Tuesday, 21 March 2017

Efficiency & Spin Axis

As someone who finds the maths and physics of Golf almost as interesting as playing I read and watch a lot in an attempt to improve my understanding. This led me to dig a bit deeper into my driver data, specifically efficiency and spin axis.

ST does not measure clubspeed it is calculated (with good accuracy). It measures ball speed with high accuracy. I therefore extracted my data and calculated bsCarryEfficiency, bsTotalEfficiency, csCarryEfficiency and csTotalEfficiency.

The PGA tour measures carry and total efficiency comparing clubspeed to carry and total distance.
For carry efficiency the 2017 range is 1.918-2.616, my average across my ST data is 2.4
For total efficiency the 2017 range is 2.391-2.936, my average across my ST data is 2.63
Fig 1
Figure 1 shows my bsCarryEfficiency distribution. The mean is 1.64, Q1 is < 1.6 and Q4 is > 1.69. It therefore seems reasonable to conclude anything at or above 1.69 is an efficient drive.

Having watched numerous youtube videos of various professionals I noticed a number of them were posting driver data with significantly greater carry distances than myself despite a similar or lower ballspeed. Spin rates and launch angle were similar so the distances should have been comparable. I therefore looked at other launch parameters and once again found myself looking into Spin Axis:
Fig 2
Figure 2 compares spin axis with ball speed carry efficiency for my driver data. PGA tour style carry efficiency would be better but I didnt want to rely on a calculation and using ball speed reduces (not eliminates) the effect of strike on the data. Within this data there is a curve (albeit with a low r2 value of .3397); increasing spin axis (left or right) generally reduces efficiency which makes sense. Using this data an efficienct drive requires a spin axis of +- 20deg (+-700rpm of side spin). My normal dynamic loft is 15.5deg so I need the face and path to be within 4.5deg of each other to hit an efficient drive (data taken from trackman university). The low r2 value will be the effect of strike and delivery, so you can have a low efficiency straight drive but you cant have a high efficiency crooked one (from the perspective of carry).

Skytrak, Trackman, GC2/GCQ
Going back to my youtube comment I decided to add some extra data; I added my trackman data from my recent Skytrak vs Trackman test, and manually captured data from youtube to see how the data compared. All of this data was produced on GC2/GCQ from various pros and also some amateurs:
Fig 3
Figure 3 shows the bsCarryEfficiency for Skytrak, Trackman and GC2/GCQ. Trackman shows a wider range which makes sense, it is measuring actual ball flight and there are lots of variables at play (open data). However, I think the GC2/GCQ data proves my observation that the youtube carry distances posted are significantly higher than ST and Trackman. This means for every mph of ball speed you put in a GC2/GCQ will likely calculate a more generous carry distance than a ST or a TM would observe.

Returning to Spin Axis
I then decided to reproduce the spin axis chart with the additional data and see how it compares:
Fig 4

Figure 4 shows the efficiency vs spin axis plot for the 3 data sets and I this this illustrates a few things:
1. The TM efficiency data is comparable to ST except the spin axis values are generally smaller. This was already noted in previous testing.
2. The range of spin axis measurements for ST and GC2/GCQ are comparable (more extreme than TM) so perhaps the ST isn't exaggerated compared to GC2/GCQ? Much more data would be required to prove this.
3. The GC2/GCQ is significantly more generous in its carry calculation, to the tune of 10 yards at my average ball speed of 155mph.

Thursday, 9 March 2017

Flightscope Trajectory Optimizer - Launch Angle


This is a great tool and I use it frequently to assess my launch conditions against a different flight model (Flightscope instead of Skytrak or Trackman), and also to experiment with different parameters so I can analyse how they affect distance.

I launch my driver quite high, on average between 14.5-16 degrees. Trackman suggests the optimum for my ball speed is closer to 13 degrees so I wanted to assess the impact and it turns out the results are quite interesting.

The coloured columns are carry, total (hard), total (soft).
This suggests high launch is not a big issue, and strongly indicates it is better to err on the side of launching too high vs too low. According to this launching at 17deg is optimum for carry, but even launching at 25deg will only cost you 6.5 yards of carry and approx 10 yards of total distance in hard conditions. Conversely reducing launch angle to 9 degrees will cost you 10 yards of carry. From this point carry distance drops rapidly with decreasing launch.

N.B. the dip in the curve for total (hard) at 10 degrees launch.

133mph ball speed
I thought it might be interesting to compare this with a lower ball speed (the average ball speed of player a from our Skytrak vs Trackman test).

A similar pattern emerges and optimum launch increases to a whopping 20-22deg. How many amateurs launch their driver at 20-22deg???

The dip in the curve for 10deg launch is much more pronounced. Not sure what this indicates???

Friday, 3 March 2017

Skytrak vs Trackman Pt 2 - Shot by shot Player A

After the first part I wanted to try and illustrate the data captured on a more blow by blow basis so this is that attempt; this will therefore be a very chart heavy! Each chart shows the Skytrak and Trackman data, aligned shot by shot.

This is the actual Trackman data with no normalisation. Differences in calculations from the ST flight model should therefore be expected (but I still expect it to follow the measured data (e.g. short shot = short shot etc).

Summary data is useful but I think the below shot by shot charts illustrate just how accurate the ST is.

