About that intentional walk (no, not that one)

The Overworked Relievers
The RAB Radio Show: May 24, 2011
Were the chances of this happening worth the IBB? (AP Photo/Paul J. Bereswill)

I try not to evaluate things in hindsight all that often here, just because it’s easy to sound smart when you already know what happened. That said, let’s have a little fun and play a game of “what if” with last night’s intentional walk. No, not the one to Juan Rivera (which made little to no sense), but the one to Jose Bautista earlier in the inning.

The game was tied at one when Corey Patterson led off the sixth inning with a double over the head of Chris Dickerson. Baseball Prospectus’ run expectancy matrix tells us that a team with a runner on second an no outs is expected to score 1.035 runs in the inning. Let’s say the Yankees pitched to Bautista and he did hit a homerun in that spot, making it a 3-1 game. He just added two runs to the ledger, but they were already expected to score 1.035 runs in the inning anyway. The net gain from the homer would have been 0.965 runs (2 – 1.035). The inning then “restarts” with the bases out and no one out, which has a run expectancy 0.4646. The total impact of the homer would have been 0.965 runs plus the 0.4646 runs, or 1.4296.

The Yankees didn’t pitch to Bautista though, they put him on first base intentionally. The run expectancy of first and second with no outs is 1.3986, so the impact of the free pass was just over a third of a run (0.3636 to be exact). That’s the situation they chose over pitching to Bautista, which in the worst case scenario (homer) would have resulted in an additional 1.4296 runs above expected. Of course Bautista wasn’t guaranteed to go deep (even if it felt like he was), the guy had “just” 19 long balls in 178 plate appearances coming into that at-bat, so the odds of him hitting one out were 10.6% based on how his season had played out to that point.

If we crudely multiply that 1.4296 worst case run expectancy by the chances of it happening, or 10.6%, we get a 0.1515 runs. That’s less than the 0.3636 runs the Yankees gave the Blue Jays by putting Bautista on, so yeah, the math says they should have pitched to him. Of course it didn’t play out according to the run expectancy, Toronto ended up pushing five runs across in the inning, making those totals of 0.1515 runs and 0.3636 runs seem silly. Remember run expectancy doesn’t tell us what will happen, just what is expected to happen based on historical data. In hindsight, pitching to Bautista and hoping he didn’t hit a homer was a better option than walking him, but that’s much easier to say now than it was last night.

The Overworked Relievers
The RAB Radio Show: May 24, 2011
  • Math

    You also have to include the run expectancy had he hit a single, double, triple, etc. Think of it as calculus and ‘everything under the curve’. Had Bautista grounded out to second, advancing the runner to third, that also has run expectancy repurcussions. It is too simplistic to look only at the possibility of a home run.

    That said, it probably would not change the answer.

    • Cris Pengiuci

      That said, it probably would not change the answer

      I’d still love to see all those calculateions. :-) Care to run through them for use?

      • Cris Pengiuci

        calculateions calculations

      • Brian in North Hampshire (formerly New Hampshire)

        reminds me of my calculus days…finding the value of everything under the curve. I can picture the graph…just not the formulas. Its been 10 years :(

  • Frank

    Mike, well thought analysis. However, IMO, the key AB was Patterson leading off the inning. Colon had him 0-2 before the count went full. Patterson is a guy Colon had to get out because he’s just not that good.

  • virginia yankee

    It is unlikely that major league managers or coaching staff that have any concept of risk / reward – chain of events calculus — they have grown up playing by the “book” whatever the “book” was at the time; or “gut feel”, or “instinct”. That is why Girari has Granderson sacrifice but doesn’t PH for Cervelli.

    – the worst thing would be a long rally which is what happened. Bautista had a OBP chance of being a baserunner and BAIP of potentially moving the runner – minus BAIP that could not move the runner. There was 10% chance he might hit a 2 run homer – I suppose you could park adjust, count adjust, pitcher adjust – etc — or just go with your “gut instinct”

    Bautista may deserve to hit .400 but he does not deserve to hit HR in every 10 at bats — other than pitcher error there is no reason to place pitches in his pull zones — not every P can locate like a Gregg Maddux, but they can try. Colon’s pitch to B in the 1st was awful — but he has shown very good control until last night. He might have done better than the chain of events delivered.

    Not only did the Yankees not play “swinging bunt ball” last night, the did not play Long Ball either.

