Key point from the article:
The caveat with QoP is that while the numbers it generates are objective, there was subjectivity involved in its development. The weights assigned to the various factors were honed over a number of years, but there’s a central question as to whether a cutting fastball can truly be compared to a sharp curve or a well-placed change up. (The system also doesn’t allow for using one pitch to set up another, which is a key part of the art.)This goes to the heart of why Moneyball and the whole stats vs. scouts debate continues to reverberate every dugout and back office in baseball. It is a power struggle between whose data to use. The "subjectivity" of scouting versus the "objectivity" of statistics. I think we can see that the lines are not so much black and white as it is shades of grey regardless of which direction you go.
from the National Post:
Moneyball 2.0? New pitching stat — courtesy of a couple of guys from Edmonton — could help identify hidden talent | National Post:
Moneyball 2.0? New pitching stat — courtesy of a couple of guys from Edmonton — could help identify hidden talent
May 29, 2015 2:20 PM ET
TORONTO — It has been 12 years since Moneyball was published, and 13 years since the first playoff appearances of the Oakland A’s team that it documented. That is to say, on-base percentage isn’t sneaking up on anyone any longer.
The things that Billy Beane championed with the A’s — the value of OBP and slugging percentage when evaluating prospects, and a decreased reliance on traditional indicators such as speed and contact — have long since been accepted by enough by people in the game that the original Moneyball conceit has largely been neutralized. That development poses a challenge for teams trying to find a statistical edge to complement their scouting: The central tenet of the Beane way of thinking, identifying the market inefficiency and then exploiting it, demands that there is still something left to exploit.
A couple of guys from Edmonton think they have just the thing: pitch quantification. Here is Wayne Greiner, chief salesman for the metric they call Quality of Pitch, or QoP, with the bold statement: “We think QoP is eventually going to carry more weight than ERA.”
As I say: Bold. The statistic has its roots in the college baseball career of Jarvis Greiner, Wayne’s son, who pitched at Biola University in Southern California before an injury put an end to that. Working with one of his professors at Biola, Greiner set out to try to grade the quality of a pitch in a way that had never been done before. We know that a fastball that travels 96 miles per hour is better than one that travels 87 mph, and one that paints the corner of the plate is better than one that crosses its middle. The vast amount of data now provided by Major League Baseball’s PITCHf/x system can say how much a curveball breaks and a sinker sinks, and when it breaks. Quality of Pitch attempts to take all of that information and boil it down to a single number that says whether a pitch was good or bad. A perfect pitch rates a 10. Anything above 5.0 is considered above average. And the allure of that single, simple number is that it can be assessed on any pitch: fastball, curveball, slider, changeup. Every pitch is graded on five factors: velocity, location, amount of break, point of break, and rise out of the pitcher’s hand. A pitch that breaks late and is on the edge of the strike zone will score better than one with little movement and that misses the plate entirely, and other such things you can probably figure out for yourself.
The practical application for the metric — the market inefficiency that it could potentially exploit — is that it’s a more pure assessment of the things a pitcher can control while stripping out the things he cannot. Greiner explains the concept this way: a pitcher is consistently throwing well, but a batter manages to fight off an inside pitch and bloops a single. The next guy is walked on a borderline call. The pitcher is unfazed and starts the next at-bat with more quality pitches, but then he hangs a curve ball that is turned into a three-run homer. This is bad for all of his normal statistics, including earned-run average, but QoP would say it was actually a pretty good stretch. Conversely, a pitcher who is not making quality pitches but is bailed out by a handful of great defensive plays behind him would have his mediocre outing reflected in the QoP numbers, if not the traditional statistics.The tantalizing prospect of QoP is its potential ability to tell teams which pitchers are consistently throwing better than their top-line numbers indicate. Like the guys who were quietly posting high on-base percentages a decade ago, that is the hidden value that QoP could unlock. Greiner says the numbers from 2014 predicted that, for example, Minnesota’s Kyle Gibson pitched better than his numbers indicated. He was in the top ten with an average QoP of 5.46, but had a middling ERA of 4.47. This season, again throwing quality pitches, his ERA is 2.72. (One of the things about developing algorithms over a number of years is that the makers don’t want to share all the data just yet. Since Quality of Pitch was first presented at a sabermetrics conference two months ago, nine Major League teams have taken an interest in the data. “I think we can get all 30,” Greiner says.)
The information would also be of use to teams trying to assess their own pitchers, like an early-warning signal for when someone’s curveball, for example, suddenly becomes flat. Greiner even says that the numbers can foretell arm trouble: if a pitcher’s QoP metrics suddenly go squirrely over a number of appearances, there’s a good chance that he’s not feeling right.
THE CANADIAN PRESS/Nathan DenetteThe tantalizing prospect of QoP is its potential ability to tell teams which pitchers are consistently throwing better than their top-line numbers indicate.
The caveat with QoP is that while the numbers it generates are objective, there was subjectivity involved in its development. The weights assigned to the various factors were honed over a number of years, but there’s a central question as to whether a cutting fastball can truly be compared to a sharp curve or a well-placed change up. (The system also doesn’t allow for using one pitch to set up another, which is a key part of the art.)
But Greiner believes the models have been tuned enough to now generate reliable data, pitch after pitch. It’s not Moneyball 2.0 yet, but it might get there.
Postmedia News
'via Blog this'
No comments:
Post a Comment