Case Study: Tour de Vale Analysis with Strava Private Segments

Using Strava private segments to compare two rides of the same route.
The green line is cumulative difference between the times I took for the two rides.
My official time was 1 hour and 28 seconds faster,
while on the segments analysed it was 1 hour and 14 seconds.
Small, non-overlapping bits between the segments are probably the cause!
Anyway, if one is an hour quicker, who cares about 14 seconds!

I rode the Tour de Vale last year (2015) and this year (2016).
I was quicker this year.

Timing was done both times with timing chips on the back of the rider number, so is likely to be pretty reliable. Of course, the timing is "total time", not any of that "moving time" rubbish that wannabes like to use to pretend they are much faster than they actually are.
We can all go faster if we keep stopping for a break every now and then!


But there are no "split" or intermediate times.
For training, it is handy to know where time is being gained, and where it is being lost.
Enter Strava, the wannabe racers' favourite.




I set up 6 segments to divide the course up into (what seemed to me like) logical segments.
The segements vary in length a bit, because they are based on logical breaks rather than distance.

Segment 1 is from just after the start to the top of by far the biggest hill on the event. it is the only section with a 13% climb, and you can't get a decent run at it, because you have to turn right (in excess of 90 degrees) just before. Often it is a standing start to to having to give way to oncoming traffic. The top of the hill is where one slows a bit, and takes in a recovery drink! the reason my segment start just after the start is because the sports stadium usually has the side gate (used in the event) locked, so it makes sense to measure from just outside that! 13.7 km.


Segment 2 is from there (the end of segment 1) to a right turn at the bottom of a hill - again it involves a stop, as you usually have to give way to the road that you turn into. 12.6 km.

Segment 3 is from there to the top of a big hill, but has a couple of other decent short-sh climbs to keep you busy. 23.1 km.

Segment 4 is down the other side of big hill (and the maximum speed achieved on the ride is almost certain to be here!), then across some more open country to a crossroads, where, again, a stop is very likely, as one is not the "pririty" traffic stream. 15.3 km

Segment 5 is from the cross roads to another cross roads - flattish, with some minor up/downs. 18.6 km.

Segment 6 is from that second cross roads to the end. A couple of modest climbs to test tired legs.
Like segment 1, the segment is measure to just outside the stadium gateas, rather than the start/finish line, because the start/finish line is actully on a running track, and they don't let folks cycle all over it except for special events! 18.6 km.

Anyway, it is easy to see from the graph that I made the most time on segments 3, 4, and 5.
As these contain the more open country riding, I will have to have a look at wind direction and spped for the two dates. I'm pretty sure Garmin store that along with the ride data. Segments 1, 2, and 6 have more tree cover. Additionally, the predominant wind direct (Westerly) will affect segments 3, 4, and 5 more, as they all head more in that direction.

I would like to claim that it was my attempts at better aero that made all the difference, but let's just see how the wind numbers looked!
Segment 6 last year (2015) didn't go as well as it could have, because I misjudged the distance to the finish, and faded somwhat late on. Unlike 2016. But my choice of long segments doesn't reveal that slowing down in 2015, only that I was quicker in 2016.

Yes, things changed between the two events.
1) I was lighter
2) the bike (a different bike!) was lighter
3) I carried a LOT less stuff (weight AND aero savings)
4) I was wearing closer fitting clothes in 2016, not "baggies" like in 2015.

I will revisit this when I have the wind data (provided I can get it).
But this is just a first look at what can be done with private segments on Strava.
You will notice that my shortest segment is over 12km, and my longest is about twice that.
When you are riding 100km, it really doesn't matter how fast you ride on any particular 100 metre to 1 km section - what counts is how you string them all together!

Update: 11th september 2016

In 2016, I had a higher average heart rate on each of the six segments, compared with 2015

Another bit of data pulled from the 6 Strava "private segments" I set up to analyse the ride.
Three things stand out:
1) in both my 2015 and 2016 rides, the first segment produced the highest average heart rate;
2) in 2015, my second highest "segment" heart rate was on the last segment, while in 2016 it was on the third segment
3) in all segments my heart rate was higher in 2016 than in 2015.

