Fridays Above Replacement

How Productive Are Productive Outs?

Last week, we had a rousing discussion about RE24.  We went over what it means, how to read the RE24 matrices, and alluded this this week’s topic.  This week’s topic, of course, is productive outs.  More specifically, are they actually productive and how can we use the RE24 framework to aid our decision making process?

The idea behind productive outs is very simple.  If you have to make an out, it would be preferable if the out would affect a positive change on the field, such as advancing a base runner.  I do not think anyone will ever argue that point.

However, there seems to be a disconnect between folks in baseball, fans, pundits, and analysts regarding the importance of productive outs.  Some players, seemingly, are attempting to record a productive out rather than just attempting to get on base and avoid an out at all costs.  My contention is that productive outs should be something that happen at random and we’re happy.  They should not be something a player strives for, except in very fringe cases.

Let us start with sacrifice flies.  A sacrifice fly, in the simplest example, is the trading of an out for a run.  This assumes the runner is not thrown out; then it becomes a trade of a base runner on third for an extra out.  That is a bad trade.  Here’s a chart with every sac fly situation (unless I missed one), along with the beginning and ending base-out states (BOS) and how much run expectancy (RE) was added as a result of the play:

chart1

The first thing that jumps out is likely the amount of red and green in the “RE Added” column.  Turns out 11 of the 16 sac fly scenarios end up decreasing run expectancy, even if a run scored.  Looking at the green scenarios lets us garner an interesting piece of data; 4 of the 5 positive outcomes come with 1 out.  As a general rule of thumb, we can say a sac fly is a more valuable play when it comes with 1 out rather than 0 outs.   The 5 worst outcomes come with 0 outs.  This is likely due to the “rally killer” effect that happens when you record an out and also trade your farthest-advanced base runner for a single run.

If a sac fly happens randomly, it’s a good thing.  It’s good to get a run on the board instead of striking out, for example.  However, if a player changes his approach in order to try and achieve the result of a sac fly, he could actually be hurting the team.  The better thing to do is stay with your normal approach and just try to get on base.  If a sac fly happens, view it as a silver lining. Never strive to record a sac fly.

Another type of productive out comes from attempting to hit behind a runner.  Again, if you hit behind a runner in your normal course of action, that’s great.  You advanced a runner.  However if you hamper your natural ability to get on base by attempting to hit behind the runner at all costs, you’ve likely hurt your team.  Here’s the pertinent table:

chart2

Perhaps a more interesting case than hitting behind the runner and sac flies are sac bunts, since they don’t happen by accident.  Because of this, we can do some math and figure out the actual run expectancy change by simply making the decision to bunt.   We are going to use a situation that came up during Brandon Finnegan’s no-hit bid on Monday night in Chicago.  More specifically, we will analyze a sac bunt situation with a man on 2nd and no outs.  At the beginning of this play, RE stood at 1.1 runs.  If the sac bunt happens as planned, the runner will reach 3rd base and Finnegan would be thrown out at 1st. This is defined as “success.” At that point, the RE of the inning would have stood at 0.95 runs; a decrease of 0.15 runs from the bunt.

To pull the thread a bit further, we know that not all bunts are successful, right?  Sometimes a pitcher strikes out.  Sometimes he gets to 2 strikes and the manager takes the bunt off.  Sometimes he’ll get a hit in that situation.  How do we make sense of all those different outcomes, you might ask? A table, of course!

chart3

For this table, I’m assuming an 80% chance of getting the bunt down.  Maybe it’s higher, maybe it’s lower.  We’ll go with 80%.  Also, once the bunt is down, I’m saying there’s an 80% chance it’ll be successful, and a 20% chance that the lead runner will be thrown out in some manner.  You can see the other outcomes and the percentages I applied to each other transition event.   If you fail to bunt and record a “bad out,” that means you created an out without advancing the runner.  Sometimes, you’ll fail to bunt and then hit behind the runner, advancing him anyways.  Sometimes you’ll get a hit!

Now, look closely at the Delta RE and Weighted Delta RE column.  Each transition event has an associated ending RE.  Calculate the different between beginning and ending, then multiply by the percent chance to occur, and we have our Weighted Delta RE metric per event.  If you sum all of those up, you end up with the total expected change in RE that is affected by simply making the decision.  As you can see, in this situation (given my assumptions) simply making the decision to bunt destroys 0.264 runs.

