The Evidence for Starting a Reliever
Over the weekend, I attended the Sloan Sports Analytics Conference, and on Friday, I attended a couple of the presentations of research papers that were baseball-oriented. The first one, presented by John Salmon (based on work he did with Willie Harrison), was titled “Bullpen Strategies for Major League Baseball“, which is obviously a hot topic in the game right now. But while the title made the paper sound like it might touch on Andrew Miller’s postseason usage, the part of the talk I found most interesting actually had more to do with starting pitchers.
In the first half of Salmon’s presentation, he talked about home field advantage, and particularly, how the data shows that one of the primary factors in a home team’s edge might be that their starting pitcher just takes the mound first. While Jeff Zimmerman wrote about this here on FanGraphs back in 2013, I must have missed that piece, because this was new information to me. But Salmon is correct that the data is clear that the gap in performance between the home and road team in the first inning dwarfs the difference in every other inning.
| Team | 1st Inning | All Other Innings |
|---|---|---|
| Home | 0.323 | 0.320 |
| Road | 0.347 | 0.330 |
| Difference | -0.024 | -0.010 |
This table covers data from 2002-2016, but in recent years, the first inning gap has been close to 30 wOBA points, while it drops to about 10 wOBA points in innings after the first inning. According to Salmon’s calculations, this huge difference in starting pitcher performance means that the first inning accounts for about 21% of a home team’s overall advantage.
So while a lot of the prior research done on home field advantage has focused on ballpark familiarity, the comfort that comes from sleeping in your own bed, or even subconscious umpire bias, Salmon suggests that perhaps a good chunk of why home teams win 54% of the time is that the away starting pitcher has to sit in the dugout and watch his teammates hit after finishing his pre-game warmup.
Zimmerman’s post from 2013 includes a couple of graphs which seem to support the point.
Jeff’s research showed that road starters generally throw a little harder than home starters in every inning besides the first inning, but that effect is not true in the first inning, which could be consistent with some kind of cool-down effect. He also showed that first inning fastball velocity declines even more if the road team builds a big lead in the top of the first inning, which of course would require even more time for the starter sitting on the bench after finishing their warm-up throws.
This isn’t a smoking gun or anything, and since this was just a portion of Salmon’s presentation, he didn’t dive too much into the possible alternate explanations for the first inning performance differential, or prove definitively that it is because of a cool-down effect. I did run the idea by a couple of Major League pitchers, though, both of whom said they could buy that the gap in time between warm-ups and actually throwing their first pitch could be part of the reason that road pitchers tend to do relatively worse in the first inning.
If the cool-down effect was indeed the main cause of this performance gap, then Salmon had a potential solution; road teams should simply begin the game with a reliever on the mound, preferably one who was a good match-up for the home team’s leadoff hitter. The reliever would absorb the penalty associated with finishing his warm-ups and then not pitching for a while, but this could potentially be offset by gaining the platoon advantage and using an arm who would specifically do well against the guy you know is hitting first. Thus, Salmon suggests, the starting pitcher could finish his warm-ups during the first batter at-bat, and come straight into the bottom of the first inning without any cool-down period.
The people from Major League teams who were also at the presentation who I talked with called that idea unrealistic, given that it forces you to burn a reliever for just one batter early in a game, limiting the ability to get specialist match-ups in late-game, higher-leverage situations. But even that part of the suggestion seems unnecessary, at least in the American League, where the starting pitcher could theoretically push his warm-up back a bit, so that his routine finished during the top of the first inning. This probably wouldn’t work as well in the NL, where a starting pitcher might have to bat in the first inning occasionally, but AL teams don’t have to worry about that complication.
This could also have some potential impacts on how teams handle things in September, when pitching staffs expand, and the cost of burning a reliever in the first inning isn’t nearly so high. This kind of potential gain might not be large enough to incentivize change in April through August, but if you’re a contender fighting for a playoff spot in the final month of the year and you have access to basically unlimited pitching changes due to the expanded rosters, it might be worth it to try something different in order to gain an edge that could help save you a run in a game that could be the difference between making the postseason and watching at home.
