What an excellent and well-thought-out metaphor...

Well, if a pitcher's strike-outs go way up, that could also be the result of facing dreadful batters that day; similarly if he walks a lot one day that could be the result of facing very patient hitters with good eyes who can foul off tough pitches... So walks and strike-outs aren't hitter-independent.

Still, I agree with you about Kennedy. I think he was a little better last night than usual (the balls in play were somewhat weaker than usual and of course, no HR), and a bit lucky, and he'd been a bit unlucky earlier in the season.

Kennedy's fielding independent ERA is 4.49

Formula: ( (13*HR + 2*(BB+HBP) - 2 * K) / IP ) + 3.20

Dice-K, who's walking more than his fair share is at 3.63

Great pitchers have low hit and walk rates, and high strikeouts rates. It's not cut and dry though. You can have a high walk rate if you also strike out more and get hit even less. You can get hit more if you don't walk many and can't get lots of strikeouts.

Thanks for the definition and explanation. However, my original point is still valid and you don't address it. You have only three metrics - BB, Ks and %strikes. There are many more aspects to a pitching performance than these. Location of pitch, accuracy (did he hit his spots), break of ball, change in speeds, etc. Clearly these aspects of pitching, if mastered, lead to strikes. But they also lead to poorly hit balls. Also, sometimes the defense works off the pitcher - maybe the Yanks and Kennedy were more in synch. Obviously, it is difficult if not impossible to measure these things but that doesn't mean they aren't relevant. Just because you can't access data on a skill-based factor doesn't mean it isn't a factor. Thus I think you misuse "luck" by never discussing the possibility that non-measurable aspects of pitching may be improved upon.

Phil, the argument against that is (loosely) covered by DIPS theory, whose main tenet is that pitchers have very little control over the outcomes of balls in play. The batting average on balls in play (BABIP) tends to vary from year to year, and over time almost all pitchers regress to the mean in this area. (The control that pitchers do exert is on slugging average on balls in play; this tends to correlate better for pitchers from year to year. Of course, this effect is much weaker than the correlation in HR allowed, and pitchers who allow lots of HR also allow lots of doubles, in general.)

There are a few exceptions, but much fewer than you'd think; most of those pitchers have excellent stuff and do well on the other metrics anyway.

Moving to a more abstract level, "luck" is a loaded word. Many people have the logic chain "luck = undeserving = bad", and it is often used in the "luck = non-repeatable" sense instead.

Until Kennedy consistently demonstrates success despite low K:BB and GB:FB ratios, there is no reason to think he's made an improvement in the non-measureable areas that is having an impact on his game.

Finally, "just because you can't measure it doesn't mean it isn't relevant" is a bogus argument for two reasons. First, the burden of proof is on the person making the statement; they have to show that it is relevant, not the other way around. Second, if we measure everything else and that accounts for 90% of the results, then the remaining unmeasurable aspects by definition are not as relevant. If the measurables account for 99%, then what we can't measure isn't relevant, because the impact it can have is so small that it will be dwarfed by the other measurable parameters.

Put anecdotally, a priest is watching a baseball game with a friend and they notice that one batter always crosses himself before each pitch. The friend turns to the priest and asks, "Father, does that help? Would it help more if you did it?"

The priest replies, "Not if he can't hit."

-----

Put more pithily still, God is on the side of the biggest battalions.

fascinating discussion. 2 thoughts in this realm i wanted to share:

1. to the point of the burden of proof being on those who raise the idea that things that have not yet been measured may be important factors (ie they have to show that the things that have not been measure are relevant), i think that's dangerous ground when you can't control for all variables (although academically and theoretically i completely understand it). in some cases this may be very very difficult. not to be too rumsfeldian, but we don't know what we don't know. so i would hesitate to say that, in baseball, the things we do measure account for 90% (or 99% or some other %) of the results. in pitching, certainly k:bb and gb:fb are highly correlated with results (ie runs allowed) and excellent predictors of how pitchers will fare in the future. however, as fans striving to understand the game on a much deeper level, we should stay open to the idea that other factors that have not been measured may have just as significant an effect. we need to keep asking interesting questions, like does location of pitch , or variation in speeds, or height of the infield grass, or is the pitcher's dad in the crowd, etc. have a high correlation with results (again, runs allowed). hey, it's long-standing baseball 'knowledge' that changing speeds and hitting locations are key. but has anyone ever actually measured that to figure out how key? is that the data we should be measuring, with k:bb and gb: fb being the result of location and speed? it's possible that the things we *know* with certainty today will change as we measure and learn more.

2. it's important to draw a distinction between how "well" one played (past tense) v. how one is likely to perform in the future. getting back to ian kennedy for a second, he performed very well in his last start. he kept runs off the board. i think as saber-heads we are too quick to dismiss performances that achieve the goal of each team (to score more than you all) if we don't like the full stat lines. when we say that justin morneau didn't deserve his mvp in 2006, well, we're probably right. but he made the scoreboard change 227 times that year (97 runs scored, 130 rbi). that accounted for than 25% of minnesota's runs. in every day baseball, that means something because at the end of each night, while we're looking at obp, ops+, whip, k:bb, etc., those guys are looking to see if they won or lost. morneau's full stat line (obp, ops, rc, etc) shows us that there were better players in 2006 than him, and that if you put those players in his situation they *probably* would have produced more runs than he. but, alas, they weren't there so we can never really know that. i think as stat-heads we need to give these guys more credit for doing the obvious things well (plating runs for hitters, avoiding runs for pitchers). we can do this while at the same time saying that we don't expect those counting stats to repeat themselves the next year.

Sabrata -

Thanks for the explanation - I like "unrepeatable" much better than luck. And we'll see how repeatable Kennedy can be.

I must admit to ignorance regarding DIPS - can't argue about that.

Re: burden of proof - it works both ways. Dave doesn't "understand" Kennedy's effectiveness (ie. it doesn't fit his theory of a good pitching performance) so it is luck. But that is BS. If a fact doesn't fit your theory then your theory needs work - you can't just say he's lucky. In science, we call this handwaving - that data that runs against my theory isn't really important. You are right in that if I want to actually convince anyone of my point I need to show my own data. I still think, and hopefully new technology and an unlazy person will help show, that other metrics like changing speeds, etc. are behind Ks and BBs and these will offer a little more nuanced insight into pitching.

"If a fact doesn't fit your theory then your theory needs work."
Yeah, that's curve fitting and while it might work in outliers it is stunningly ineffective in practice if the theory is valid.

I don't understand how you can call Kennedy lucky in a start when he allowed 8 baserunners in six innings while only allowing one run. Strikeouts to walks ratio is not the sole category to look at when trying to determine whether a pitcher is progressing. One can look at Kennedy getting out of a bases loaded jam with one out as progression.

Secondly, I really don't understand how you can define luck in baseball. The only way to look at luck in baseball is if a sure homerun hits a bird on its way out of the stadium and is caught on the warning track. That is luck. If you talk to any pitcher, a hard line drive out in a game is still an out. The only thing that one can take from a pitcher getting hard outs is that he may run into a team that get those hard hit balls to drop into the gap. But a win is a win is a win, especially when you go six innings, pitch yourself out of a jam, and allow only one run. I'm sure any team in the MLB will take that.