    This is 1966 heading for 1968. Or 1981 heading for 1996. PEEMPT – get rid of Cashman and Girardi before it is too – late — Get a GM and Manager with ability to make deals and manage by what they SEE – not what they hope to see; or what they have seen in the past.

    • Ted Nelson

      “It is unlikely that major league managers or coaching staff that have any concept of risk / reward – chain of events calculus”

      That’s pure speculation. Girardi has a famous binder he’s always looking at… what do you think he keeps in there? Comics?

      “he does not deserve to hit HR in every 10 at bats — other than pitcher error there is no reason to place pitches in his pull zones — not every P can locate like a Gregg Maddux, but they can try.”

      You are contradicting yourself… Guys should pitch to Bautista and try to locate their pitches correctly, when guys do pitch to Bautista he hits the crap out of the ball… see the conflict?

      This is the big leagues… no one “deserves” anything, but people still do things. 10% is a number Mike pulled out of his ass. It’s not just about HRs. Bautista has 27 total XBHs and 22 singles. It was a tie game. Even a single would have likely scored the run, while an XBH would have scored the run and left another man in scoring position with 0 outs.

      “This is 1966 heading for 1968. Or 1981 heading for 1996. PEEMPT – get rid of Cashman and Girardi before it is too – late”

      What on earth are you talking about? Cashman does get guys he has seen do well in the past–CC, Tex, A-Rod, etc. However, in trades these guys tend to cost a premium. He’s also gotten guys who he’s seen do well in the past, but have struggled lately like Granderson and Swisher at a potentially discounted price.

      • Mister Delaware

        “That’s pure speculation. Girardi has a famous binder he’s always looking at… what do you think he keeps in there? Comics?”

        Agreed, and I think that’s the annoying-to-infuriating thing about Girardi. He has shown that he is a smart guy who does grasp the percentages and its safe to assume the binder contains all the relevant data we’re citing here, yet he’s getting worse and worse when it comes to decision making. Just looking at the two most obvious types of managerial meddling …

        2008 IBB%: 0.60% of PAA
        2009 IBB%: 0.45%
        2010 IBB%: 0.61%
        2011 IBB%: 0.89%

        2008 Sac%: 0.50% of PA
        2009 Sac%: 0.48%
        2010 Sac%: 0.52%
        2011 Sac%: 0.62%

        At the current rates (28.3% of the season done), he’ll break his past sac bunt record by 8 (39 to 31) and IBB record by 19 (56 to 37). Obviously not huge raw numbers, but as percentage increases, they’re material.

        • Ted Nelson

          You absolutely could be right, but at the same time his binder might not have the same data we’re looking at… it might have better data. (Ideally, it should be an iPad or something that can help in a more dynamic way.) He might have access to similar data, but specifically looking at the Blue Jays this season with Bautista coming up (too small a sample, maybe, but maybe this season and last, or the second half of last).

          In making your hypothesis entirely dependent on run expectancies, you are making an implicit assumption your table is as good as or better than what Girardi has access to. Maybe it is. But the Yankees are a multi-billion $ business and might have been data than us.

          You have also apparently decided that every single bunt and every single IBB is a bad move… Not sure if that’s the case or not.

          • Mister Delaware

            I don’t think they’re allowed electronics, especially something two-way like an iPad, in the dugout.

            I imagine the Yankees and Girardi absolutely have better data than us, but that doesn’t assure Girardi is properly using it. And, assuming the binder can’t account for every hitter/runner(s)/outs/innings/score scenario, biases are always going to be in play which is where the Girardi debate lies.

            (To answer the last question: I absolutely don’t think every IBB or sacrifice bunt is a bad move. Cervelli forced to hit with runners on 1st and 2nd in the 9th of a 1 run game? Absolutely bunt him. If Girardi sacs 10 times, I’m sure I agree with 3, think “I wish he hadn’t done this” but don’t flip out at 3 or 4 and then lose my shit on the other 3 or 4. Not because I’m 100% right, but because my personal understanding and interpretation of the percentages dictate that reaction.)

            (Pretty sure that response was 100% clear and not subject to any reach interpretation. Of which I’m super proud.)

            • Ted Nelson

              Probably true. Maybe eventually MLB will hand out sanctioned iPad data machines to every manager. (Seriously.)