And then, there is October. With frequent off-days and the highest stakes possible, teams are already experimenting with wildly different pitching usage plans in order to gain any advantage possible. If road teams could significantly improve their pitching performance in the first inning by simply using a reliever or two to handle the first inning, and then bring in the starter to begin the second inning, that’s a little less crazy when the impact of a single win or loss has a dramatic impact on your likelihood of bringing home a championship. In the winner-take-all Wild Card game, for instance, teams should already be unloading their bullpens, and that game might be where this idea has the most benefit.
Of course, we don’t know for sure that the difference in first inning performance is due to this time between warm-ups and taking the mound, and teams generally prefer not to mess with routines to maybe gain an advantage of unknown size. I wouldn’t expect this to become a trend any time soon, but when the Wild Card games roll around, it will be interesting to see if the road teams decide to mix things up a bit in an attempt to diminish the home team’s large first-inning advantage.
Dave is the Managing Editor of FanGraphs.
Why can’t the away SP spend the top half of the 1st in the bullpen completing his warmup?
“The people from Major League teams who were also at the presentation who I talked with called that idea unrealistic, given that it forces you to burn a reliever for just one batter early in a game, limiting the ability to get specialist match-ups in late-game, higher-leverage situations. But even that part of the suggestion seems unnecessary, at least in the American League, where the starting pitcher could theoretically push his warm-up back a bit, so that his routine finished during the top of the first inning. This probably wouldn’t work as well in the NL, where a starting pitcher might have to bat in the first inning occasionally, but AL teams don’t have to worry about that complication.”
I think you’re all minusing this solely because Dave has the main sentence hidden in the middle of the paragraph. “(t)hat part of the suggestion seems unnecessary, at least in the American League, where the starting pitcher could theoretically push his warm-up back a bit, so that his routine finished during the top of the first inning.”
So HappyFunBall’s suggestion does seem to fix the whole problem. Totally. Entirely. As noted below, if the NL starter does come to bat in the first inning, you’re already a happy camper anyway.
Dave, that quoted section refers to the proposal to use a quick opening pitcher: it says nothing about delaying warmups without using an opening pitcher.
What value does the opener add that couldn’t be accomplished by simply delaying the road starter’s warmups by 5-10 minutes?
“But even that part of the suggestion seems unnecessary, at least in the American League, where THE STARTING PITCHER COULD THEORETICALLY PUSH HIS WARM-UP BACK A BIT, so that his routine finished during the top of the first inning. ”
emphasis mine, as apparently it was necessary.
Using a reliever for the first batter doesn’t really do anything but maybe add a minute to the starter’s warm-up time. He still wouldn’t be able to get ready while the reliever is pitching so what’s the point really?
What if it’s Pedro Baez.
Exactly what I was going to post. There’s no need for the road SP to leave the bullpen prior to the bottom of the second.
He should continue warning up through the top of the first, and just trop straight from the pen to the mound like a RP.
Beat me to it. Yesh, i mean, duh.
If you were the NL team, you could just “start” one of your other starting pitchers, so that if needed, he would bat. If there is a rule that says the starting pitcher has to face a batter, he could just throw an intentional ball, and then bring in the actual starter. This, of course, would be awful, because you’d have the whole half inning change, one pitch, and then another break for a pitching change, which would make this process take even longer.
I would think that if anything like this caught on, MLB would kill it somehow.
“face a batter” means finish a plate appearance. So he’d half to issue a full intentional walk (now done without throwing any actual pitches), which would really mess up any advantage gained, especially because the leadoff guy is usually either the fastest or one of the fastest guys on the whole team.
But because he doesn’t have to throw anything now, you could use your 25th man, the fringe bat who won’t play except injury or extra innings.
If an NL team has to bat his pitcher in the first inning, then it sounds like a good problem, as they will have guaranteed to have scored at least 3 runs by the time the pitcher bats (if you bat the pitcher 9th).
Yeah, I thought this too that the advantage is a moot point if the NL pitcher is batting in the 1st inning given the likely lead he would have to already have.
But I did then wonder… can you move from the bullpen to the dugout during an inning? In other words, could the NL pitcher warm-up anyway during the top of the 1st and then move to the dugout once, say, the 7th batter gets up? If not, then the pitcher would have to be in the dugout to start the game and can’t warm-up later.
Easy fix. Have your throw-day starter throw immediately before the game begins. Sit him in the dugout. Then if the pitcher comes to bat, he can do it and even pitch to a batter if need be in the bottom of the inning. Then bring in your regularly scheduled starter.
