Monday, January 4, 2010

Rox Talk - Anatomy of a Start

In the Anatomy of Every Start by Raphy on the baseball-reference blog he provides a table of data for every starter's pitch line. I hate tables so I graphed it. I did make one simplification so that I could graph it more easily. The simplification was that I performed a weighted average on the runs scored per third of an inning. In this way I got the graph below. The data shows the innings along the x-axis and the runs weighted average on the y-axis. Each circle then represents the number of starts. Of the 4,860 starts by MLB starters last year, 1,083 starters went exactly six innings! As you would expect the shorter the start the more runs on average were scored compared to the later innings. Interesting to see how average runs given up through 4 2/3 is pretty even and then nosedives to the 9th inning. My guess is this shows that when a pitcher has it, teams generally don't get to him and that when pitchers don't have it the early innings clearly show this. It is nice to see that 78% of the starts do last 5 innings or more. Also interesting to note that only 30% of the runs scored are done before the 5 inning.

As is my custom I look at the 2009 Rox starting pitching to see how they compared. The purple dots show the same information. Once again it is interesting to see the Rox data trend the same way after 6 innings. So in addition to the above information I also graphed innings pitched versus team winning percentage. Data suggests that for a team to have a 50/50 chance of winning it helps for the starter to go 6 innings. Purple data shows how the Rox compare.

And finally Raphy also supplied Game Score data. Once again I graphed it coming up with a win rate term which is simply wins divided by number of games. The graph has team win rate in blue and pitcher win rates in pink. Interesting to see that a Game Score of 50 will get the team to about a 50/50 chance of winning while a Game Score of 60 is required for the pitcher to get a 50/50 chance.

No comments: