All about Gear Effect - 6

Dave Tutelman  --  March 19, 2009

Rolling right along - is GRT great?

Figure 6-1
If you recall, my investigation of driver face roll was instigated by Marcel Bal wanting to know what gear effect said about Tom Wishon Golf Technology's GRT (Graduated Roll Technology) clubface. Tom's design stems from the premise that, while bulge has an understandable function because of horizontal gear effect, there is no obvious need to have vertical curvature. The downside of clubface roll is that low hits produce "worm burners", drives with extremely low trajectories.

Figure 6-1 shows the reasoning that leads to GRT. If you put constant loft on a driver instead of roll, you will get constant launch conditions, because the launch conditions are a function of loft, clubhead speed, and angle of attack. So low-face hits will not be low ropes, and high-face hits will not be sky balls.

This concept, unfortunately, is doomed from the start. One of the launch conditions is spin. Now we know that, due to vertical gear effect, the spin will be very different from a low hit to a center hit to a high hit, if the loft is constant. The function of face roll is to change the loft across the face to get a better set of launch conditions based on the height of impact. For instance, a high hit will have a lot less backspin than a center hit. So you want somewhat more loft higher on the face, to (a) raise the launch angle to better match the reduced spin and (b) increase the backspin a little. By balancing (a) and (b), you get better launch conditions and more distance. Similarly, a low hit will have more backspin than a center hit, so a lower loft is called for.[1]

Tom has explicitly taken the position that vertical gear effect contributes relatively little spin. If that were the case, then GRT would be a great idea. But a look at the Hotstix data shows this not to be the case. The data was taken from a driver with some face roll. If vertical gear effect were negligible, then backspin would be lowest for a low-face hit, and increase as impact moved up the face. Instead, exactly the opposite was observed; there is a significant decrease in backspin as impact moves up the face. The only reasonable explanation is enough vertical gear effect to overwhelm the change in backspin due to loft. This casts serious doubt on the idea of a constant loft driver.

A closer look at GRT

Let me start this discussion with a disclaimer. Most of the conclusions below are based on a physical model. The prototype testing that I have done is limited and anecdotal; those tests support the conclusion of the model, but I need to be as cautious about my own anecdotal evidence as I remind others to be.
Height on
+0.8" 15.7 14.7
+0.4" 13.2 13.2
center 10.7 11.8
-0.4" 8.1 10.7
-0.8" 5.6 9.6
While it is clear from physical arguments that a true zero-roll driver clubface must have performance problems, GRT is not zero-roll.The GRT clubface in practice is as close as Wishon could reasonably go to a zero-roll clubface. It has a 15" roll for the top third of the face, and a 20" roll for the bottom two thirds. The table at the right shows the loft progression up the face, for the optimized roll and GRT. The GRT does not have a flat face by any means, but it has a lot less loft change than the model says is ideal. (Note that the two faces were chosen to have the same loft at +0.4", the height that gives maximum distance for a 100mph clubhead speed.)

Tom's rationale, as I understand it, is:
  • Large-head drivers have tall faces. With the usual face roll of 9"-12", this causes in a big loft difference between low-face hits and high-face hits. The result is too high and too low trajectories to be very good.
  • The initial goal was the flattest face possible for the bottom 2/3, to keep the loft at optimum everywhere on the face. Because of the difficulty of fabricating a perfectly flat face, a curvature of 20" radius was adopted.
  • The top third of the face has a 15" roll, mostly because many golfers fit themselves -- through ego -- into too low a loft on their driver. So a bit of roll might help them. There was also a feeling that high-face roll might be advantageous in and of itself, but that was a secondary consideration.
  • This may produce ballooning for clubhead speeds over 115mph, but is beneficial below that.
I'm not sure Tom will agree with my synopsis. You can read a very detailed account in his own words in his forum. He also talks about the testing they did, and which golfers benefit most from GRT. Wishon has always been very ready to share technical information with the clubmaking community.

If the gear effect model is valid (and it has been validated), then it looks like GRT is not an optimal design. The model says that roll should be about 8"-10", not 15"-20". That conclusion immediately precipitated a discussion on the TWGT forum. The discussion pushed me to evaluate GRT more carefully. Here's what I found...

First, some personal experience. This is not a careful, controlled test, but rather a single anecdotal data point. Still, it told me what to examine, just in case the model had not.

In the summer of 2008, I got into a swing pattern that had me hitting my driver shots thin more often than not. Entirely too many drives at "quail height". So I got a Wishon 915CFE driver head with a GRT face to apply a band aid while I tried to fix my swing. I built it to the same specs as my usual driver, which has 10" of roll.

The GRT face definitely gave me better-looking drives. The worm-burners were a thing of the past. When I took both clubs to the driving range, it was clear that the trajectory difference was the GRT face and not my swing. The 915's loft did not fall off low on the face, and you could see the difference in the launch angle.