              No it doesn’t.

              In simply listing his total IBB and sacs by season you’re not calculating which type of IBB and sacs he’s increasing… hence why I made that leap. Have to hope someone in the FO is monitoring it, and if they’re not happy they communicate it to Girardi… who knows if that happens, but you have to hope.

              Is the Bautista IBB one where we need to flip our shit? Marginal change in run expectancy (I calculate 1.235 expected if you don’t IBB him… have no bothered to extend that to what you then expect for Escobar and beyond contingent on what has happened with Bautista) in a tie game to take the bat out of the hands of a 1.330 OPS guy and put it in the hands of a .771 OPS guy.

              • Mister Delaware

                Nope, the Bautista walk was in the first category since he was being followed by Escobar and Rivera and it was in the Stadium and Colon throws a decent amount of GBs and considering the Jays have some decent relievers to run out behind Villenueva. That the IBB ended in the worst case as Bautista walked to the plate (both he and Patterson scoring) sucks, but I took no issue with the process.

      • http://youcantpredictbaseball.wordpress.com bexarama

        Ted, you’re cool, but don’t try arguing with virginia yankee. It’s one thousand percent pointless.

      • http://notanevilempire.com Steve O.

        That’s pure speculation. Girardi has a famous binder he’s always looking at… what do you think he keeps in there? Comics?

        Actually, yes. Joe Girardi has been know to have many Batman, Superman, Wonderwoman, Catwoman, Spiderman and Ironman(among others)comics in his binder. He reads them in between innings, and sometimes during the opponent’s at bats.

        Shocking, I know.

  • Kevin M.

    This is the same site that recently complained about not letting the other team’s best hitter beat you, right? Time to get your stories straight.

    • hogsmog

      Not letting their best hitter beat you is not the same as always IBBing their best hitter. IBB is sometimes the right choice, but never most of the time.

  • Damian

    My understanding of run expenctancies is that they are derived from historical data and are meant to show outcome probabilities under different circumstances WITH ALL THINGS BEING EQUAL. Jose Bautista is the Jays’ only really good hitter, so all things are not equal, and the intentional walk decision has to incorporate that information. That said, you’ve provided an interesting take on the matter.

  • LeftyLarry

    Ridiculous moves by Girardi.

    It’s the 6th inning and that’s not BAbe Ruth facing Mickey Mouse.

    You do not walk the guy to make it first and second with nobody out and set up a big inning.

    The great likelyhood is, guy flies out and if he hits a HR, do what, it’s the 6th ining and you’re 2 runs down.

    If COlon is good enough ot face 3 men with first and second and no outs, he’s good enough to pitch ot a big power hitter and get him out.

    2 walks to create a big inning, was ridiculous and frankly, the more I watch Girardi, the mroe ocncerned I get.

  • Matt

    I took Mike’s crude methodology, except used Bautista’s stats with runners on base because the approach of him and the opposing pitcher likely both change in those situations. Obviously these are very rough calculations because they are based off of small sample sizes and make a ton of latent assumptions, but let’s play around with it for fun.

    His breakdown with runners on this year is like this:

    83 PA
    26 Other Outs
    25 BB
    11 Strikeouts
    11 Singles
    3 2B
    6 HR
    1 HBP

    I don’t have data on his other outs, but let’s take from Fangraphs that he hits 40.4% GB, 38.3% FB in medium leverage situations, and from Texas Leaguers he has hit only 4 ground balls to the right of 2nd base this season for outs, so let’s assume that it’s 4 of his 37 ground balls that would move a runner up, or roughly an 11% chance. Let’s also assume that half of his fly balls could advance a runner, since he tends to drive the ball well much of the time.

    So if we take all of this together, what do we get?

    13%- Single- Let’s say 50% chance it scores the runner from 2nd to give us something
    30%- Walk
    13%- Strikeout- Let’s say the runner doesn’t advance on this outcome
    4%- Double- Let’s assume the runner scores
    7%- Home Run- Both runners score
    1%- Ground ball to the right side or behind 2nd base
    12%- Ground ball to the left side
    6%- Fly ball that doesn’t advance runner
    6%- Fly ball that advances runner
    8%- Other out that does not advance runner

    Taking the run expectancy matrix from 1999-2002:

    6.5%- 1.904
    6.5%- 1.953 (1 that scored + runner at 1st 0 outs)
    30%- 1.573
    13%- 0.725
    4%- 2.189
    7%- 2.555
    1%- 0.983
    12%- 0.725
    6%- 0.725
    6%- 0.983
    8%- 0.725

    Multiplying these all out gives an expected runs for the inning of 1.341 as soon as Bautista steps to the plate. 1st and 2nd, no outs is 1.573 according to the same matrix. So it seems that it did increase the run expectancy of the inning as you suggested.

