With the current winning streak (last night not withstanding), I've been thinking about the Twins' starting pitching and wondering if there was any correlation between how I intuitively saw them doing and their game scores.

Now, truth be told, I don't know if game scores are a useful statistic for measuring a starting pitcher's performance, but I find that I always check them every morning on ESPN's box score.

To find an explanation of what a game score is and how it is calculated, look here.
 Essentially a game score begins at 50 and adds points for positive things the pitcher does (outs, K's, finishing innings late in the game), and takes away points for mistakes made by the pitcher (hits, run, ER's, HR's, BB's). The resulting number is a measure of how dominant that particular start was for the pitcher. 50 is average.

Using data from Baseball-reference.com up through last night's game, what do game scores tell us about the Twins' starting pitching this year?

To begin with, I'm not going to include Liriano, as his three starts are (for this purpose) irrelevant. That leaves Livan, Boof, Baker, Blackburn, Slowey, and Perkins. My sense this season is that Livan has given us some ups and downs but has been relatively consistent. Boof, while at times pretty good, was ultimately pretty bad. Baker seems to me to be the closest thing we have to a staff ace. That leaves Blackburn, Slowey, and Perkins--the three young guns. They've all had growing pains, but it seems to me like they're showing more positives than negatives, with Slowey and Blackburn showing some strong potential.

So if these are my intuitive assessments, (feel free to disagree in the comments below), how do they compare to the game score data?

The top Twins' pitchers thus far this year, by average game score:

• Slowey--55, 13 starts
  
• Baker--54.8, 13 starts
  
• Blackburn--49.5, 17 starts
  
• Perkins--47.1, 12 starts
  
• Boof--43.4, 12 starts
  
• Livan--42.4, 19 starts
  
• (Franchise--31.7, 3 starts)

Slowey and Baker are the top two pitchers, which seems right to me. Blackburn is up there as well. The most surprising discovery is that Boof has actually been marginally better/more dominant than Livan this year. So what happens if we look at their 20 game splits throughout the year?

Slowey

• Games 1-20: 42, 1 start
• 21-40: 50.5, 2 starts
• 41-60: 57.8, 4 starts
  
• 61-80: 51.5, 4 starts
  
• 80-100: 67.5, 2 starts

Baker

•    1-20: 56.8, 4 starts
• 21-40: 46, 2 starts
• 41-60: 51, 1 start
• 61-80: 54.8, 4 starts
• 81-100: 61.5, 2 starts

Blackburn

•    1-20: 54.3, 4 starts
• 21-40: 42.3, 4 starts
• 41-60: 53.3, 4 start
• 61-80: 42.3, 4 starts
• 81-100: 74, 1 starts

Perkins

•    1-20: 0, 0 starts
• 21-40: 51.5, 2 starts
• 41-60: 40.8, 4 start
• 61-80: 47.3, 4 starts
• 81-100: 55, 2 starts

Boof

•    1-20: 46.8, 4 starts
• 21-40: 45.8, 5 starts
• 41-60: 35, 3 start
• 61-80: 0, 0 starts
• 81-100: 0, 0 starts

Livan

•    1-20: 52.4, 5 starts
• 21-40: 44.8, 4 starts
• 41-60: 30.3, 4 start
• 61-80: 36.5, 4 starts
• 81-100: 48.5, 2 starts

What does this tell us? Slowey and Baker have been pretty good all year, and are trending upward. Blackburn appears to alternate between being pretty good and being a little inconsistent, which makes sense for a rookie. All together he's been pretty good as well. Perkins, at this point, has started a little slow but is also trending upwards. It remains to be seen whether his ceiling is as high as Slowey's or Baker's, but at the worst he's looking like a great end of the rotation guy.

The most interesting discovery is that, for the most part, Boof and Livan are mirror images of each other. Both started strong through the first 40 games of the season, and then both imploded. Boof, as we know, got demoted to the bullpen, but their game scores suggest that Livan just as easily could have been sent down. In fact, through the second 40 games of the season Livan's average game scores were down by 15. It is too soon to tell if he's bringing those numbers back up, or is just the lucky recipient of above average run support. In either case, Livan is far from our "ace" using this measure of dominance--in fact, he is our #5 starter.