Ball Speed

I think this is highly impressive and shows excellent correlation between both systems. The difference is normally negligible peaking at approx 7.7mph in the worst case.

Club Speed
As a calculated parameter this is surprisingly close to the measured Trackman value.

Vertical Launch

Horizontal Launch

I think this shows the subtle difference in alignment between the 2 units. If you subtract approx 1.5 deg (turn the ST 1.5deg left) they match even more closely. I do not believe these 1.5deg affect the spin axis measurement but they do affect the offline projection.

Total Spin

I think this is the most impressive chart of all, Skytrak can measure spin rate with phenomenal accuracy. It is this which makes me think the Spin axis deduction is not calculating as expect (as opposed to capture). The sizeable deviations are caused by Trackman not returning total spin.

Spin Axis

As previously noted this is the only troublesome measured parameter. I think this chart clearly shows the Skytrak is measuring Spin axis (I would even say accurately) but for this player it is predominantly skewed in a negative direction. For myself (Player B) I am pretty sure we will see the reverse (skewed positively) because our shot shapes are opposing.

Flight Duration

A calculated parameter for Skytrak. This is impressively close considering we were hitting range balls with a slightly reduced distance.

Descent Angle
A calculated parameter for Skytrak, impressively close to Trackman for the majority of shots.

Peak Height

A calculated parameter for Skytrak, impressively close to Trackman for the majority of shots.

Smash Factor

A calculation based on a calculation so potentially risky. However, it seems impressively close to the Trackman value. 


As previously noted the wind was slightly helping and from the right. You would therefore expect the ST shots to finish right of the actual TM shots? However, if you take into account the circa 1.5deg alignment difference between the units (ST alignment approx 1.5deg right of TM target) I think these are almost identical.

Carry Distance

Carry distance looks spot on considering they were range balls.

Total Distance

As before total distance also tracks as you would expect (for a range ball).

Skytrak vs Trackman Part 1

I recently conducted some testing to compare my Skytrak to Trackman. I have read a number of comparisons (mainly by golf magazines etc) but none of them come across as very rigorous and they didn't share complete data so that others can peer review/scrutinise.

The objective of this testing was simple, assess the accuracy of my Skytrak against the industry standard Trackman and present full results so others could peer review.

Myself and a playing partner hit balls with a variety of clubs collecting the raw data from both systems.
The balls were range balls (I will cover this later), but they were high quality srixon range balls. In Trackman ball conversion terms they are the equivalent of hard (reduced spin with short irons, elevated spin with long irons/woods).
The weather was excellent (for Scotland in Feb!), approx 6C with a light wind helping slightly from the right.
Skytrak software version was 2.5.4 on and Ipad Air 2.
Trackman 3 in Outdoor mode.

Results (Player A)
Source Data Link

These are the results averaged across all shots and shows excellent correlation across the data except spin axis/side spin. This leads to reduced accuracy for the offline figure. However, I think it is important to remember this is range ball data hitting outside so this is comparing what actually happened to the ball vs the Skytrak flight model. 

Initial Thoughts
  • My Skytrak behaved impeccably throughout this test. Trackman missed 1 shot whilst we were setting up, and Skytrak missed 2 shots during this test. I think this was down to my playing partner hitting the followup ball too quickly.
  • Trackman frequently missed spin data (you can see some of this in the detail worksheet). Skytrak did not miss once, and the results match so closely I found myself trusting the Skytrak data over the Trackman (The trackman spin numbers appeared to fluctuate more).
  • Some people have complained about the delay on Skytrak. For plenty of the longer shots we found ourself waiting for Trackman to finish tracking the ball and present its full data rather than the Skytrak!

Horizontal Launch
This was one measurement I expected to vary because of differential alignment between the systems. Despite this I believe the Skytrak data is very accurate and consistent. The spec for Skytrak is 2Deg and it was within this tolerance.

Spin Axis
The spin axis seems to be exaggerated frequently leading to excessive curvature on shots. My initial response on witnessing this was the Skytrak struggling to read/calculate the side-spin component accurately.  However, given the accuracy of the total spin (or backspin component) and horizontal launch angle it seems highly improbable the raw data capture was to blame. It therefore seems likely there is something weird in the way that raw data is converted into spin axis.

I raised this issue with support and was advised to reinstall the Skytrak software which I have done. At some point I will need to retest and see if the spin axis is captured more accurately.

With the exception of the Spin axis phenomenon described above I was surprised by how closely my Skytrak matched the Trackman data. I think the important parameters (ball speed, v-launch, h-launch, total spin) were measured with highly impressive accuracy especially when you consider the cost differential (2k vs 25k) and the fundamentally different ways in which these units operate.

I will also say that I couldn't use a Trackman at home due to its space requirements, but if I wanted an outdoor ball tracking device and cost wasn't a factor I would definitely have one. The Trackman software is phenomenal and the amount of data it captures with unerring accuracy is amazing. When conditions allow there is no substitute for measuring what actually happened vs what should have (even though the latter is very accurate!). It is the benchmark.

Next Steps

  • Try to summarise across all clubs, shot by shot.
  • Player B data.
  • Open pandora's box and delve into some of the raw Trackman data!
  • Experiment with normalisation and see how that affects the data!
  • Examine some individual shots.
  • Find a GC2/GCQ to test against!