Re(de)fining The Quality Start
In a recent Dugout Central article, John Bowen asked if the quality start is a bad statistic. A lively discussion ensued, and the general consensus was that it rewarded mediocre pitching, at least for those barely meeting the requirements. In this article I present data on how often a team wins given how long the starting pitcher lasts and how many earned runs he allows, so that people can judge for themselves what the criteria for a quality start should be. I also give my own opinion, and suggest two possible redefinitions of the quality start.
Using data from the last 5 years (2006-2010), taken from baseball-reference.com, the following table lists the winning percentage of a team if the starting pitcher goes at least the number of IP shown (decimals refer to thirds of innings) and he gives up no more than the number of ER shown.
When a pitcher hurls a quality start (at least 6 IP and no more than 3 ER), his team wins 67.8% of the time. Other things we learn, for example, are that if a pitcher lasts at least 5 innings, on average his team wins 56.1% of the time, and if he lasts 5 innings and allows no earned runs his team wins 85.2% of the time.
But the cumulative totals don’t tell the whole story, since the 6 IP/3 ER qualification includes many performances much better than that (for example, a 9-inning shutout is also a quality start). To really determine if a particular start is worthy of being called “quality,” we must look at how the team did for exactly that number of IP and ER, shown in the following table. (Note that the numbers vary quite a bit for 0 ER, 5 or more ER, and more than 8 IP due to low statistics for those cases.)
Here we see that the minimum quality start, 3 ER in 6 IP, actually is a losing proposition for the team, which wins only 48.3% of the time with that performance by its starting pitcher. If you think a quality start should enhance a team’s chances beyond a coin flip, then 6 IP/3 ER is definitely not good enough.
It’s interesting that pitching 5 1/3 or 5 2/3 innings and allowing 3 ER is actually better for the team than pitching 6 innings with 3 ER. A similar pattern can be seen elsewhere in the table. I think this is because teams that are behind tend to pinch hit for the pitcher, which means that the starter will have an even number of IP and the team will tend lose more often. Teams that are ahead tend to replace their pitcher mid-inning, to preserve the lead, and might tend to win more often.
Even extending the quality start qualifier to at least 7 IP still doesn’t assure even a 50% chance of winning (although 6 2/3 IP does, due to the fractional inning effect mentioned above). If you say that a quality start should give the team a 60% chance of winning (roughly three out of five), a pitcher allowing 3 ER should go at least 7 1/3 innings.
If you require a 67% probability of winning (about two out of three), then you would need 8 1/3 inning with 3 ER, maybe 6 2/3 IP for 2 ER, and 5 IP for less than 2 ER — fewer than 5 IP with less than 2 ER can also give the team a good chance to win, but a pitcher should at least qualify for a win to also earn a quality start.
It doesn’t seem feasible to have the same maximum number of ER for all IP in a new quality start requirement – at least if you want a quality start to have a uniform benefit to the team. If you want something that’s easy to remember, you could require that the number of IP would have to be at least 5 more than the number of ER, with a maximum of 3 ER. That would mean a quality start would give a team an 80% chance to win on average, and about a 70% chance if a pitcher meets the minimum requirement. The last two sets of criteria mentioned above are listed in the following table. On average, both give the team about a 75% chance of winning.
|Criteria 1||Criteria 2|
|0 ER||>= 5 IP||>= 5 IP|
|1 ER||>= 5 IP||>= 6 IP|
|2 ER||>= 6 2/3 IP||>= 7 IP|
|3 ER||>= 8 1/3 IP||>= 8 IP|
Under Criteria 1 there would have been 8552 quality starts in the years 2006-2010; for Criteria 2 there would have been 7893. This is compared to 11,874 standard quality starts and 24,296 total starts in those years.
For me, either one of these definitions for a quality start is much better than the current one. What do you think?