There was a downside, however. At the range and on the course, the carry distances were not very different between GRT and a conventional-roll driver. The GRT drives looked a lot better in the air, but they landed no farther up the range. And the total distance -- where the ball stopped -- was significantly in favor of the roll face driver. Eventually the GRT driver left the bag, and I lived with low line drives until my swing came around in late Fall.

It is worth adding that I should be an ideal candidate for GRT. Tom reports that his testing showed that it might produce ballooning for clubhead speeds over 115mph (higher than the average on the PGA Tour), but was advantageous below that. He also said that the 20" roll radius solved that problem. My aging swing is 80-85mph on a good day. Suffice it to say that I was disappointed that, while GRT solved my trajectory problem, its effect on my overall driving distance was negative.

Can the model supply an explanation? Turns out it can. I used the model to plot carry distance vs face height, for both the GRT roll and the ideal roll. But, because my GRT driver got a lot less total distance -- even when the carry distance was similar -- I also looked at angle of descent. (Angle of descent is the primary factor in runout, how far the ball goes after it lands.) Here are the results.

Figure 6-2

The carry distances are pretty similar -- within a couple of yards -- for strikes above the center of the clubface. They depart on the lower half. By 0.4" below center, GRT is giving up ten yards of carry to the ideal loft.

Height on
Loft Spin Loft Spin
+0.8" 15.7 2600 14.7 2250
+0.4" 13.2 3050 13.2 3050
center 10.7 3500 11.8 3900
-0.4" 8.1 4000 10.7 4800
-0.8" 5.6 4400 9.6 5650
But carry distance is only part of the story. The angle of descent is even more telling. Except for strikes quite high on the clubface (above +0.4"), the GRT roll produces higher angles of descent (AoD) -- considerably higher for the lower half of the clubface. An informal industry rule of thumb[2] says that you are losing substantial overall distance if you let your AoD get above 40. GRT produces an AoD above 40 for all impacts from the center of the clubface down.

Why should this be the case? The table at the right is the same as the table we saw above, but with columns for spin as well as loft. From the center of the face to the bottom, where GRT is helping the trajectory, it is also adding lots more backspin than the roll clubface. In addition, the lower launch angle of the roll clubface produces a trajectory that prevents ballooning and keeps down the angle of descent. (Launch angle can be estimated by eye from the table; it is 85%-90% of the loft.)

This would explain the results I saw from my GRT driver. Yes, I got a pretty trajectory even from a low-face hit, but it was associated with a lot of backspin. For a GRT face, the backspin was 5000rpm for impact a half inch below center. For the ideal roll driver, it was only 4000rpm. That, together with the lower launch angle, result in a much lower AoD and a bigger runout.

What about Wishon's caveat that there might be ballooning for low hits, if the clubhead speed is 115mph or more?
  • 115mph is around -- actually, slightly above -- the Tour average clubhead speed.
  • The calculations above are based on 100mph clubhead speed, so they should see the benefit of GRT if Tom is right. But they do not.
  • My own clubhead speed is dropping with age; nowadays it is 80-85mph on a good day. I should certainly have seen the benefits of GRT. I saw quite the contrary.
  • Elsewhere in this study, there are calculations of the sensitivity of optimal roll to clubhead speed. Speed does not seem to matter much. From 80mph to 120mph, the difference in optimal roll radius is less than 2".

Finally, let me mention -- in case you haven't noticed already -- that I have said nothing about fairway woods and hybrids. This discussion is only about drivers. Fairway woods have much shorter faces and shallower centers of gravity, so gear effect is not nearly as much of an issue.

Re-evaluating the model

So far, the model says that GRT is not as good as the commonly-found face roll in commercial drivers. Wishon has criticized this work as being just math without benefit of prototype testing. He claims that his prototype testing validates the superiority of GRT, but has not put forth any actual data.

Be aware that not all tests agree with Tom's. My own experience and at least one other forum poster have observed problems from too flat a face, especially on below-center hits.)

Added in July 2012: I exchanged email with a longtime Internet golf friend with whom I have actually played several rounds, even though we live at far corners of the country. Dale has a Wishon driver with a GRT face, and I had suggested he do a test run using impact tape to distinguish a high face hit from a low face hit. Here is an excerpt from his response.

OK, Dave, you were right! You wrote in an earlier email, “A GRT clubface will launch the ball at nearly the same launch angle as the high-face hit. But, due to gear effect, there is a lot more backspin on the ball. This results in some ballooning that will hurt carry distance (even though it still looks like a good ball flight). It will also result in a substantially higher descent angle, seriously curtailing rollout after landing.” [Emphasis added is Dale's.]
So I paid much more attention to impact this past weekend, and I am missing below center. Ball flight, as you point out, still looks fine, so I didn’t realize I was missing low until I started verifying at your urging. That explains a lot, and tells me what I need to work on. The ball flight and lack of roll you describe above exactly matches the drives I’m disappointed in.

