If a starter throws many innings then he helps his team by reducing the chance of his bullpen blowing a lead. A starter that throws 5 innings and allows zero runs has the same ERA as a starter that throws 8 innings and allows zero runs but clearly everyone would rather have the second pitcher. And obviously, if a starter prevents runs from scoring then he makes things much easier for his offense.
I was wondering whether there was a simple statistic that measured when a starter went deep into a game while allowing only a few runs and therefore gave his team a high chance of winning. The first stat that came to mind was the quality start. A quality start is a statistic for a starting pitcher defined as throwing at least six innings while allowing three or fewer runs. This statistic was originally created in 1985 by sportswriter John Lowe. Rob Neyer discussed the quality start with John Lowe and they discussed his reasoning. The article stated that:
The foremost attribute of this statistic is that it shows exactly how many times a pitcher has done exactly what his job is – pitch well enough for his team to win.If a pitcher goes at least six innings and gives up no more than three earned runs, he has done his job. He has taken his team to at least the seventh inning, the stage where the good relievers usually can start coming in. And even if his own team is being shut out, it is behind by no more than 3-0 (barring a faulty defense that has caused some unearned runs). A 3-0 deficit can be erased by one or two big hits.
In this bullpen era, a starter's job is not to throw a complete game. It is not even necessarily to win the game. It is to make a quality start -- to give his team a decent chance to win.
On first glance, this statistic seems to be an easy way to measure each time a starter goes deep into a game while allowing few runs. Joe Posnanski notes that teams win about 65% of their quality starts. Baseball Reference questions the validity of the quality start but notes that teams win over 65% of their starts. Over a 162 game season a team that wins 65% of their games would go roughly 105-57. If that isn’t success then I don’t know what is.
There’s just one small problem. It is possible using the Baseball Reference Play Index to determine how many times a team won a game given certain circumstances such as the number of innings a starter pitched and the number of earned runs that he allowed. From 2000-2014, when a starter throws six innings and allows three earned runs his team has won 1,459 games and lost 1,670 games. That is good for a 46.6% winning percentage while the starters themselves have a 47.1% winning percentage. When a starter goes seven innings and allows three earned runs over that same time period his team has a 48% (1073-1164) winning percentage. I don't understand how a starter can be considered successful if he puts together a performance that would cause his team to lose more than half the time.
It appears that a starter can throw a quality start but still not give his team a 50% chance to win. I propose creating a new stat that measures this based on the following conditions.
- It should be defined by the number of runs allowed instead of the number of earned runs allowed.
Regardless of fault, if a starter allows six runs his team will probably lose. This statistic is trying to measure whether a team has received a pitching performance good enough to allow them to win and if a pitcher has allowed six runs then his team will most likely not win. Unearned runs may not fully be the fault of the pitcher but ultimately they count for the same amount on the scoreboard.
- It should be defined based on both the historical starter winning percentage as well as the team winning percentage for a given scenario.
The winning percentage of a team when they receive such a start should be high enough presuming an average lineup, defense and bullpen for that team to make it to the playoffs. There have been three teams with more than 90 wins that wouldn’t have made to the playoffs if there were two wild card spots since 2002. I think a team winning percentage of 56% (about 91-71) should be a reasonable cutoff line.
In order to give a starter credit for such a start it is necessary to make sure that it was his contributions that led to the win and not the bullpens. Determining the starters winning percentage for a given scenario (wins/total games) determines whether the starter had a significant role in his teams’ success. If the average starter has a 33 game starts in a season then if he wins 42.5% of his games he’d win 14 times. I think that’s a good amount of wins for a starter and a reasonable cutoff line.
Since this statistic measures whether a starter pitched well enough to make it to the playoffs, I propose calling it a Playoff Caliber Start. It should be awarded to a starter when he throws enough innings and allows few enough total runs for his team to historically win 56% of its games while the pitcher would win 42.5% of his games. Now all we need to do is look at the data to see when these conditions are fulfilled. The data for team wins is below. The y-axis is innings pitched and the x-axis is runs allowed.
Teams win 56% or more of their games when the starting pitcher either throws at least five innings while allowing two or fewer runs or throws nine innings while allowing fewer than four runs.
The data does indicate that when a starting pitcher throws nine innings and allows five runs his team wins 56% of the time. However, there were only nine cases of this happening from 2000-2014. Given that this happens with such low frequency I don’t consider such a scenario a Playoff Caliber start.
When a starter throws between seven and eight innings while allowing three runs his team only wins 51.7% of the time. When a starter throws between eight and nine innings while allowing three runs his team only wins 53.2% of the time. When a starter throws between seven and eight innings while allowing four runs his team wins 38.8% of the time and when throwing between eight and nine innings while allowing four runs his teams win 39.5% of the time. None of these scenarios reach the 56% cut-off and therefore shouldn’t be considered a Playoff Caliber start.
Next is the data for pitcher wins. The y-axis is innings pitched and the x-axis is runs allowed.
I didn’t include scenarios that resulted in a team win percentage of less than 56% because they aren't eligible to be Playoff Caliber starts. Tthe only scenario where a starter would win fewer than 42.5% of his starts while his team would win 56% of his starts is when he throws fewer than six innings and allows two runs. Starters that threw five innings and allowed two runs won 473 games, lost 223 games and had 553 no decisions. Over a thirty-three game season this would result in a 12.5-5.9 season which is reasonably good. The fact that such games are more likely to result in a no-decision than a win indicates that such a start shouldn’t be considered a Playoff Caliber start.
The guidelines for a quality start are a simple and easy way to determine whether a pitcher gave his team a chance to win. However, there are certain cases where they are inaccurate and could benefit from modifications.
The guidelines I propose indicate that a start should be a considered a Playoff Caliber start when the starter successfully does the following:
- Throws at least five innings while allowing zero or one runs
- Throws at least six innings while allowing two runs
- Throws nine innings while allowing three or four runs
When these conditions are met there is at least a 56% chance that the team will win and a 42.5% chance that the pitcher will win. If a pitcher only throws Playoff Caliber starts that meet these guidelines he will have a strong season and will put his team in position to reach the playoffs.