People are notoriously bad at recollecting things they hear and see, especially when they happen a long time ago. For example, the Innocence Project reports eyewitness misidentification occurs in approximately 75% of convictions that are overturned.
I submit that people are also poor at understanding the things that they do, even if they are experts at it and did it successfully for a long time. Professional golfers were once asked whether the orientation of the club face on a golf club or the direction of the swing was the primary determinant of the initial direction of the ball. In other words, if you swing to the right, but your clubface is pointed to the left at impact, which way does the ball start out? I forgot the numbers, but a significant percentage of PGA golfers answered incorrectly. If you care, it is mostly the angle of the clubface which determines the initial direction of the ball.
Tonight in the Braves game, in an AB by McCann with Kershaw pitching, Ron Darling, an excellent pitcher during his career and a Yale graduate to boot, remarked when the count went to 2-2 and Kershaw had thrown several fastballs, “He is surely going to throw the curve ball (or slider) now.” On its face, that is an absurd statement. If Darling knows that to any degree, then surely so does the batter, who happens to be a catcher! So that can’t possibly be correct! Of course, Kershaw threw another fastball. Darling immediately said, “Well, he decided to go with one more fastball and then the off-speed pitch.” Are you kidding me? Same shizit, different day. If Darling is that certain now, then surely so is McCann.
All pitchers, and especially great ones like Kershaw, randomize their pitch selections precisely so the batter cannot figure out what is coming with any certainty. Now, if in a certain count and certain situation, a certain pitcher throws 80% fastballs, then obviously the batter can “look” fastball and be right 80% of the time. But, still he does not know any more than he has an 80% chance of getting a fastball. He does not know, or should not be able to deduce, anything other than that 80/20 split based on the previous pitch or pitches. That is what it means to randomize your pitches. That you cannot tell what is coming based on prior pitches.
The concept of “set-up pitches” is largely a fallacy other than the fact that they may change the percentages. For example, if (and that is a big IF) throwing a high inside fastball actually makes a breaking pitch more effective on the next pitch, even if the batter knows that it is more likely to be coming, then you might throw 30% breaking pitches whereas if the previous pitch were a low and away fast ball or another breaking pitch, then maybe you would throw only 20% breaking balls. Let me put it another way. A pitcher throws a high inside fastball. Now the count is 2-2 and the pitcher normally throws 50% off-speed and 50% fastballs at a 2-2 count with this batter and this exact situation. Are we to believe that he can throw 60% or 70% curve balls now, and yet if the last pitch were something else, he would throw 30% or 40% curve balls? If that were the case, then the batter would now know which way the pitcher was leaning based on the last pitch. That can’t be correct unless somehow the curve ball is more effective after a high inside fastball than it is after another pitch, even at the same frequency. That might be the case (I am not saying that it isn’t), but the batter can surely neutralize that by simply forgetting about the last pitch. Plus, again if the pitcher now throws the curve ball more often, the batter has the luxury of knowing that and he can now look for the curve ball, thus making it less effective, presumably.
I hope that was clear, because it is a very important point.
Anyway, the point is that Darling, as a successful pitcher, clearly randomized his pitches in all situations, as do all pitchers. All he can tell you, as an ex-pitcher, are the percentages at any given point. He cannot tell what a pitcher is going to throw with any certainty unless those percentages reflect that.
And more importantly, those percentages should almost never be predicated on what was previously been thrown – only the count, batter, game situations, etc. If those percentages are predicated on previous pitches, then the batter can more easily figure out what you are going to throw AND those percentages will become sub-optimal (again, with the caveat that it might be true that a certain pitch makes it harder or easier to hit a certain subsequent pitch even if the batter knows that, which he surely does). In other words, if a pitch is predicated on previous pitches, for example, if you throw 5 fastballs in a row, you are more likely to throw an off-speed pitch, even at the same count, then you are not randomizing your pitches. That IS the definition of randomization. Darling should know that, but somehow when words come out of his mouth, he doesn’t.
That is why when you think you are getting good analysis from ex-players, because they are ex-players, you often are not.