in this situation before sending him.
      2. Even in the unfavorable outcome (caught stealing), you still have a 95% chance of winning.
      I like looking at it this way.. I wonder if how that 83% changes when the Twins are down by 1 run or 2 runs, or if the game is tied?
    1. DeepFriedTwinkie's Avatar
      DeepFriedTwinkie -
      Quote Originally Posted by Seth Stohs View Post
      I like looking at it this way.. I wonder if how that 83% changes when the Twins are down by 1 run or 2 runs, or if the game is tied?
      I've actually already done the calculations where there is just a runner on 1st for a few late-game scenarios. Could easily update it to look at the 1st and 3rd scenario, though.

      Attachment 4257
    1. jorgenswest's Avatar
      jorgenswest -
      Quote Originally Posted by DeepFriedTwinkie View Post
      We can account for the situation (bottom 6th, one out, up four) and calculate the breakeven rate by comparing win expectancies (The Win Probability Inquirer):
      WE before steal: 97.12%
      WE if successful steal: 97.38%
      WE if caught stealing: 95.86%
      Implied breakeven rate: 83%

      This shows two things:
      1. You should believe that Hicks has an 83% chance of stealing 2nd in this situation before sending him.
      2. Even in the unfavorable outcome (caught stealing), you still have a 95% chance of winning.
      How did you set the run scoring environment given the next three hitters were Florimon, Carroll and Dozier?

      I would think it must be different than if Mauer was batting followed by Willingham and Morneau. Looking the their runs created per game, it probably is at least a 3 run difference between the two trios. I am not sure how the run scoring environment correlates to runs created.

      Wouldn't using a lower run scoring environment also decrease the break even point?
    1. DeepFriedTwinkie's Avatar
      DeepFriedTwinkie -
      Quote Originally Posted by jorgenswest View Post
      How did you set the run scoring environment given the next three hitters were Florimon, Carroll and Dozier?

      I would think it must be different than if Mauer was batting followed by Willingham and Morneau. Looking the their runs created per game, it probably is at least a 3 run difference between the two trios. I am not sure how the run scoring environment correlates to runs created.

      Wouldn't using a lower run scoring environment also decrease the break even point?
      I used a 4.0 run scoring environment (3.0 to 6.5 scale). Here are the rates in the most extreme run scoring environments:

      3.0 => 83% breakeven
      6.5 => 79% breakeven

      That's backwards of what I would have thought originally, but now that I think about it the run scoring env applies to both teams. A higher RSE makes it more likely that the Brewers can rally from four down. I can see where that makes it slightly more tempting to go after insurance runs.
    1. jorgenswest's Avatar
      jorgenswest -
      So you can't really set the run scoring environment for the context of which hitter is at the plate since it also reduces the Brewers overall run scoring environment.

      With Florimon batting and his high ground ball tendency I think the break even point changes. He is much less likely to hit a sacrifice fly and more likely to hit the ball on the ground keeping the runner at third or hitting into a double play.
    1. DeepFriedTwinkie's Avatar
      DeepFriedTwinkie -
      Quote Originally Posted by jorgenswest View Post
      With Florimon batting and his high ground ball tendency I think the break even point changes. He is much less likely to hit a sacrifice fly and more likely to hit the ball on the ground keeping the runner at third or hitting into a double play.
      I would agree with that. I also wonder about the likelihood of a bad throw to second trying to get hicks, which allows doumit to score from 3rd.
    1. Kwak's Avatar
      Kwak -
      Hicks did not "go on his own". It might have been a R&H with the batter missing the sign. I think that pitcher employs a slide-step. But to me, up 3 with the opponents having questionable pitching--the steal is foolish. Play patiently and wait for BBIP, the inning was built and now it's time to see if there will be a hit parade to bury the Brewers with an avalache of runs.
    1. Willihammer's Avatar
      Willihammer -
      Quote Originally Posted by jorgenswest View Post
      So you can't really set the run scoring environment for the context of which hitter is at the plate since it also reduces the Brewers overall run scoring environment.
      Not when you are looking at it from the perspective of Win Expectancy, no.

      From the perspective of expected run value you can at least narrow the calculus down to being lineup-specific. The formula for that is simply:

      Break Even Rate = 0.590 + 3.33 x (HR/PA)

      But, that too doesn't distinguish between a late and close scenario or a situation like last night with 1st and 3rd, 1 out, up 2 in the bottom of 6, or any other base/out/inning situation you can think of. Also, the constant and coefficient here are derived from the entirety of 1950-2012 (and yet manages to hold a r-squared of .685).

      So that's why I think for us who like to play armchair manager, you have to combine the two approaches somehow, possibly by weighting the WE values against the HR propensities of hitters, since the breakeven SB-HR relationship is so strong and HR rates can vary so wildly hitter to hitter.
    1. USAFChief's Avatar
      USAFChief -
      A run expectancy chart is interesting trivia, but tells you little about a specific game situation. It also doesn't give you much information on which to base managing decisions.

      Specific game situations are not equal. The "run expectancy" with 2 good baserunners on base, Cabrera/Fielder/Martinez due up, and Worley on the mound approaching 100 pitches;

      is not the same as

      2 poor runners on base, Hicks, Florimon and Dozier due up, and Justin Verlander on the mound.

      You will never convince me those two situations are exactly the same. Yet that's what you must believe if you believe a run expectancy chart gives you useful information with which to make strategy decisions.
©2014 TwinsCentric, LLC. All Rights Reserved.
Interested in advertising with Twins Daily? Click here.