In order to develop a more comprehensive ranking assessment for our starting pitchers, I ran one last piece of analysis on these numbers, calculating the standard deviation for each pitcher. The standard deviation reveals the spread of numbers within a given sample, and is essentially a measure of consistency. The lower the standard deviation, the more consistent the numbers in the sample are, (or that a pitcher is more consistently performing at or close to their average level). Standard deviations are in the same units as the sample, allowing easy and accurate comparison. In this case, they are unit less because game scores have no units.

Using their standard deviations, (where lower is better), here are the Twins' most consistent pitchers:

• Baker, 9.8
• Perkins, 12.7
• Blackburn, 16
• Boof, 17.9
• Livan, 18.4
• Slowey, 19.7

Baker is clearly more consistent than the rest of the staff. The only noteworthy observations here is that Perkins is also pretty consistent, just at a lower performance level than Baker. Livan, Boof, and Slowey are on the other extreme. Their substantial standard deviations suggest a large fluctuation in their start-to-start performances. Slowey's standard deviation--more than double Baker's--is the most surprising to me considering how dominant he seems to have been this year.

Given this I decided to run the numbers again, this time taking out the highest and lowest game score for each pitcher. These modified standard deviations rank the pitchers accordingly:

• Baker, 6.2
• Perkins, 9.3
• Slowey, 12.1
• Blackburn, 13.4
• Boof, 14.4
• Livan, 16.4

With their best and worst games taken into account, Slowey jumps to the the third most consistent pitcher. This makes sense when you consider that his best game score was an 89, but his worst was a 10. Baker and Perkins continue to be mirrors of each other, just at different levels of performance. Perkins' best and worst games scores are eerily similar to Baker's, just lower (63 and 20 versus 72 and 33 for Baker). Boof and Livan improve, but are still pretty bad.

Looking at the differences between the regular standard deviation and the modified standard deviation gives us a shorthand measure of the consistency of the consistency. While certainly open to interpretation, I think it is really interesting to look at this number in conjunction to the Twins' pitchers, because seems to tell in a nutshell what each pitcher is about. I could be completely wrong about all this, but see what you think:

• Livan, diff. = 2
• Blackburn, diff. = 2.6
• Perkins, diff. = 3.4
• Boof, diff. = 3.5
• Baker, diff. = 3.6
• Slowey, diff. = 7.6

Most of the staff receives the same adjustment, with the major exceptions of Livan and Slowey. Livan has been consistently inconsistent, with lots of good games and lots of bad ones, in relatively equal proportions. Slowey, on the other hand, has had few games extremely divergent from his average. The primary reason for his 7.6 difference is that he has pitched far more games above his average score than bad ones below it. The bad ones he did pitch happened to be pretty awful, which tilted the scores accordingly.

So, in summation, I think each pitcher's game scores correlate pretty well with the overall picture we have of them this year. Factoring in their standard deviations as a measure of their consistency, I'd rank the starting pitchers this way.

1. Scott Baker, 54.8 ave. GS, 6.2 modified SD (consistency rating)--consistently good and trending upwards
   
2. Kevin Slowey, 55 ave. GS, 12.1 SD--most dominant of our pitchers, just needs to continue gaining experience. Greatest potential to become a true ace, even without a supposed "out" pitch.
3. Nick Blackburn, 49.5 ave. GS, 13.4 SD--Pretty solid thus far, with a slightly higher demonstrated ceiling than...
4. Glen Perkins, 47.1 ave. GS, 9.3 SD--Perkins is very consistent, but simply hasn't shown to this point that he can be a truly dominant starting pitcher. His best game score of the year is the lowest of all the starters, but the difference between his highest and lowest games scores is second best behind Baker (43 to Baker's 39. Slowey has the greatest discrepancy at 79. Livan, Boof, and Blackburn are at 59, 57, and 56 respectively). Perkins is what Livan should be--a solid innings eater that you know exactly what you're going to get from. A poor man's Scott Baker.
5. Tie--Boof and Livan. Both have been wildly inconsistent and bad this year. If I had to call it, I'd give the edge to Livan because I do think he's had a positive influence on the younger guys. Livan, in many ways, in the anti-Santana--terrible but great record due to terrific run support.
   

So what do y'all think?

PS--Just for kicks, here's Matt Garza's line for this year:

• 53.4 average GS, 15 modified SD (which compares him most closely to Slowey).