IGT 4/18/08 Santana against Hamels

[Guest Triple Dee]

I tried to edit my previous post, but it opened a new post. Ignore this.

[batmagadanleadoff]

Triple Dee wrote: I can't quite figure out why you find the idea that the Phillies may have figured Heilman out, and consequently he should be used sparingly against them, such a difficult proposition to come to terms with.

I'll bite. Because it's an idea. A theory. That's all it is. Because you don't have enough evidence to take your idea beyond the idea stage. Because 36 innings is nothing more than a tiny blip in the grand scheme of things if you want to use those 36 innings to evaluate Heilman's effectiveness against the Phils. And it wouldn't matter if those 36 innings are the most Heilman's pitched against any one team. It's still only 36 innings and 36 innings isn't enough to prove whether or not Heilman's struggles against Philly is anything more than random coincidence. After all, he's gotta have his worst stats against some team.

This is not to say that under other circumstances, 36 innings wouldn't be enough to make some concrete judgement about some pitching phenomena or other. If a pitcher throws 36 innings and allows a HR in every single one of those 36 innings, I think it's safe to say that the pitcher is susceptible to the long ball and needs to make whatever corrections are necessary if that pitcher wants to continue to pitch in the major leagues. Or ordinarily, one wouldn't be able to gather much meaningful information from flipping a coin exactly 36 times. But if I were to flip a coin 36 times and get heads every single time I flipped that coin, my instincts would tell me that the coin is rigged even though it's theoretically possible though unlikely to get 36 straight heads from flipping a fair coin.

But you don't have this extreme outlier sort of data that would give greater meaning to a small sample size, so your 36 innings (Heilman's 36 innings) aren't enough yet.

Now it might one day emerge that Heilman's struggles against Philly are not random. Perhaps Heilman is tipping off his pitches is such a way that only Philly is noticing. Or maybe the Phils cheat whenever Heilman pitches. But all that would mean, is that you had a hunch, and you followed your hunch and went with it, and it turned out that you were right. But you still need more proof. You have a theory and you are entitled to place as much faith in your theory as you wish. But faith is not evidence.

[Guest John Cougar Lunchbucket]

Triple Dee wrote: I can't quite figure out why you find the idea that the Phillies may have figured Heilman out, and consequently he should be used sparingly against them, such a difficult proposition to come to terms with. You're either disagreeing because you don't believe it's true (even though the statistical evidence supports the opposite conclusion), or you're disagreeing for the sake of disagreeing.

No, I'm disagreeing because your argument is flawed: It's entirely too casual a conclusion that doesn't illustrate any cause-and-effect, and advocates a response that won't solve the problem it purports to point out.

Let's assume for a minute that you're right -- Heilman's stretch of ineffectiveness in fact proves that the Phillies have "figured him out." Why must we go on to assume that Heilman can't go back and "figure out" what the Phillies have figured out? That's what effective pitching is; and Heilman is an effective pitcher, as supported by the preponderence of data except which you've handpicked to make this argument.

You also have to acknowledge that buying into that idea requires swallowing a bunch of other ridiculous notions -- such as the idea that if the Phillies had truly figured Heilman out, then their failure the other night with the game on the line must have been spectacular. If you're Manuel, do you pinch-hit the next time Jenkins and Werth need to oppose him? How come?

The argument also doesn't solve any problems, but to replace Heilman's innings with the everpopular "someone else" as if that someone is impervious to being figured out themselves, necessarily any better than Heilman, rested/warmed, etc etc.

Mostly though I disagree with the general idea of managing by small subsets of outlying nonpredictive data. If Heilman is too ineffective to trust against the Phillies, then he's too ineffective to trust against anyone.

[Guest Triple Dee]

batmagadanleadoff wrote: Because 36 innings is nothing more than a tiny blip in the grand scheme of things

And this is where you're wrong, Grasshopper. Statisticians consider a sample space of 30 as the minimum for achieving a normal distribution. It's called the Central Limit Theorem.

[batmagadanleadoff]

Triple Dee wrote: And this is where you're wrong, Grasshopper. Statisticians consider a sample space of 30 as the minimum for achieving a normal distribution. It's called the Central Limit Theorem.

I think this is absurd. I've been reading about baseball analysis through statistics for about 30 years and I've never heard of anyone making the kind of judgements you're making about Heilman based on 30 innings. According to you, it would therefore be statistically meaningful data when a batter gets hot and bats .500 over a stretch of 30 at-bats. I bet this happens a few times every single season. According to you, that run is supposed to be a strong indicator that the batter is gonna bat .500 over the rest of the season.

Would it make any difference to you to consider that the entire population of major league caliber baseball players is, in and of itself, not random at all, but culled from the extreme right side of the curve... so that when I go on and on about how Rey Ordonez sucked, it's still understood here that in the grand scheme of things and when compared to the population as a whole, he was still one of the very best baseball players in the world?

[Guest AG/DC]

Well, I disagree wth him also, but 30 innings /= 30 at-bats.

[Benjamin Grimm]

The number "30" can't be significant on its own, independent of the unit of measure.

Do we need 30 innings? 30 at bats? 30 pitches? 30 games? 30 years?

[batmagadanleadoff]

AG/DC wrote: Well, I disagree wth him also, but 30 innings /= 30 at-bats.

My understanding is that 30 refers to the unit. And that any unit is acceptable, assuming that the theory applies in the first place. It's not necessarily the unit of measure, but also what you are trying to measure with what particular unit. But sure, there are necessarily more batter/pitcher confrontations in an entire inning rather than in a singular at-bat, which is obviously a component of the inning itself. But this is why, given what we know about baseball, we all know on an intuitive level that 30 at bats or 30 innings pitched isn't concrete evidence for a whole lot of things, but might constitute proof of some things.