    • Matt

      Interesting to note is that the same matrix says that on average, runner at 2nd base, 0 outs yields an expected 1.189 runs.

      So under these assumptions, Bautista is expected to generating 0.152 runs over the average performance in that situation.

      • Matt

        *to generate

    • Ted Nelson

      Where did you come up with the 50% number for a fast runner like Patterson scoring from 2nd and advancing on a fly ball?

      • Matt

        I just made them up so I would have something to work with. If someone could get a figure for that, it would make the numbers more accurate obviously.

        I don’t think it would change the outcome substantially though because those situations couldn’t comprise much more than 1/4 of the results together.

        • Mister Delaware

          If you ever trying to reverse engineer linear weights, figuring out the advance percentage is easily the most frustrating part.

      • Matt

        I changed both of those outcomes to advancing 100% of the time and it increases the run expectancy pre-IBB to 1.359, so a slight increase but not enough to change the result.

        Obviously there are things which can’t be accounted for in what we did here. For instance, are Juan Rivera and the like going to provide average performance? Is Jose Bautista better positioned to succeed against Colon because of his fastball approach and his lack of control in the strike zone?

        If the hitters that follow are below-average, as some of them were, then perhaps the two outcomes end up being closer than these numbers suggest. The difference though seems to be in that walking Bautista sets up the possibility of a big inning, while giving up a HR kind of just kills the rally right there.

    • yankees28then29

      So it is basically never correct to intentionally walk someone with man on second and no outs.

  • Ted Nelson

    I’m no expert, but I don’t believe you are using the run expectancies in any meaningful way.

    Saying that on average a team gives up X runs says nothing about what they give up with Bautista coming to the plate vs. the great Yunel Escobar… whose OPS is only, oh, 560 points lower than Bautista’s this season. Nothing about Escobar having 6 GDP this season, the next batter Rivera having 5, and Bautista having 1. Or with a 39 year old starter in his 6th inning.

    When it suits your point, you seem to favor an in-depth analysis… you always harp on over-using (and even just using 1/2 the time) reliever… some sort of vendetta against bullpens almost… so when looking at pitch counts rather than IP comes up, you just jump right in. No mention of warm ups being constant for a 5 or 10 pitch inning when you mention warm-up pitches, though… because that contradicts your theory on bullpen usage.
    In this case, actually looking at the specific situation would fly in the face of your theory… so why not just look at the league average?

    Looking only at HRs for Bautista–as if all other hits wouldn’t also be theoretically worse than a walk… and even a long AB that resulted in a walk wouldn’t be worse than an IBB in terms of Colon’s pitch count–is a stretch.

    You don’t like to say things in hindsight? Come on…

    • V

      You’re way too antagonistic, d00d.

  • king of fruitless hypotheticals

    Add more math–what if we bring in Robertson to face Bautista? Sure Colon may have looked ok to that point…but if we could have put our best reliever (non-Mo division) on the hook at that point, what are the chances Robertson beats Bautista? Then what?

    • Mister Delaware

      Perfect name + post.

      • king of fruitless hypotheticals

        Except for ‘Hopeless Homer’ and ‘I Don’t Understand Sabremetrics’ I might be the most appropriately named poster here…

  • in the know

    the hilarious thing about baseball fans trying to use “math” is that none of them understand it, so they just bastardize whatever concept to justify what they’d think anyway. this is to say… they’re still just fans.

    • Mister Delaware

      That assumes a lot of fans don’t have a snap opinion, run the numbers, prove themselves wrong and never post at all. I do that weekly.

      • Ted Nelson

        And assumes that mathematicians/statisticians/etc. are never baseball fans…

  • Monteroisdinero

    And for some good news off topic: Scranton wins a day game behind a solid pitching effort by Warren and homer #17 by JoVa as he continues his chase of Bautista from dotf.

