THE BOOK--Playing The Percentages In Baseball

One interesting variable- speed.  “HIT_BALL_FX_LW” on the exact same batted ball is not the same for every hitter.  A hard line drive in the gap from Carl Crawford is clearly more valuable than the same batted ball from Jim Thome or Adam Dunn.  JC Bradbury ran into this same problem with PrOPS in THT a few years back. 

Any ideas how to deal handle this? 

On the flip side, there’s an interesting situation for fielders as well.  One fielder could play a ball into a double, where as another fielder might hold the runner to a single.  Of course, this is also dependent on the runner’s speed and aggressiveness.  How much of the result is due to a speedy runner, and how much is due to a poor fielder?

I am not sure this is so much of a big deal for the fielder, because it will average out over any reasonable sample size and the effect isn’t that large to begin with. For the hitter I think we need some sort of speed score.

Another issue is LH batters who hit into a shift. Once we have a few years of this data, we can see whether those batters tend to fall short of what Peter’s formula would predict. If so, the shift is working.

Of course when David Ortiz and Jason Giambi underperform Peter’s formula, we don’t know whether it’s because of the shift or because they are slow runners. Also we have to remember that the shift may cause hitters to change their approach.

If I remember correctly reading Peter’s article the other day, there was a question about the speed of the batter/runner.

As is being discussed about hang time, can we use a stopwatch to measure the time for a batter to get to first base? Measure those times when the batter runs full speed through the base.

The immediate benefit would be to find how often the batter would be expected to beat the throw to first. We could get a good ranking of players, and over time, see how much speed to first players lose as they age.

Electronic timers for the run to first base would probably work better, at least for when they hit the bag.

If teams are concerned about players hustling to first, I’m surprised they haven’t implemented these.  Put the times up on the screen when they bat:  league or team avg speed to first, player’s avg speed to first, and the speed of the player’s last run to first.

Everyone that has mentioned that speed of the batter is an important skill that is a factor in whether a hit ball becomes a hit and what kind of a hit it becomes is correct.  There was a question at THT about this, and I had also considered it when I was writing the article.  As I said in the article and in response to the question at THT, there will definitely need to be fine tuning of the bins when we have a couple of years of Hit f/x data to ponder.

However, and I should have made this point more clear in the article, Skill Determined Batting Runs was not meant to measure ALL the batter’s offensive skills, just his skill in being able to hit the ball in the best way to make it a hit.  It gives a measure of that player’s true talent for that ONE particular skill.  To translate SDBR back into a projection of a player’s future performance you would need to add back in the effects of other factors including park effects, quality of the opposition pitching, unusual fielding positioning, and the speed and handedness of the batter.  Somewhere in my box of unpublished metrics I have a measure of speed skill that included all of a player’s offensive speed runs, stealing, taking the extra baseon other batter’s hit balls, AND his speed’s contribution in getting infield hits, bunt hits, and getting extra bases on his own hit balls.  Maybe its time to drag that metric out; Skill Dependent Running Runs.  I think Juan Pierre lead by a large margin. 

The best approach is to measure batter speed as a separate offensive skill.  The reason is that it is likely that his speed skill will age at a different rate than does his ball hitting skill, or his pitch recognition skills.  The only reason that speed should be incorporated in SDBR is if future data show that certain batters with speed skills actually change their approach to hitting, and instead of hitting balls as hard as they can at the vertical angle and horizontal direction that would maximize its linear weight for the type and location of the pitch, that they PURPOSELY hit it less hard to the left side of the infield in an attempt to achieve an infield single.  This may be true for Ichiro and a few other very fast left handed hitters and they may have to be projected with this special skill in mind, similar to the projection of knuckle ball pitchers.

While the data in hit f/x is ultimately the way to go in terms of being able to better project actual hitting results (singles, doubles, etc.) in the short term, and everything that Peter says in the article is correct (I think), I don’t think you would ever want to substitute this formula for a traditional one, even in the very short run (OK, maybe in the very, very short run).  The reasons are pretty much articulated in the article and in this thread so far, and are obvious.

In addition to the parameters of the batted ball, you need the position of the fielders, the speed of the batter, and park and even ambient effects (like weather).  And even then, you better be real accurate with your data.  If you are not, and some hitters are able to control where they hit the ball (which I don’t necessarily think they can) then you are in trouble.  Basically, I think a combination of the two ("real" results and the batted ball data) is the best way to go.

Without at least the speed of the batter and some measure of power for the batter (as a proxy for fielder positioning), I would not even think of using a formula like this to replace a traditional lwts one.

You could simply create more bins, or just break the existing bins into 3 separate ones, for slow, medium, and fast runners, and then again, 3 more times for low, medium, and high power.  And of course you need separate bins for LH and RH batters, although even then fielder positioning can vary a lot depending upon whether said batters tend to pull the ball or not.  The two biggest things you have to control or account for are speed of the batter and fielder positioning.

Then again, this is like a park or other “adjustment” where even with all its problems you can ALWAYS use it to tweak a traditional lwts formula.  If you did nothing else but adjust for those bloop and scratch hits and the line drive outs, you would be making an improvement.  Then again, you don’t necessarily need something as accurate as hit f/x for that.