Player substitution and winning: is less still more?

For decades in major league baseball, a predictable inverse relationship existed between the number of substitutions made during a game and the likelihood of winning that game. Teams that didn’t make substitutions were more likely to win than teams that did, with that winning percentage declining with each additional substitution made.

In today’s game, with at least three pitchers (starter, setup man, closer) in every team’s game plan everyday, the expanded pitching staffs necessary to sustain that approach have reduced bench size and, presumably, limited opportunities to use tactical substitutions on offense. Or, have they?

After the jump, more on the relationship between player substitution and winning.

First, let’s look at the number of pitchers used in a game, and how well teams have fared as a result. Here’s the National League since 1946.

NL Pitchers Used 1946-2014The vertical bars are the percentage of team games each season in which a specific number of pitchers are used, measured against the left axis. The lines are the winning percentage when using that number of pitchers, measured against the right axis. Thus, when a pitcher records a complete game, his team will have about an 85% chance to win, a relationship which has held throughout this period.

Of course, the number of those complete games has dwindled to almost none, going from up to 40% of games to just 1% or 2% now. In fact, the proportion of games today with two or three pitchers used is almost the same as the proportion of complete games 60+ years ago. With two pitchers used, teams started winning more often than losing about the mid-1950s, improving to about a .600 winning percentage from the early 1980s to the mid-1990s, before dropping back to just above .500 in this century. But, look what’s happened to the winning percentage with 3 or more pitchers. Consistently below .400 until the early 1980s, today there is essentially no difference in winning percentage based on pitchers used in the 90% of games in which a team uses three or more hurlers.

The next chart shows the AL over the same period.

AL Pitchers Used 1946-2014

Similar picture to the NL, with the notable difference in the winning percentage when using just a single pitcher. About the same as the NL around the .850 level prior to the DH rule, starting in 1973 that winning percentage immediately dropped down to the high .600s, even as the number of such games shot up 50% (from 20% to 30% of games), indicating that not having to remove a pitcher for a pinch-hitter didn’t really help a team’s chances to win, as the necessity of a pinch-hitter merely emphasized the likelihood of losing such games. The number of AL complete games is now only slightly more than in the NL, but the winning percentage in those games still lags the senior circuit as those extra complete games are, in most cases, complete game losses, as indicated in the chart below.

Complete Game Losses 1946-2014Now, let’s like at game results based on the number of batters used. Here’s the NL picture.

NL Batters Used 1946-2014
The trend to lower winning percentages when using more than one batting substitution has held throughout the period, though that advantage has dwindled with the continuing convergence of the winning percentage lines. The biggest change has been a drop in winning percentage since the mid-1990s when using one substitution or less, and that, also starting about that same time, using one substitution has produced better winning results than using none. The mid-1990s was the start of the sudden upsurge in offensive production that peaked in 2000; but, why that would be reflected in this result is not immediately apparent to the author.

The shortening of benches has not resulted in fewer batters being used in NL games. On the contrary, as pitchers other than starters or middle relievers are seldom permitted to come to bat, other players must take the pitchers’ turns at bat in the late innings.

The AL story is somewhat different.

AL Batters Used 1946-2014Here, the convergence of winning percentages lines is not gradual, but sudden and immediate with the introduction of the DH. As in the NL, the winning percentage when making exactly one substitution has become superior to when making no substitutions, even though the number of games without an offensive substitution has crept up gradually to around 30% today, compared to about 25% immediately after the DH rule was adopted. Similarly, the number of games with 11 or fewer batters, around 65% to 70% of games from the mid-1970s to the early 1990s, is now consistently around the 80% mark, probably an indication of reduced bench size in this century.

What sort of offensive substitutions are being made? The argument against shorter benches is how that limits a manager’s ability to pinch-hit in critical situations. However, the data do not support that supposition, at least not in the NL, as can be seen below.

