Game Scores for Pitchers

[viper]

In the never-ending battle for truth, justice and the American way, I ask about the daily GSC for pitchers as shown in box scores.

The formula is:

GSC = 50 + 1*Outs + 2*IP after the 4th + 1*K - 2*H - 4*ER - 2*UnearnedRuns - 1*BB

I ask the top intellectual minds of the Mastersball staff to tell me just how indicative of a pitchers performance this number is. I really have no clue as to what is considered good, bad or indifferent. I do know that Braden's perfect game scored a 93 as did a one-hitter by Johnny Cueto. The Jason Marquis zero-out, 7-ER, 4-hits and a walk debacle scored a 13 [realize you get 50 for being able to make your way to the mound].

[AllstonRockCity]

Interesting. I have always wondered how that was calculated.

That means that Pedro's CG 1H 1ER (chili davis solo shot) 17K performance at Yankee Stadium scored a 98, 5 points better than Braden's perfect game. That seems to be proof that this formula is inherently flawed. That game was a masterpiece by anyone's definition, but a perfect game is instant history.

[viper]

but 17 Ks is impressive.

Probably any score over 80 is out of sight. To date there have only been 33 scores of 80 or more. The total number of starts is around 1000 [30 teams at 33+ games each]. I would like to see a complete list of all starts to see how numbers are in various groupings.

I am of the belief [gut not facts]that pitchers go in spurts - both good and bad. They get into grooves like hitters. You need to ride the hot streaks and get off the train when the cold begins.

[Todd Zola]

Interesting. I have always wondered how that was calculated.

That means that Pedro's CG 1H 1ER (chili davis solo shot) 17K performance at Yankee Stadium scored a 98, 5 points better than Braden's perfect game. That seems to be proof that this formula is inherently flawed. That game was a masterpiece by anyone's definition, but a perfect game is instant history.

Sorry, but from a sabermetric viewpoint, I completely disagree with this viewpoint.

The real perfect game is 27 strikeouts. Pedro's performance was considerably closer to perfect than Braden's.

Braden fanned 6. He got 4 outs on liners, 7 on grounders and 10 on fly balls.

From a strict probability angle, that should have resulted in 9.5 hits and 1.5 homers.

Now consider that he would have needed 9.5 MORE OUTS

Yes, a perfect game is historic. It is beautiful. It is also extremely lucky.

EDIT I just realized with the way I have my Excel tool set up to do this calculation, Braden would not have needed 9.5 outs. What the data says is if he went he whole game and got 27 outs with his GB/LD/FB in the same proportion, he would have allowed 9.5 hits and 1.5 HR, sorry for the original misstatement. The larger point still holds true.

[Guest]

I think to answer the question you have to start with a simple question:

Do you want the game score to reflect simply what the pitcher did in that game or to reflect the underlying performance?

To follow up on what Todd says, if a pitcher throws a no-no with 0 K and 8 BB, he might score well in that metric but it isn't necessarily a performance that reflects repeatability, if that makes sense.

So it depends what you want to measure.

[viper]

I guess my post was two-fold in purpose.

First, I just wanted to know what you all thought about this statistical computation relative to how it reflected a pitchers performance.

additionally ..
In three of my six league, you start every rostered pitcher. It is too late to undo my draft so I have to live with my selections. These leagues have a fairly high 1100 minimum inning requirement which precludes too much of a LIMA approach. You need at least six SPs to get to 1100 especially given injuries. In two of those three leagues IP replaces Ks so SPs take on an even greater relevance.

In my other three league, you have bench so you get do some mixing & matching. For these league you need some way to make your start/bench decisions. I try to start my "hot" pitchers and bench my "cold" one. I want some way to determine hot and cold. The only real information I have is the box score. Maybe it is impossible from this but it is all I have. I also try to avoid going against the top hitting teams.

I don't even know what factors are the most important during the season. Plus not many places provide much help in this area. I can try and bully pitching but I'm looking for the better ways to manage pitching.

Maybe I am presenting this wrong and I should quit trying.

[added note] Back in the middle 80s I was a systems analyst for Marriott. I did computer system design for their Inflite division. I told their manager that I could program anything he could explain to me. But I told him that many things could not be explained. Maybe this is one of them.

[Guest]

I sure hope you don't quit trying.

As a decision-making tool I'd be much more interested in K/9, K/BB, and HR/9 than anything else you put out there in a calculation.

[viper]

for those three stats, on a game by game basis [or a rolling 3, 4 or 5 game basis] what are the good, middle and bad numbers?

Maybe 7, 2.5, 1

[Guest]

I like that - could go down to 6 for K, but that's the range.

Especially if you're seeing pitchers who have good K, KBB figures with tough HR numbers with matchups in good pitchers parks that can help.

[Todd Zola]

I see several issues.