[Guest AG/DC]

Well, there's some theory out there that performance over 150 at-bats (or is it plate appearances?) is an indication that what a batter is doing is f'real.

[Guest Triple Dee]

John Cougar Lunchbucket wrote: Let's assume for a minute that you're right -- Heilman's stretch of ineffectiveness in fact proves that the Phillies have "figured him out." Why must we go on to assume that Heilman can't go back and "figure out" what the Phillies have figured out? That's what effective pitching is; and Heilman is an effective pitcher, as supported by the preponderence of data except which you've handpicked to make this argument.

Look I agree with you on one point; If the Phillies have figured Heilman out, the onus is on him to re-adjust. But why do you say the data is "handpicked"? A better argument would be that the data is too broad, like Namor suggested.

John Cougar Lunchbucket wrote: You also have to acknowledge that buying into that idea requires swallowing a bunch of other ridiculous notions -- such as the idea that if the Phillies had truly figured Heilman out, then their failure the other night with the game on the line must have been spectacular.

Not really. The odds are always stacked in the pitcher's favor. I accept that Heilman's record against the Phillies was one of a multitude of factors that could have influenced the outcome. But I cannot agree that it is not probative.

John Cougar Lunchbucket wrote: Mostly though I disagree with the general idea of managing by small subsets of outlying nonpredictive data.

Fair enough and you are entitled to that. Unfortunately, this is the way baseball is now played. Baseball managers/analysts use even smaller subsets of data. But I find it somewhat confusing that people criticize Willie for managing on a hunch, but find it acceptable when he ignores things like this .

[batmagadanleadoff]

AG/DC wrote: Well, there's some theory out there that performance over 150 at-bats (or is it plate appearances?) is an indication that what a batter is doing is f'real.

Yeah. Even 30 at bats might be a meaningful # of at bats under the right circumstances. It's also dependent on what you're trying to figure out. A rookie might come up and hit 4 or 5 hr's in his first 30 at bats and just about blister the ball in almost every other non Hr at bat. Now these 30 ab may not be proof enough that the rookie's going to be setting hr records, but those 30 ab's are more meaningful when trying to gauge if the rookie is capable of hitting major league pitching.

[Guest Triple Dee]

Benjamin Grimm wrote: The number "30" can't be significant on its own, independent of the unit of measure. Do we need 30 innings? 30 at bats? 30 pitches? 30 games? 30 years?

The ERA statistic measures earned runs according to innings.

Not to bore you with the finer details of the Central Limit Theorem, but the result of the proof actually mandated a larger sample space. But apparently from empirical evidence, it was shown that that 30 was sufficient.

[batmagadanleadoff]

Let me follow up here some more because some of these formulas are not as formulaic as they've been made out to be. It's not good enough to simply say that 36 innings pitched is good enough to decide this. It's not just the unit of measure that needs to be considered, but also what exactly it is that you're setting out to establish, and also what kind of level of comfort do you want to have with your final results - how reliable are your findings.

To illustrate, suppose you were given a coin and were charged with determining whether or not that coin was a fair and balanced coin; the kind of coin that has the same chance of coming up either heads or tails when flipped in the air. And so you take that coin and flip it 36 times, coincidentally the same number of innings pitched that Heilman has accumulated against the Phils. So you flip the coin 36 times and get 18 heads and 18 tails.

Well, what do these results indicate about the coin? I would say that the 36 flips are meaningful in proving some things and not meaningful for other things. For example, the results allow you to rule out the possibility without any doubt, that the coin might have been rigged in such a way that it would always come up heads on every single flip. Or tails on every single flip.

But could you use those results to rule out the possibility that the coin was rigged in such a way that on every flip, there is a 52% chance of getting heads and a 48% chance of getting tails. Obviously not. I wouldn't be comfortable asserting that the coin is truly balanced based on those 36 flips. Because it's not so statistically unreasonable to flip a 52/48 biased coin 36 times and get 18 heads and 18 tails.

Thirty six innings isn't enough data to prove that the Phillies have Heilman's number, especially given all the rest of Heilman's stats against the Phils.

[Guest Triple Dee]

Do you know what a normal distribution is?

[batmagadanleadoff]

Triple Dee wrote: Do you know what a normal distribution is?

What are you, The Riddler?

You know what? You win. I concede. Heilman sucks against the Phillies and it's not a coincidence. And as soon as Beltran's last 30 or 40 at-bats against any specific team go bad, I'm gonna put it in Willie's suggestion box that Beltran should never play against that team anymore.

[batmagadanleadoff]

Triple Dee wrote: Look I agree with [Lunchbucket] on one point; If the Phillies have figured Heilman out, the onus is on [Heilman] to re-adjust.

But how could Heilman ever readjust when you've conclusively established that the Phillies have his number? Didn't you just prove with his 36 innings that Heilman can't get the Phillies out effectively? So how is Heilman's performance supposed to be able to change? And if he readjusts, for, oh, I don't know, the next 36 innings or so, does this mean that Heilman then has the Phillies number? And that maybe Heilman should then pitch every single game that the Mets ever play against the Phillies instead of never? And then could the Phillies readjust to to the readjusting that Heilman did against them? And if all of that ever happens, which cluster of 36 innings pitched would, in your opinion, get to matter the most?

Once you acknowledge that Heilman must readjust or that he even can, doesn't this undermine your claim that Heilman shouldn't pitch against Philly? That you have allowed for a scenario where Heilman and the Phils might go back and forth, trading the upper hand against each other in a game of leapfrog, is proof of the random nature of the 36 innings you relied on. For now, those 36 innings only prove that Heilman sucked against the Phils over his last 36 IP against them.