Pinch-Hitters Used 1946-2014The chart is showing pinch-hitters used over the course of a season on a per team basis. Thus an average figure above 250 times a season in the NL indicates a bit more than three pinch-hitters used for every two games, a figure which has declined only slightly since a 1980s peak. A different story in the AL where the number dropped sharply, as expected, after the DH rule came into effect, and then has declined steadily since the mid-1990s, likely a result of shorter benches.

However, as reader David Horwich points out, the steady rate of NL pinch-hitting over the past 40 years does not mean that NL managers are not challenged by shorter benches. Rather, as shown by the chart below, batting appearances by NL pitchers have declined steadily over this period (admittedly due, in part, to expanded inter-league play) meaning more NL pinch-hitting is of the “automatic” variety (i.e. batting for the pitcher) rather than for a matchup advantage at a key point in the game.

Batting PAs by Pitchers 1974-2014

Finally, how has the use of pinch-hitting specialists been affected by shorter benches? Here’s that picture.

Pinch-Hitting Specialists Used 1946-2014The pinch-hitting specialist has become pretty much an NL-only phenomenon, increasing in the senior circuit starting in the mid-1970s, likely in response to the decline in offense and a greater focus on small ball. The number of such specialists has consistently averaged about one-and-a-half per NL team (indicating about half of NL teams employed both a right-hand hitting and left-hand hitting specialist) until the current decade when that number has dipped to about one per team, probably a reflection of reduced bench size. AL pinch-hit specialists have been rare since the DH rule, with only about one team in five having such a specialist through the 1990s, before reducing to none about ten years ago, with Dave Hansen the last AL player to have 40 PH appearances, in 2004 and 2005.

My thanks, as always, to the indispensible Baseball-Reference.com, for the data used for the illustrations for this post.

0 0 votes
Article Rating
Subscribe
Notify of
guest

6 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
David Horwich
David Horwich
9 years ago

Interesting post, Doug. I’m not sure the “Pinch Hitters Used” graph provides enough information to support a conclusion either way about the effect of shortened benches on the use of pinch hitters. On the one hand, we see a decrease in AL use of pinch hitters over the last 20 years, which is at least suggestive that shorter bench = fewer pinch hitters. But although the NL rate has hardly changed over those same 20 years, it could be that this is primarily a product of the increasing number of pitchers used, creating more essentially automatic pinch hitting for pitchers,… Read more »

Doug
Doug
9 years ago
Reply to  David Horwich

Thanks David, What is not quite so much work is to come up with PH frequency by BOP. With so much double-switching happening now, you never know what spot the pitcher is batting from in the late innings, but it would probably provide some indication of whether there was less pinch-hitting happening now than in the past. Intuitively, there *must* be less pinch-hitting for other positions because NL pinch-hitting has been pretty constant for the last 40 years or so, but the number of batting PAs by pitchers has declined steadily, from close to 450 per team in the mid-1970s… Read more »

Voomo Zanzibar
Voomo Zanzibar
9 years ago

Craig Kimbrel, career

289. IP
1.43 ERA
.903 WHIP
_________

Orval Overall, 1909

285. IP
1.42 ERA
.996 WHIP
_________

Paul E
Paul E
9 years ago

Doug, Thanks for the interesting post. This starter, set-up, closer “thing” was the developed to the extreme by Tony LaRussa with the Oakland A’s in the late 80’s – early 90’s. It worked well for him and that was the end of it-everybody adopted the formula… As a kid, I watched Gene Mauch do things like move Rojas to LF (pulling Alex Johnson), move Amaro to 1B (removing Dick Stuart), bring in Wine at SS (for Amaro), and bring in Taylor at 2B (for Rojas)- all in the 8th & 9th innings! It worked…or seemed to work in the sense… Read more »

brp
brp
9 years ago

Does the “batters used” for NL include pitchers who didn’t actually bat, or what? Or I am reading the chart wrong – it seems absurd to me that 75% of NL games included 13 or more batters when most teams only have 12 or 13 position players. I know 1-2 pitchers will bat during a game but are the NL teams really emptying the bench that often? The later graphs show it certainly isn’t because the pitchers are getting PAs.