Why IP Is a Poor Indicator

Innings pitched (IP) seems to be the standard for judging a player’s workload. Sure it will tell you how deep into a game a pitcher went and it’s often used as a measure of pitcher durability, but it tells you nothing about a pitcher’s effectiveness. A far more useful stat is the pitch count during each particular outing, or even better pitches per innings pitched (P/IP). I think we can all agree that all innings are made differently. A pitcher can throw three pitches or it can take 61 pitches as evidenced by Steve Trachsel (1997 – Chicago Cubs) and still get credit for 1 IP. Actually I think it’s possible to throw zero pitches and get 3 outs, but I don’t have the motivation to look up the rule at this particular moment.

Here are some stats for three players in the 2015 season.

Player GS W L IP
Player 1 27 11 10 159.2
Player 2 26 12 7 171.2
Player 3 30 12 10 169.1

All the players in the table above have very similar peripheral statistics, aside from an IP difference of 12 between players 1 and 2. From looking at these stats it’s a toss-up as to who has had the most successful season — do you choose player 2 since he has the most IP or player 3 since he’s made the most starts? In the table above Chris Heston is player 1, Matt Harvey is player 2 and Yovani Gallardo is player 3. What really separates the players is the pitch counts and P/IP.

Chris Heston – 2461 Pitches and 15.4 P/IP

Matt Harvey – 2533 Pitches and 14.8 P/IP

Yovani Gallardo – 2959 Pitches and 17.5 P/IP

Chris Heston has 12 IP less than Harvey but has thrown 72 fewer pitches this season. Harvey and Gallardo have thrown about the same amount of innings, but Gallardo has thrown 426 more pitches this season. The reason I chose Harvey as one of the pitchers for this comparison is due to the very public feud between the Mets, Boras and Harvey. In case you missed it, there was a disagreement with the innings limit imposed on Harvey in his first season after Tommy John surgery. Boras wants the Mets to stick to 180 IP while the Mets thought it was more of a soft cap. I wanted to look at the relationship between the IP in a season and the total number of pitches thrown. Luckily this data was readily available for download via FanGraphs, but only pitch counts back to 2002 were available. Below is a plot showing all pitchers who threw more than 100 innings in a season compared to their pitch counts. The data has a linear relationship, with the red line showing the mean and the outside black lines are the prediction intervals where we would expect 95% of the observations to fall within.

Now based on the 180 IP limit imposed on Matt Harvey, a linear model predicts that a pitcher would throw 2867 pitches in a season with an upper limit of 3158 and a lower limit of 2576. Now this means that at 180 IP we can reasonably expect a pitcher to throw between 2576 and 3158 pitches. Now for a guy coming off a major surgery, doesn’t a range of 582 pitches seem a bit extreme? It basically amounts to a difference of 5 complete games’ worth of pitches. In the plot below I also highlighted an innings range based on the range of innings where a pitcher throws 2867 pitches in a season. Now most importantly this range extends from 160 to 200 innings.

The medical team could just have easily set a limit anywhere between 160 and 200 IP. This is why an innings limit doesn’t work well in this situation; there is just too much variability in the data. In the future it will probably be a better idea for team officials and the medical staff to discuss a pitch limit over a season instead of an innings cap. Since the main goal of limiting a pitcher’s workload is to reduce stress on his arm I think the plot above does a good job showing that innings limits will have very little effect on actually managing a pitch count. Harvey is obviously thinking about the long term here because I know he doesn’t want to go through another surgery. After a second Tommy John the chances of a pitcher returning to the majors drops to somewhere around 30%, not to mention the drop in potential future earnings.

So I’ve shown you why I don’t think IP is a good indicator and now I’m going to show you why I think pitch counts and P/IP should be more important statistics.  Based on the linear model shown in plot 1 the formula to predict pitches in a season is as follows: Pitches = IP*14.5 + 256.9. Now the intercept for this model is 256.9 which suggests that if you don’t throw a single inning in a season you would still be expected to have thrown 257 pitches. Obviously there is something going on at the lower inning totals, but we are going to ignore that for the purpose of this article. As an added note, the lower prediction interval from plot 1 has an intercept of -33.975, so we are very within range of showing 0 pitches for 0 IP from this model.

Player IP P/IP P/IP Rank Actual Pitches Expected Pitches Difference Predicted IP
Chris Heston 159.2 15.4 24 2461 2565 -104 152
Matt Harvey 171.2 14.8 11 2533 2739 -206 157
Yovani Gallardo 169.1 17.5 84 2959 2708 251 186.1

Heston and Harvey both rank very high in P/IP among qualified starters while Gallardo is dead last among qualified starters. Efficiency is key here. Should Harvey be directly compared to Gallardo based on IP? No, absolutely not, Harvey is among the most efficient pitchers in the game this year. He has been able to get through innings while keeping his pitch count down and most importantly reducing stress on his arm. An inverse prediction based off pitch counts was used to predict the IP in the table above. Based on their pitch totals from this season Harvey and Heston have “thrown” less than their IP totals suggest and Gallardo has actually thrown quite a bit more. This has a big effect on that innings cap imposed on Harvey for this season. His stats show that he’s thrown 171.2 IP, but based on the number of actual pitches he’s thrown in game situations his number may be closer to 157 IP. Does that mean he should have the equivalent of 23 IP left in the tank for this season? Well that’s not up to me, but IP should less important than total pitches.

One thing I didn’t look at this article was the proportion of pitches thrown throughout the 2015 season. It’s been in the back of my mind, but I don’t have a reference for what the most stressful pitches are on a pitchers arm. I think it’s safe to assume that all pitches are not equal. Let’s think a Dickey knuckleball vs. Chapman fastball. The amount of effort needed for each pitch type is likely highly dependent on the pitch speed and type, but to simplify things here I’ve just assumed that all pitches are equal. We also need to realize that all pitchers are not equal, whether it be mechanics or individual variation in abilities. I was curious to see where Mark Buehrle’s pitch count (leaderboard here) lined up with all other pitcher since 2002 and lo and behold he’s thrown the most pitches since records became available. Obviously he doesn’t throw as hard as many of the other guys in the league, but that hasn’t stopped him from being a workhorse and one of the most effective pitchers over the last decade.

Print This Post

newest oldest most voted

Nice work – you put into words what I thought about when I read the news about the Harvey dilemma.

Ed Ward
Ed Ward

The same amount of pitches for one pitcher may cause more damage than another. Why not use the pitch velocity and control as the indicator when he should be rested or shut down rather than an arbitrary number?


Has anyone found a site that counts total pitches thrown for minor-leaguers? Unless I missed something obvious I thought that I had searched them all. I am pretty sure MLB.com still shows ‘NP’ for the majors.


One thing I think that would be interesting to look at would be the standard deviation of pitches per game.

Pitchers who throw 110 pitches a game on average but only throw between 105 and 115 might be better off than those who throw 100 pitches per game on average but anywhere between 75 and 125.

Being regular in rest and in workload would probably be beneficial for injury prevention.


If we’re getting granular, yes pitches matter more than innings.

But, as mentioned, it likely matter what pitches are being thrown as well.

Furthermore, there is the concept of “stressful” pitches. Pitching with men on base, pitching later in games when tired … put more stress on a guy’s arm than throwing free and easy with the bases empty early in the game.

That’s why biomterics are so important. Rather than just trying to guess the best proxy for a player’s workload, MLB needs to find better ways of tracking what’s actually going on in a pitcher’s arm as the season progresses.