    /sorry but we needed some

  • Dave

    Not to mention that Giardi has a minute to make these calculations. We have an entire night, plus don’t have to make any other decisions or be thinking several plays ahead. I don’t really have a problem with his decision here. He had a much better idea of we do than a lot of the variables that statistics just aren’t going to cover — what kind of stuff Colon has, what the scouting report is saying about Bautista, etc. Some of the decisions inevitably drive us mad because we know they don’t work, but that doesn’t mean that under the circumstances they aren’t perfectly defensible or that, when they don’t work out, Giardi and Cashman should be instantly fired because the Yankees are only in 1st.

  • The209

    Two things I’m wondering…

    * If the run expectancy with man on 2nd, 0 outs = 1.035 for the ‘average’ hitter in their database — what is the number when the batter is hitting homeruns at 4x (or whatever multiple) the rate? No one has those numbers — so all of these types of analysis are rather arbitrary aren’t they?

    * Also, you say, “If we crudely multiply that 1.4296 worst case run expectancy by the chances of it happening, or 10.6%, we get a 0.1515 runs.” But it doesn’t make sense to do that …. the database already multiplied the chance of a homerun happening (albeit at a rate lower than 10.6%, assuming the average of all data) to come up with the run expectancy for that given situation, didn’t they?

  • Tim

    Why would you even consider run expentancy in a situation like this? The entire concept of run expectancy was an interesting one to me, which led me to buy Tom Tango’s “The Book”. I read through it and on several occasions nearly threw it out the window. The concept of run expectancy is unambiguously flawed and essentially worthless for one reason – IT DOESN’T TAKE INTO ACCOUNT ANY OF THE PARTICIPANTS IN THE SITUATION THAT IS OCCURRING. It doesn’t consider who’s hitting, who’s pitching, who’s available on the bench, who’s on deck, what inning it is, what the score is, what the pitch count is, the style of pitcher, hitter tendencies, etc, etc, etc. According to Tango’s calculations, as reported by Mike above, walking Bautista put the run expectancy at 1.399 for the inning. Well, that number would be the same if the next three batters were Yunel Escobar, Juan Rivera, and Aaron Hill as if they were Babe Ruth c. 1927, Barry Bonds c. 2003, and Hank Aaron c. 1970. How in the name of Mo can you consider any calculation that works like this valid?

    In order for run expectancy to REALLY work, there has to be some factor included for who’s pitching, who’s scheduled to hit, etc. You cannot with a straight face tell me that Toronto had a better chance of scoring with the three stiffs coming up with first and second over Bautista coming up with a runner on first. If you do, you just don’t know anything about baseball. As it turns out, they scored 5 times – but that is because guys who aren’t very good hitters ended up getting hits. It happens about 23 times out of every 100 AB for those guys.

    IMO, using run expectancy as a tool to make decisions solely is flat out idiotic, and any manager doing that for the team I GM would be promptly fired. It’s barely even directional as a statistic, when you think about it. I have no aversion to sabermetrics, and I do subscribe the most of the analyses that have come from the SABR- revolution. But this one is pretty much garbage.

    If you play out that 6th inning 200 times, walking Bautista 100 and letting him hit the other 100, when all is said and done, does anyone really believe the Yankees would be better off in the latter 100?

    • Tim

      Oops, meant Bautista coming up with a runner on second.

    • Mister Delaware

      “If you play out that 6th inning 200 times, walking Bautista 100 and letting him hit the other 100, when all is said and done, does anyone really believe the Yankees would be better off in the latter 100?”

      This makes me wish I had access to the simulators BP uses. Because I know my answer* but I don’t know if its close to right.

      * My answer is an annoying “what does ‘letting him hit’ entail?” If the first option is a 100% BB rate and the second is a 0%, like he has to hit/K, I’d probably take my chances with the latter (64.7% chance of an out, SLG less than 1.000 means he’s average less than the one base you’d be giving away). If the second option still uses his 22.7% walk rate, I take the former and move on to Escobar and Rivera.

      • Mister Delaware

        Nevermind, I’m taking the IBB no matter what “letting him hit” means. .867 bases per PA, major league average is .439. Once you back out his numbers from the MLB average, he’s almost twice as dangerous as the median player. That’s stupid.

    • The209

      I was wondering the same thing (comment above yours).

      It’s been years since I took statistics, but this sort of analysis seems like it’s missing a lot of basics.