But really all this points to, if anything at all, is a modification in warm-up routines. Not wholesale tactical changes.
But if your starter is in the announced lineup, then you would lose him. I believe this is the applicable rule: 5.10 Substitutions and Pitching Changes- 3.05(a) The pitcher named in the batting order handed the umpire-in-chief, as provided in Rules 4.02(a) and 4.02(b) (Rules 4.01(a) and 4.01(b)), shall pitch to the first batter or any substitute batter until such batter is put out or reaches first base, unless the pitcher sustains injury or illness which, in the judgment of the umpire-in-chief, incapacitates him from pitching.
Thus the throw-day starter is the “announced” starter. Of course that means your announced pitcher is a sham every single time. Or it means you’re committed to pitching the sham starter for at least one batter. But the notion of throw-day starters coming on in relief is not a new one, and by naming him your 1st Inning Starter you’re accomplishing the same thing. In fact, you’re adding an element of predictability to it all.
That’s true. But then Vegas might have a conniption if you don’t pull the starter for the presumed guy. Some guys like to go on 4 days rest all the time, some guys like or need that extra day if there was an off day speckled in the last few days. I just don’t know that the math ever really makes this worth it.
If you started a reliever wouldn’t you still have to announce it before the game? Therefore giving the home team a chance to change their leadoff hitter and erase the platoon advantage?
How much of an advantage is this really? We all know SPs are creatures of habit. Is this slight advantage worth the SP changing his routine and upsetting him? I think these kinds of studies, while mathematically show a slight advantage, ignore the psyche of the human player.
Not necessarily disagreeing, but taken to the extreme, this argument could be used to justify keeping hitters in the same lineup spot every game.
I feel like I’m missing something obvious. Why do Away Pitchers consistently have a higher fastball velocity than Home Pitchers after the first inning?
Maybe the harder throwing in the first inning causes a little early fatigue?
Most likely it’s that home pitchers are more relaxed and don’t overthrow.
Another explanation for the correlation between big leads and lower velocities is that the away pitcher simply wouldn’t throw at max effort with a 5+ run lead. Someone made this point in the comments to Zimmerman’s 2013 article, and I find it pretty persuasive.
But is this point accurate? It reminds me of the Jack Morris pitching-to-the-score argument.
I wouldn’t say that all pitchers necessarily pitch to the score or anything, but I would say that it is very unlikely you’re getting the same effort in a 5 run game in June that you’re getting in a tie game in October.
Given that pitcher compensation is based on stats, it seems unlikely to me.
And that’s a wonderful thing about baseball — when a team is down 20-2, its players still have an incentive to try their hardest.
Another factor to highlight is there is a strong likelihood the reliever used for first inning would be a below and/or average pitcher. If teams utilize this strategy, I doubt the manager would put in Andrew Miller for one or two innings of work if he can put in Boone Logan. Because, obviously, a manager would want to use Miller for higher leverage innings.
The effect is the wOBA difference would be greater. Using the above example, Kluber is obviously better than Logan. There is a greater chance Logan posts a higher 1st inning wOBA than Kluber, thus negating the intended benefit.
With Kluber pitching, instead of being down 1-0 on the road, you open it up to possibly being down 3-0 and a wasted reviler in Logan.
Definitely opening it up to more risk by opting for relievers in early inning, low leverage situations. Not a fan of this strategy.
While I like the “reviler”, I don’t think this argument makes sense. This is an average-leverage situation (LI = 1), sure the manager saves his high-leverage guy, but why would he feel moved to pitch his junk guy?
Please forgive the potential stupidness of this question, but who announces their starters first? Visitors or Home? Pitchers or hitters? How would you ensure the platoon advantage with your one-out reliever starting if the opponent changes the lineup without penalty?
Don’t they exchange lineups at home plate at the same time?
I mean, right? So how would you know who was leading off? As a manager, do you anticipate their leadoff hitter, then get burned when your opposite number anticipates your move and leads someone else off? (I think I just sprained something in my brain.)
I’m not arguing the mathematics of the idea; just wondering about the mechanics of it.
I’ve always thought that starting a reliever made sense for another reason: you know that the opposing team’s best hitters are due up.
