You are on page 1of 4

Dan Stauffer

Ms. Mckelvie
Statistical Reasoning in Sports
16 June 2015
Brad Lidges 2008 Season
Brad Lidges 2008 season was one of the most impressive seasons for a closer in
the history of baseball. 41 saves in 41 chances, a 1.95 ERA, and a 2-0 record are some of
the impressive fruits of Lidges arms labor. While two future Phillies in Roy Halladay
and Cliff Lee sat atop the American League Cy Young Award leaderboard, Lidge was the
only reliever on the board in the National League, coming in at fourth in the voting.
Perhaps one of the reasons for Lidges impressive year was his heightened performance
in high-leverage situations. Without runners on base, Lidge allowed an opponent batting
average of .208, and with runners on base, Lidge allowed only a .179 opponent batting
average (all according to Fangraphs.com). To see whether Lidges ability actually
increased in these high-leverage situations, we need to first state the two basic options
(a.k.a. the null and alternative hypothesis):

H0- Brad Lidges ability was the same both with and without runners on base.
Ha- Brad Lidges ability was better with runners on base than without runners on
base.

Next, well need to gather our data and establish some things. In this case, well
be using 0.05 as our significance level, as thats the typical number used in cases like this.
Well also need to present Lidges statistics from 2008, which can be done clearly and
easily in a two-way table:

No Runners on Base Runners on Base Total


Outs Gotten 114 92 242
Hits Allowed 30 20 50
Total Batters Faced 144 112 256
100%
90%
80%
70%
60%
50%
40%
30%
Percentage
20% Hit
10% Out
0%

Circumstance
And to allow us to
visualize the data even more easily, a segmented bar graph:

Although the difference in Lidges performance may seem small, in the context of
baseball and batting averages, the difference between .208 and .179 is huge. It does seem
that Brad Lidges ability increased when there were runners on base, as compared to
when he pitched without runners on base. To look at this more in depth, well perform
some simulations, using a significance level of 0.05. The test statistic in this case can be
found by taking (the number of outs gotten without runners on base over the total number
of batters faces without runners on base) minus (the number of outs gotten with runners
on base over the total number of batters faced with runners on base); thus, we get
(114/144)-(92/112)=-.0298.
To simulate the distribution of the test statistic with note cards, we would make
242 note cards labeled out and 50 labeled hit. Then, we would make one pile with
144 cards, symbolizing Lidges performance without runners on base, and one pile with
112 cards, symbolizing Lidges performance with runners on base. We would then
calculate the percentage of hits in each pile, subtract the second piles percentage from
the first piles, mark that value on a dot plot, and repeat. Using an online applet to
simulate 100 trials,
this dotplot
represents the results
of the simulation:

With 100 trials performed, 23 dots fall below the specified test statistic of -.0298;
thus, the p-value in this case is 23%. Assuming that Lidges ability is the same with and
without runners on base, there is a 23% chance that he would have a difference of
opponent batting average the same or larger than the one he carried during the 2008
season. Because the p-value of 23% is significantly larger than the previously
determined significance level of 5%, we fail to reject the null hypothesis. We do not have
convincing evidence that Lidges ability is greater with runners on base than without
runners on base.
This simulation wasnt very flawed. The p-value was far enough removed from
the significance level that we can confidently say Lidges performance was likely not due
to a change in performance. The only real issues that can be taken with the simulation
are the sample size used. It typically takes a pitcher about 630 batters faced for his
batting average against statistic to stabilize. Lidge only faced 144 batters without runners
on base and 112 batters with runners on base. To make this study more accurate and
solid, we would need a larger sample size to confirm Lidges batting average against in
the two situations we looked at.

You might also like