You might be thinking, “Dangit, Jeter! Finnegan is a pitcher!  He’s probably going to get out anyways if you let him hit!”  Well, perhaps.  But, he also may record a productive out which would be equivalent to a bunt.  He might single and knock the run in.  He might homer.  To wit:

chart4

The percentages on this chart were calculated a bit more “scientifically.”  In 2015, the average NL pitcher hit .132/.159/.169 with a 2.6% walk rate and a 37.2% strikeout rate.  Finnegan, by virtue of hitting safely in his first two starts this year, seems to be an above average hitting pitcher.  I’m going to make the not-so-bold claim that Finnegan could be a true-talent .152/.183/.194 hitter with a 3.0% walk rate and a 31.6% strikeout rate.

Also, the average NL pitcher had the following batted ball characteristics:  a 61.1% ground ball rate, a 5.8% bunt hit rate, a 26.3% pull rate, and a 40.2% up-the-middle rate.  All of that data is used to determine Finnegan’s chances of recording a productive out, such as hitting behind the runner.  If this happens, he advanced the runner anyways, so it was equivalent to having a successful bunt.  On the chart above, you’ll see most of the possibilities of what can happen broken out, as accurately as I can estimate on my lunch breaks.

So, with everything added together, letting Finnegan hit would result in…a decrease in RE of 0.152 runs.  Hrmmm.  That’s not very fun.   It’s not as bad as making him bunt (-0.264, as a reminder), though.  If you have a pitcher than can handle the bat, it’s generally going to be a better to let him hit than make him bunt, at least in a no-out, runner on 2nd situation.  However, pitchers hitting is a negative for the team.  There’s really no way to gussy it up

THE TAKEAWAY

Really, the only thing I hope we learned today is that productive outs are fine.  They represented a silver lining that is nice to have if your at-bat didn’t end up getting on base.  However, going out of your way to induce a “productive out” is generally a bad thing to do in most situations.

Doing a little research on my lunch breaks this week got me to this point.  What could a front office with an analytics team do?  Much more, I’d hope.  Every manager should understand how to break down the pros and cons of his decisions.  Maybe he won’t have tables like these in the dugout to look at, but he should know already the principles described herein.

Next week will be the last in the RE24 mini-series and we’ll discuss weird situations and try to put it all together.  Thanks for reading!

23 thoughts on “How Productive Are Productive Outs?

  1. Thanks, Patrick. It seems like a bit more than a “little” research, and your appreciative audience is grateful. One would think that, in a rebuilding year when expectations are low, it would make sense to try using this sort of information to determine game strategy. Nothing to lose, right?

    • Yeah, definitely nothing to lose! Like I said in the closing, I don’t even think a manager needs to ever see a chart like this… he just needs to understand that every decision he makes needs to weighed against all possible opportunity costs of making said decision.

      I am from the school of thought that more information can’t ever hurt. People cite “analysis paralysis” but I think that’s just a catch phrase people use to sound smart.

  2. Informative article, the only thing I think it is missing is game time situation. To be a sacrifice fly or bunt is much less valuable in the early innings than it may be in late inning situations. I have no data to back that up, it just seems practical to be that a team ought to be going for a big inning early rather than just trying to get 1 run across. But in a late inning particularly the ninth where that run may tie, allow a team to go ahead or actually be the game winner then that is more of a productive out. I do agree that that statistical date aside the logic of an sacrifice is more productive with one out than no outs.

    • You are correct. Context is missing. Doing all the topics on 1 article makes the article too long. Was planning on addressing this next week, but I might wait until later in the season. Three RE24 articles in a row might be a bit much, even for the hardcore statistically minded folk.

  3. Good stuff. In the Finnegan situation, there were not league average hitters coming up next. The next two hitters were going to be Billy Hamilton, whom they had to leave in for defense, and DeJesus, who was playing because Cozart was hurt. The Reds were up 3-0, so Price probably was playing for one run–one guy’s getting a hit or lucking into a productive out, rather than hoping for 2+ hits out of the guys coming up, which would have scored more than one run.

    The RE24 data is interesting, but in practice the team’s strategy is often dictated by down-and-distance things like the score and inning, who is expected to hit, and how the defense may react such as by changing pitchers.

    • Yep. There’s plenty missing, as the topic is too large to cover in a single article.

      Whether or not league-average hitters were coming up next is inconsequential, since we’d still be comparing the same baseline. For example, rather than lowering RE by .264 and .152, perhaps it’s only lowered by .230 and .120, or something. Still, the relationship remains even if the numbers are different.

      Generally speaking, the numbers themselves don’t matter. What matters is their direction.

  4. As you said, there are some fringe cases where it’s ok to go for a productive out but these are fringe cases based on who’s at the plate, how the defense is playing the hitter, who is pitching and his tendencies to miss bats or induce ground balls, who is actually the baserunner, inning, and score. There has to nearly be a perfect storm of those factors that work against swinging away for the bunt to be called for. This was especially true in the Finnegan situation where the defense was defending against the bunt. When managers start letting hitters swing away in this situation more often and it starts working, we’ll see it more and more. Right now though, the game is still “Runner on 2B, no outs, pitcher up, you bunt!! It’s frustrating sometimes.

  5. This is a great article and a great series. I agree, and have thought for a while, that breaking baseball down into these base-run states and looking at how different events change the run expectancy is the clearest and easiest way to think about baseball strategy.

    The one issue I have with the Finnegan example is that you don’t include the possibility of a double play. With a terrible hitter at the plate, like most pitchers are, and one with a 60+ percent ground ball rate, one of the main reasons for him bunting is to avoid making TWO outs. I suspect that if you put in a healthy double play chance into that table, the bunt would be just about equal if not ahead of letting the pitcher swing away.

      • In the Finnegan scenario above, the runner was on 2nd correct? Not that a double play is impossible, but extremely unlikely regardless of who the hitter is.

    • Double plays are big concerns with pitchers batting in most cases. In this case, the runner was at second base, so almost no chance it would come into play.

  6. Then there was the old Dusty method of bunting the pitcher with Stubbs on 1st (80% steal rate) when every pitcher but Arroyo was 50/50 at best in getting it down properly. You don’t have to go to MIT to figure out that the math isn’t working out there! Dusty would be doing the same with Hamilton (batting 8th of course) if he was still here

    • Good call. Using an 80% “getting the bunt down” rate seemed a bit high for the ‘ol Redlegs over the last several years.

    • The argument not to steal (just playing Devil’s Advocate) is if the bunt fails you still have the runner on base. But despite the flawed sacrifice logic that when it actually works- it put you at a disadvantage, I see it not work quite frequently (not sure if the Reds are terrible bunters- but as a casual fan they sure seem so)

      But to my point- the average hitting only gets 2 cracks at the sacrifice, because they waive it off in fear of the third strike foul. But if your strategy is so sound to S to begin with, are there any statistics that shows the positive outcomes to reinforce waiving off the bunt? I know a batters’ obp goes way down with 2 strikes- so you put them at a huge disadvantage giving them only one strike no matter what the count.

      • I haven’t seen any analysis like that. But as you note, hitting with 2 strikes is a huge disadvantage. Joey Votto goes from something like one of the Top 25 hitters in history to a “league average” hitter, once he gets 2 strikes.

  7. Still reading the article, and love your writing, but I had to take a small break to be a grammar Nazi regarding a weird quirk of English.

    “…it would be preferable if the out would affect a positive change on the field…” – this is one of those rare cases where “effect” is correct as a verb. Generally, as I’m sure you know, effect is a noun and affect is a verb. There’s an infrequently used exception in both cases. Affect (as a noun) can refer to an emotional state. And when something causes a change, it’s correct to say “effect a change.”

    See: https://en.wiktionary.org/wiki/effect#Verb

    You may already know this and just made a typo, or maybe you’d never heard of the uncommon usage of effect. If you hadn’t, you have now!

    • Just curious, are you an editor? If so, there need to be more like you. My sister was an editor a long time and is a writer. There are so many things that go by the boards these days because the editors are either not very good or they are way, way too busy. I’m not even talking about a blog like this where grammar mistakes really aren’t that big of a deal. I’m talking about major printed publications and news sites.

    • Thanks for pointing out; certainly wasn’t a typo. I wasn’t aware of that case! I am now! 😉 Check back next week when I try to use it correctly!

  8. I suspect part of the reason this sort of analysis isn’t more common is that you can so quickly get into an overwhelming number of permutations. To suggest even further complicating the assessment, there are in my mind two different but valid ways to consider the implications: Run Expectancy and Odds of at least One Run. Depending on the game situation, either may be more important. Late in a game, scoring one run to tie or to go ahead may be more important than risking scoring none in an attempt to score several. I’d be curious to see how these data look, especially in terms of the sac bunt and the stolen base, when the change is to the chances of scoring at least once rather than to total run expectancy. And whether or not you get to that analysis, thanks for this good article Patrick!

  9. Bruce Bochy sent Jeff Smardzija up to bunt down 2-1 in the top of the 7th, with a runner on 1B and 1 out. Even if he got the bunt down successfully, he would have been looking at 2B and 2 outs. Smardzija did the worst thing though and that was bunting into a DP. The right move from the beginning there would have been sending up a PH and having him hit away. Smardzija was pitching well and his pitch count was reasonable, but Bochy realistically could have only expected one more inning out of him.

    I mention this because sometimes we lose site of how managers go about managing. Bochy played against the RE24 charts because he wanted to get another inning out of his starting pitcher. Just not a good move there.

    • In this scenario, I’d have to imagine that Price should do the same thing and bunt the pitcher considering the state of the Reds bullpen…

Comments are closed.