Still, Tom's excellent track record of innovation invites further investigation of why his tests of the GRT clubs seem to be at odds with what the validated model predicts. Given Tom's record, there's enough chance that I am applying the model badly. So let's check in depth what might plausibly be changed about the model, that might cause it to show an advantage for GRT.
I re-evaluated a bunch of aspects of the computation I used to find optimal face roll. The details are in the appendix, but here is a summary. The aspects evaluated were:
  • Change the formula for vertical gear effect spin, 25yVb. This would be the most effective way to prove in GRT, because GRT seems predicated on an estimate of gear effect spin much lower than the model predicts. Vb and y are clearly proportional factors in spin, so the only thing left to change is the coef=25. Changing the coefficient enough to make GRT look good produces results completely inconsistent with the measured Hotstix data. This is a non-starter, because it requires a model at odds with real, measured spin data. We'll have to look at less all-encompassing changes.
  • Assume that shaft tip flex will limit gear effect spin. This was reported by Upshaw, but we could verify only relatively small effects, up to maybe 15% of gear effect spin. Looking at the practical issues (e.g.- keeping the shaft in a reasonable range for shaft fit), I tried a very generous 5% reduction of gear effect spin.
  • Try a lower estimate of the backspin produced by loft. We used a backspin formula that produced a best fit to several sources of backspin data. If we just fit to the source reporting the lowest backspin, we get 7% less spin. This also allows a smaller coefficient for the gear effect spin, in order to stay consistent with the Hotstix data. This is a fairly plausible change, since it corresponds to one of the sources of backspin data.
  • Increase the corrections for clubhead rotation. Our model used the midpoint of the clubhead rotation to estimate the correction. While this is sound, I tried a [physically impossible] estimate based on the full rotation, to see what it would do to the optimal roll.
  • Factor in the rolloff of COR away from the center of the face. As COR decreases, so does ball velocity Vb, a direct factor of gear effect spin. Making Vb fall off away from the center will reduce gear effect.

Figure 6-3
Figure 6-3 is a graph of all the factors and what each contributes to roll radius. Evaluating whether each could justify GRT:
  • Tip stiffness is slightly significant, but not very plausible. It requires ignoring proper shaft fitting to choose a shaft that minimizes gear effect.
  • Reduced backspin from loft is significant and reasonably plausible. As the only factor with both characteristics, I have temporarily adopted it for the GRT analysis above (Figure 6-2). So it is already factored in when you look at those results; there is no more to be had from the 7% backspin reduction.
  • Rotation correction of loft is very significant, but not that likely. It is easy to justify a half-rotation correction (as all the roll calculations in this article have done). It is impossible to justify the full correction in the graph. It is a stretch, but might be worthwhile, to take half the advantage (corresponding to three quarters of the rotation).
  • Rotation correction of speed is negligible, so we won't even worry about its plausibility.
  • Falloff of COR is also negligible. Although it is very plausible, its contribution is barely visible.
If we add up the plausible factors -- reduced backspin and half the advantage of rotation correction -- we can drive the optimum roll to about 10", but no further. Dividing things up low and high, as GRT does, the optimum roll is 9" high on the face, and 11" low on the face. That is not at all close to GRT.

Even if we were to add up all the factors, no matter how implausible, we are left with 12" high and 15" low, still well short of GRT.

So I don't see anything wrong enough with the model to justify GRT. Tom and I will have to remain in disagreement.


All this leads me to the conclusions:
  1. The concept of a flat face -- a single loft at all face heights -- is decidedly inferior to a proper roll.
  2. TWGT Graduated Roll Technology does have some small curvature; it is not flat for practical reasons. Still, the proper roll is considerably more curved than real-world GRT.
  3. There is a significant and noticeable distance penalty, both carry distance and runout, for a low-face strike on a GRT driver. The model says so. I have experienced it.
  4. One thing the GRT implementation got right: it looks like the lower 2/3 of the clubface wants to have about 20% less roll than the top third. But GRT uses 15" and 20"; the model works hard to justify even 10" and 12".


  1. For a raw-data measured example, the Hotstix data shows that moving impact up one inch reduces the backspin by 1300rpm, while the launch angle actually increases by more than 5 because of face roll. A quick calculation shows that, if the face had been flat, the change in backspin would have been well over 2000rpm.
  2. For instance, see the Golf Digest article on the 40 AoD rule of thumb. In fact, the article quotes Tom Wishon. "What's amazing is how visible it is," Wishon says. "You can easily see the difference between an angle of descent of, say, 45 degrees and one of 38. Anything under 40 degrees is a quality angle of descent for enhancing the roll. The secret to total distance is to use the driver that allows you to hit the lowest trajectory that carries close to the farthest."

Last modified - April 7, 2009