The reason for the first obsservation is easily explained: The biggest hill is in the first segment! Indeed, the segment end at the top of the only "category" climb of the route (it is a big category 4 hill, with a short section reaching 13% gradient).

The second observation is more revealing. The big increase in heart rate on the last segment of the 2015 ride suggests I had energy to spare at that point, and had therefore underpaced myself for the earlier parts of the event. Indeed, memory tells me that is the case, as I set an informal pacing target of 135 bpm in 2015, only intending to breach that for the climbs. the second segment is a bit high for that 135 bpm target, but it contains the second biggest climb (about 60m at about 8%), as well as a shorter 10% climb, and a long, gradual, uphill to the second highest point on the Chilterns (Ivinghoe Beacon). The first and third segments have the bulk of the climbing, and that is reflected in the heart rate figures!

The third observation is also revealing. My heart rate was higher in 2016 than 2015. Period.
That may well be a reflection that I consumed twice as much caffeine in 2016, compared to 2015. In both cases, I took a can of "Red Bull" before leaving home, but in 2016, I consumed another can during the first two-thirds of the event. as caffeine is notorious for increasing heart rate, it is difficult to know whether I am measuring heart rate or caffeine consumption!

The "poor man's power meter" - Strava's power estimation.

Not having "real" power data for my ride, it is the turn of "the poor man's power meter" - Strava's algorithms.
Of course, they are a bit flakey, because they make loads of assumotions about rider position and rolling resistance.
So let's not get too hung up in the actual numbers.

For a start, if I swapped my 35mm Marathon Plus tyres for something like a 28mm Continental Grand Prix tyre, I would expect to see a drop in rolling resistance, and thus more speed for the same effort.
Similarly, Strava isn't going to know that my 2016 set-up was deliberately quite a lot more aero dynamic than my 2015 riding position.
I rode 2015 with full mudguards, a pair of panniers, and a "tops"-type riding positions, while in 2016 I rode a stripped-down bike, with a riding position closer to "drops" for much of the event, using my bar-end "aero horns" to get much lower than before. I also drafted other riders for about 50% of the 2016 ride, while in 2015 it was much less often).

And yet ... the estimated power numbers don't actually look that silly.
as I said above, I underpaced a lot of the 2015 ride. I also remember that I surged on the last segment, but rather ran out of steam well before the finish line, so that is going to show as more average power for a given speed that a more evenly paced effort would have been.

A look down Strava's PR for the overall route shows very few folks who ride the course have power meters (!), and, of course, all I know is whether the rider has power info, not how it was generated. Remember that there is a PowerTap PowerCal heart rate belt that uses an algorithm to calculate power from heart rate (actually, it calculates power from the differential of heart rate - that is to say, it calculates estimated power from the "rate of change" of heart rate, rather than heart rate itself). So I may just be looking another "guesstimate" algorithm when I compare it with Strava's own "guesstimate".

Anyway, the slowest rider with an actual power number beside their data on the overall placings for the entire 100km has an overall heart rate of 145 bpm, with a power of 154W.
Strava puts me (using their algorithm) at (2015) 138 bpm, 106W, and (2016) 148 bpm, 126W.
the rider with a power of 154W finished about 2 minutes in front of me, which for a 4hr+ event is minimal (!).
Another rider finished about 4 minutes in front of me, with a 162 bpm, but only 112W of power. It happens to be a lady with a lighweight bike, Given she would likely weigh a good bit less than me (I'm over 6 ft!), have a light bike (the UCI limit is under half what my bike weighs!), and is very fit with an decent aero position, and "race" tyres, rather than my Marathon Pluses (!), that is perfectly possible. It also adds credence to Strava's "guess" at my power! the numbers ARE realistic.
The next rider had an hr of 133bpm, with 134w for the power, and was 8 minutes fater than me (still not that much over 4-and-a-bit hours).

Looks like it is time for another graph!

ALL the rider data from Strava for those riders with recorded "power" readings.
I have added Strava's algorithm "guess" of my 2016 ride (marked with an arrow)

So there you have it.
You can see that the "trend line" has a few outliers, but that Strava's "guess" of my avaerage power over the 100km route is VERY close to that trend line.
As I said above, the lowest power was a lady on a lightweight bike.
The rest are all guys.
Guys tend to be bigger than ladies (just the way it works!), and at 6 ft 2" (188cm) and weighing in at about 187lbs (85kg) on the day of the ride, I am unlikely to be the smallest or lightest rider!