Why not run out a 7th-inning type (especially if your team has a deep pen) to start against the other team’s 1-2-3 hitters, factoring in platoon issues, and then go to the starter after. That way you ensure you have a higher quality pitcher in against the top of the order at least once?
We know that teams score more in the 1st inning because their best players are coming up. Why not do something about that directly?
Managers have to manage pitchers and bullpens over an entire season. If we played one game in a vacuum (like the wildcard as mentioned) then this could work.
Why wouldn’t the reliever have the same cool-down penalty?
Relievers are used to both warming up on the quick, and occasionally having to make do when an inning continues longer than expected. He could just start his warm up in the top of the 1st as if it were any other random half inning.
Combining this point and one made above, a fundamental flaw in this proposal emerges.
First, let’s acknowledge that we’re setting aside the fact that most relievers are busted starters for a moment. We’re basically saying that relievers can come in and throw to one or two batters before the ‘real’ starter toes the rubber, because relievers are ‘used to’ a lesser warm-up than starters are. They have some mental/physical ability (which we don’t fully understand) to do that better than starters can. But somehow, putting a starting pitcher on the mound with one out in the first, or a runner on base, or worse, a run on the board for the home team will have no effect on his performance. Huh?
I’m all for experimenting here, but to suggest that that this idea of putting a reliever, even a good one, on the mound to throw to the first batter or three is a way of optimizing your pitcher usage seems specious to me. There are a lot of things at play here that we don’t understand, and we’re relying on one of them (that relievers need less warmup for…some reason) to base an assumption around another (that starters can enter the game after the first batter has been faced and perform as they normally do).
I’d really like to see a team manage a game this way:
1st Inning – 7th inning type (someone suggested this as well)
2nd – 7th Innings – “Starter” – this would/could offset platoons for at least the first 3 or 4 batters. It also helps alleviate the 3rd time through the order penalty.
“Closer” or more accurately, Best RP, comes as early as the 7th and when a threat begins. He’s trained to finish that inning and pitch the next inning.
or
I also would love to see a team experiment with a full stable of SPs that are very clearly not Aces. They pitch 9 batters at a time and are expected to be available every 3-4 days. If a team platoons against you heavily, they pitch to one batter, and you switch the platoon to your advantage.
I’m not sure why John Salmon (or Dave Cameron for that matter) doesn’t realize that his solution – a reliever starting the game – isn’t really a solution but just the same thing with a different pitcher. Again, the reliever would still have the same cool-down effect as a starter and the starter would still have the same problem of timing his warmup to be ready for the 2nd batter (instead of the first as it is now). Unless the reliever times his warm-up properly which if that’s the case why wouldn’t you just have the starter do that?
In general it takes a starting pitcher much more time to get ready than it does a reliever.
But pitching to one batter doesn’t really do anything for the real starting pitcher. What is the most it can add – three minutes? If the batter puts the first pitch into play then it does nothing but add a few seconds. So you are right back to the same problem that starting a reliever is the solution to. Then the solution would be for the reliever to pitch a full inning – but again, that’s no real solution since the starter would still be uncertain of when exactly he’s going to pitch so you haven’t really solved anything. And so on.
Most relievers can be ready in 5-10 minutes. Much easier to plan for than the 45 minutes to an hour plus that starters typically take to get ready.
I understand that but how does a reliever pitching to one batter improve the starter’s warm-up process? What Salmons and Cameron seem to be saying is that the solution is in the 5 minutes (at most) for the first batter of the bottom of the 1st. That makes zero sense. If it takes a starter 45 minutes to get ready how can 5 minutes make a difference? Either way he’s going to be warming up through the top of first inning so why would you start with another pitcher?
Dave, Did you wonder if the reason a starters speed declines over the game is due to them being bad pitchers and probably shouldn’t be starting the game in the first place? I looked at the per inning pitch speeds of the top 15 WAR per/GS starting pitchers over the last 10 years and 11 of them had higher speed in the 9th than the 1st. Verlander had the highest differential (1.88 mph). Darvish also had over 1 mph faster and 3 others just under 1 mph. The 4 pitchers that had a lower speed were Kershaw (-.5); Wainwright (-1); Kluber (-.22) and Price (-.87).