1. There is too much noise over a small sample to use it effectively to gauge future performance, let alone gauge a single upcoming effort. And even if you filter out some of the noise, aka luck with respect to hit rate/HR/LOB rate and focus on skills, the difference between a 7K/2BB game and a 5K/4BB game may be one missed call by an ump and a 3-2 foul tip not caught followed by ball four. Of the pitcher may have been facing a team that swings at pitches out of the zone so he gets more K and fewer BB, but when throwing those same pitches against another team, they lay off.

2. Even if we could quantify the likelihood of success, it is still a human being taking the mound. He may not have slept well, his child could be sick, etc.

3. Become a fly on the wall in a room where they are setting Vegas odds and I bet you get more info than any fantasy baseball site can provide.

[viper]

so it's a hopeless task?

[Guest]

Why not merge the two thoughts?

You accept the inherent limitations to either approach and you basically stick with hot hands, aware of the fact that underlying skills over the long haul SHOULD win out.

No doubt in looking at P decisions I am looking at the underlying peripherals, which don't always work, but work best for me.

[Todd Zola]

Nothing is hopeless.

Well, being right 100% of the time is.

But being right 50.1% of the time is not.

Just lower your expectations

[viper]

I wish I had a way to create expectations

[Guest]

Help me follow here - is that you want an actual metric? Or are you looking for a more general "what should I be paying attention to?"

[AllstonRockCity]

Pedro's performance was considerably closer to perfect than Braden's.

I understand where you are coming from, but I disagree. Pedro could have lost that game, an impossibility for Braden.

A pitcher's job is to get outs. Yes, striking opposing batters out guarantees those outs and takes any 'luck' out of the equation. From a repeatability/predictability standpoint, how a pitcher gets those outs is important, but it seems silly to grade a guy on how he got those outs on a single game basis.

But that Pedro game I'm referencing wasn't even a shut out and I have a real hard time saying that it was a better performance than a perfecto. Obviously though, that Pedro gem is one of my fondest baseball memories. As soon as I saw that formula, it was the 1st and only game I wanted to plug into it.

I guess what I'm trying to say is, if we are talking about projecting performance how the outs are obtained is paramount. When grading a single performance how the outs were obtained is irrelevant(?) just as long as they are obtained.

[viper]

is both an answer?

I am a believer in the "Bully Hitting, Manage Pitching" mantra. During the season, if you did not draft hitting you are toast. If you favor hitters in the draft, by definition you slight pitching. You do the best you can but you cannot have your cake [hitters] and eat it too [pitchers]. I slightly changed my drafting of pitchers thoughts for my last several drafts by looking closer at K/W ratios. I made 2.5+ a major factor. These staffs are better.

During the season, I tend to keep my players longer than I should based on historical ability. Gil Meche was injured but once he got going, I felt his history was that he would be adequate. Even when his outings were bad, the question was "how bad?". I started to try and create some type of metric. I feel the MLB Quality Start is a joke. I don't really like Shandlers but it is probably something I am looking for. I'm trying to find a way to judge recent past performances. I then want to look for trends. I looked at the Game Score for the late substitute A's pitcher last night. That alone convinced me that this formula has minimal value - at least for pitchers who don't reach a certain inning count.

Not sure I answered your question.

Why does the song "To dream the impossible dream" come to mind?

[viper]

Harvey Haddux lost a 12 inning perfect game. Now there is something we will never see again - mostly because no pitcher will ever go 12 innings let alone start the 13th.

[Todd Zola]

I guess what I'm trying to say is, if we are talking about projecting performance how the outs are obtained is paramount. When grading a single performance how the outs were obtained is irrelevant(?) just as long as they are obtained.

I understand the sentiment it depends on what you are "scoring".

If someone takes a civil engineering multiple choice test with 100 questions with 4 answers each and knows 92 answers then guesses correctly on 2 as probability dictates, they get a 94.

If someone else takes the same test and knows 80 questions and guesses correctly on 16 of 20 questions, they get a 96.

The latter person's grade is recorded higher, it is on the books.

But if I were asked which student I would want designing a bridge I travel on every day, it would be the former.

[Guest]

When grading a single performance how the outs were obtained is irrelevant(?) just as long as they are obtained.

Then just look at IP and ER, since ultimately that's all that matters when it relates to winning games. Not sure it's the way to go to see how a pitcher is really doing, but it grades the start correctly.

Viper - you raise an interesting point, one which I am unsure I am equipped to answer very well, but in the interest of furthering the conversation, I will raise it...which is that we preach the theory of underlying peripherals...and long term it works, you're even seeing it with your teams. But in the short term, pitchers "convert" those peripherals to ERA/WHIP differently than on the long term (when there is reversion to the mean) - the idea you raise is whether or not that conversion represents some sort of hot/cold streak or whether it is random chance.

Every year I overhaul the projection system in some way. This has been my project for 2011. However, the idea of whether I can recommend in the short term there is some predictive ability to looking at this...I am open to it.