When I have time, I will draw some charts of each of the other 7 riders above over some of the longer "public" segments that make up the route, just to "validate" the "Strava algorithm". As long as it is consistent, it doesn't matter if the numbers are a bit high or low.

Certainly, at a quick look, my "Strava guess" numbers appear to be pretty typical :-)

Still, on Strava, and their "power" algorithms.
It is one thing to be about right on a ride of three to five hours.
Things like wind level out somewhat. There are uphills and downhills.
But what about a shorter segment?
What about one that takes about three minutes, rather than 180 to 300 minutes?

The fastest 16 "power" riders on a three-ish minute Strava segement.
There are faster riders, but the 16 chosen all have "power" data recorded in the KOM table.
The 17th "diamond" is me.
BTW, I am 78th out of 553 riders on the segment, and all the "power" riders shown are quicker than me.

All over the place.
This time I am NOWHERE near the "trend" line. Indeed, I am the "outlier".
So why so different this time?
Well, this segment has a slight uphill, and a slight curve, but if you get the wind behind you you will go a LOT faster!
Indeed, this is much more a "typical" Strava segment, where the fast times are set with a ferocious tailwind. My "Strava guesstimate power" has probably been affected by having a tailwind!
One other thing - this segment pretty much demonstrates why Strava KOMs are, on the whole, "junk". And, remember, at 2.1km long, this segment is one of the longer ones on Strava, Shorter segments are likely to be even worse!

Next time I will look at the 7 riders (and myself) from the first graph to see how they compare over shorter elements of the Tour de Vale course. At least we all rode it with the same sort of wind assistance!

Update: 22nd September 2016
Not more power stuff.
instead an analysis of the weather.
Head (or tail) winds can make quite a difference - indeed they are the determining factor on most Strava segments.

I have a Garmin 310XT.
It's great.
But it doesn't record either temperature (like the fancier devices) or wind direction and speed (not sure ANY common sports devices do!)

My dataflow is that I capture the data with the 310XT, upload it to Garmin Connect, then it automatically transfers to ("free") Strava (for social and some analysis) and ("premium") Training Peaks (for training and in-depth analysis).

So how does that help?
Well ...
Garmin Connect records weather data.
Only a single, simple, weather report for a point on the day.
Not sure if the weather is for the start point, the average, or the end point.
But it is a start.
Unless you are riding a very long ride on a day with a cold, clear, morning, that heats up a lot during the day, or there is a sudden swing in the wind, then it is a VERY good start.

So, I just happen to gave the Garmin Connect records for the 2015 and 2016 Tour de Vale :-)
And here is what was recorded:

2015 TdV data from Garmin Connect - note the weather record on the top right of the map.

2016 TdV data from Garmin Connect - note the weather record on the top right of the map.

 Same temperature (although 2015 felt hotter on the second half of the route, due to the direct sunlight, but there were some decent overcast sections as well.
Same wind speed, with the only difference being the wind was a notch further towards the West in 2016.
That pretty much ties up with my recollections of the rides (I remembered the sunny/overcast mix in 2015, even though it isn't shown in the Garmin Connect data, which just shows sunny).

On a circular route, surely the wind cancels out?
No, actually.
The first half of the course has quite a lot of wooded sections, including the all major climbs on the route. the second half of the route is more open, with fields and hedgerows (and not huge Bocage-like hedgerows either!).
Look on the 2015 Garin Connect picture. I have put a black mark near Pitstone. The mark is actully at the top of Ivinghoe Beacon, and is followed by a long descent, and straight. The desvent itself has some tree cover, but by the time you are on the straight, it is just you and the wind. And that remains the dominant feature for most of the rest of the event.

Btw, the times and distances for both Garmin Connect records are a good bit longer than my actual event times (about an hour in each case) and event distance (about 15 km in each case). The Garmin record includes my ride to the stadium and back, not just the event itself.

Conclusions:
The wind speed and direction is unlikely to have played a big part in my times.
The temperature may have played a modest part, as although the "background" temperatures look the same, the 2015 event was sunnier.