Launch Monitor Data: What to include?

This is the place for any questions and discussions about TrajectoWare.

Moderators: fschmidberger, dtutelman

Post Reply
Anthony
Trusted user
Posts: 19
Joined: Mon Jun 11, 2007 11:37 pm

Launch Monitor Data: What to include?

Post by Anthony » Tue Mar 25, 2008 2:40 pm

Dave,

Hello Once Again,

I was going to send you this question through a PM since it is not directly related to trajectory, but figure maybe someone else might benefit from it. Feel free to delete it.

I was wondering if you could give me your opinion regarding averaging the results I include from my LM. What I mean by this is that when I put someone on the LM and they hit, lets say, 12 shots. If there happens to be some outlying data (let's say a worm burner) I delete it from the data. I do this to get more accurate results. I'm thinking maybe all shots should be included as this is something that probably happens on the course as well. Also, I sometimes throw out a few of the highest and lowest shots to get info on the better shots hoping to maximize the persons decent shots without figuring in the not so good shots.

I'm hoping you could post your opinion on this when you have a chance, as to what is "mathematically or scientifically correct" regarding something like this.

Anthony

dtutelman
Site Admin
Posts: 107
Joined: Mon Apr 09, 2007 8:55 pm
Location: New Jersey, USA
Contact:

Re: Launch Monitor Data: What to include?

Post by dtutelman » Sun Apr 06, 2008 1:49 pm

Sorry about the delay. It's a busy time for me.
Anyway...
Anthony wrote:I was wondering if you could give me your opinion regarding averaging the results I include from my LM. What I mean by this is that when I put someone on the LM and they hit, lets say, 12 shots. If there happens to be some outlying data (let's say a worm burner) I delete it from the data.... Also, I sometimes throw out a few of the highest and lowest shots to get info on the better shots hoping to maximize the persons decent shots without figuring in the not so good shots.

I'm hoping you could post your opinion on this when you have a chance, as to what is "mathematically or scientifically correct" regarding something like this.
Anthony,
Data analysis is an art in itself. I have taken a one-semester course on the subject (years ago), and suspect it barely scratched the surface. Here are a few thoughts on the subject:

Median, not average - When you take an average, you are trying to find a "central tendency" to the data. You really don't care what the mathematical average is; you want to know what the "typical" value is -- whatever that means.

If you don't have a good handle on the mechanism generating the statistical variation (and you probably don't for an arbitrary golfer's swing), then you are probably better off using the median than the average. One nice thing about the median is that it automatically throws away outliers (you don't have to), unless outliers constitute more than half the data.

Here are a few references about the median -- how to compute it, and what it means: Impact tape - Most data analysis problems present you with a bunch of numbers and ask "tell me something about this set of numbers: typical tendencies, variation, etc." But we as clubfitters are better off than that. We actually saw the shot. We know if it was a terrible swing. In short, we have more information at our disposal than just the numbers. You indicated you already make use of this "side information" (as data analysts would call it) when you talked about "worm burners".

You have another very important tool for gathering side information: impact tape. This can tell you at least as much about the quality of the hit as the LM numbers themselves. I would use it as follows:
  • Start by fitting the driver to minimize the number of mishits. More specifically, find the driver that gives the best pattern on the impact tape. At this stage, you are not paying much attention to the LM except to see if the trajectory is OK (slice, hook, etc), but rather to visuals of the swing and what the impact tape is telling you. The things you are focusing on are club length, measures of heft (swingweight, MOI, total weight), grip size, etc. These affect the quality of impact rather than the distance numbers.
  • When you know the specs of the club, then and only then do you start with things like clubhead CG and flex profile, and only then do you worry about the sort of optimization that a LM can help with. But it is still very OK to keep the impact tape; it won't have a significant effect on the LM numbers. And the impact tape is the best way to tell whether to drop a data point. Think about it: an on-center hit with a club that fits the golfer is not one you want to reject as an outlier -- even if the LM numbers were totally different. The fact that such a shot can exist tells you something valuable.
Hope this helps,
DaveT

Anthony
Trusted user
Posts: 19
Joined: Mon Jun 11, 2007 11:37 pm

Post by Anthony » Thu Apr 10, 2008 1:52 am

Nothing is ever easy is it?

I think for me the way to proceed is to certainly disregard any obvious mis-hits (skulls and the like). As far as loft optimizing I think I'll look at it as two scenarios. One that is optimized for the best of swings and contact and the other for lesser qualitty swings (not mis-hits though) and then decide based on whether I feel the player will most likely be improving or not.

dtutelman
Site Admin
Posts: 107
Joined: Mon Apr 09, 2007 8:55 pm
Location: New Jersey, USA
Contact:

Post by dtutelman » Thu Apr 10, 2008 2:12 am

Anthony wrote:Nothing is ever easy is it?
Hey, you asked.
:-)
I think for me the way to proceed is to certainly disregard any obvious mis-hits (skulls and the like). As far as loft optimizing I think I'll look at it as two scenarios. One that is optimized for the best of swings and contact and the other for lesser qualitty swings (not mis-hits though) and then decide based on whether I feel the player will most likely be improving or not.
Whatever floats your boat. I think there were a few points worth extracting from the dissertation above. In order of importance:
  1. You need to get a good fit before you even try to optimize. By a good fit, I mean a driver that the subject swings well, measured by ball speed and consistency of good ball speed. Otherwise, you'll get what optimization experts call a "local optimum", which is often far from the real optimum.
  2. Impact tape is an important tool in getting the fit, before you start to optimize. It can be helpful (but is not essential) in deciding which hits to throw away while you optimize.
  3. If you are averaging a collection of hits, the median will probably give a better answer than the average.
Hope this helps,
DaveT

Anthony
Trusted user
Posts: 19
Joined: Mon Jun 11, 2007 11:37 pm

Post by Anthony » Fri Apr 11, 2008 5:00 am

I didn't mean to gloss over the good fit aspect of all this. I just figured that it goes without saying. I certainly feel that impact tape is indispensible - certainly an eye opener for most folks.

I also plan on sticking with the median also. I 'll have to call Zelocity one of these days and find out what they use in there software. I'd be surprised if it is a straight average. Also, they now have a descent angle which I have yet to look at. I'd be curious as to how it compares to trajectoware.

Anthony

Post Reply