I tried to look at the worst WAR per/GS pitchers. And while the majority had declines, I ran into a problem with the data for two main reasons. 1) limited stints where they pitched into the 8th and 9th inning and 2) Every bottom 15 starter relieved games with the average being 20 appearances each. This would skew their late inning speed data. Look at Hector Santiago. He never made it into the 9th as a starter, but he finished 26 games. So his 1st inning speed was 91.52 while his 9th inning speed was 94.19.
So in regards to the thought of using a reliever to start a game if you don’t have quality starter for a playoff game, why not use the 4 and 5 starter and have them go all out for 2-3 innings? Santiago is almost 3 mph faster in spot situations. This could get you to the 4th – 6th inning where you can have the relievers do their job.
So, just a quasi thought exercise here to see if I understand this right.
Road pitchers are on the whole, 0.010 worse on the road, and are 0.024 worse in the first inning. So really, they are only 0.014 worse in an the first inning, than they are the rest of the game. So, if the back of the envelope calculation tells you 20 points of OBA is worth ~10 runs/600PA, then it stands to reason that 14 points is worth ~7 runs/600PA.
A brief glance at the stats from last year says every MLB team pitched ~1500IP, and Total batters faced were ~6000. So, the average inning has 4 batters faced, give or take.
In short, that means that is giving up ~7 runs extra above average in the first inning, per 150 innings. Or, 1 run above average every ~21 road games, or ~4 times a year.
I can’t see where the math ever makes sense to say that you, on average, get better performance from running what would have to be quality reliever out there in what can only be a low leverage situation.
The only way I see something like this being reasonable, is if you were on the road against a team like the 2016 Indians with a rested bullpen in a do or die game where they can effectively make it a 4 inning game and as a manager you had a bullpen where your ~3rd option is better than your currently scheduled starter. That is the only way I can see the leverage calculus of making sense.
Your math looks fine, but it’s not fair to expect this tactic to be worth a lot of wins per year. It’s aiming to carve a slice off home-field advantage, which itself is only a couple of wins per year. The question is on the order of, could we turn one loss into a win.
Travis and Jeff deserve all the acclaim they get, but to me Dave continues to write the most thought-provoking articles.
You learn something new every day — I was completely unaware of that Rule 3.05 (a). Without that, there would be an even bigger loophole; the visiting team could name, say, their best pinch-hitter as the starting pitcher – and if he was a good enough hitter, even lead him off – then replace him with the real starting pitcher in the bottom of the 1st. If you had a big 1st inning, he might even get two ABs out of the deal. Oh well!
Was this supposed to run on April 1st? Or is the second half about starting the game with a reliever an inside joke?
If not the potential solution Salmon presented is quite possibly the dumbest thing ever written on fangraphs. There can be absolutely no difference between the starter facing the first batter or the second if he is warming up during the top half of the first inning. What possible difference could the few minutes it takes to pitch to the first batter make when it takes 10 times that time for the starter to warm up? No matter what the starter is still going to be guessing when he first starts pitching – having a reliever pitch to the first batter doesn’t change that at all!
The more I think about it, the more I think it makes sense for teams to employ one or two fairly good relievers as 1st inning specialists and send them out for one inning on pre-scheduled days, particularly on the road. In addition to the advantages discussed in the article, I can think of the following benefits:
(1) With decent planning, you can get more innings out of good relievers pitching them this way because you won’t ever get them up and sit them down or use them for a batter or two.
(2) You never waste innings for these pitchers. They get guaranteed innings in important spots against good hitters.
(3) Because you sometimes know after the first inning whether a game will be close, you can avoid wasting starts by your best starters by holding them out if the game is out of touch in either direction.
(4) You can get your starters deeper into the game before the third time through the order penalty kicks in.
(5) You can better control starter innings because you will have one more inning of information about the leverage situation when you reach the point where you have to make borderline judgment calls on whether to pull him.
(6) If the pitcher’s spot comes up in the top of the second inning (frequent, usually indicates scoring opportunity), you can use a pinch hitter rather than having to let the starter bat.
(7) If you are playing a team playing a generally righty or lefty heavy lineup but with the opposite orientation at the top of the order, you can set things up to have the platoon advantage the first time through the order against both parts of the lineup.
I like this idea a lot! In fact …
http://www.baseballprospectus.com/article.php?articleid=31133
Didn’t see your article. I obviously agree. A lot of our arguments overlap but not all.