Good day possums!
If you are a running nerd, there is nothing more fun after you've done all the training than analyzing the results to see if you can learn anything. You sit back with your gradually tapering tired legs in your cushy chair with a nice drink and hack away.
Since I recently completed my second coached training cycle for a marathon, I thought it would be cool to compare various data from my Garmin watch and heart rate monitor to see if we can see anything interesting.
In the last months I've been adding to my python library that lets me crunch the *.tcx files and do various extractions, filtering and plotting (btw: using pylab/matplotlib for the also nerdy)
For a lot of you your eyes may just glaze over about now, but hang in there and let me see if I can walk you through all the results.
The first thing you need to now is what does the initial data look like? For each run I have the data for every 'lap'. My watch is set to auto-lap for each mile, so I have at lots of 1 mile laps. Sometimes, e.g. on the track I'll push the lap button and get shorter laps.
For each lap I have the time, distance, ave heart rate and from these I can obviously calculate pace as well.
I decided to look at two sets of runs for each marathon buildup: a 90 day set of runs and a 42 day set. The 90 days results will pretty much show the average of all the training during each cycle and 42 day is the last 7 weeks of the cycle when you are focusing on more tempo work. Hopefully, we'll see the fruits of our labors in that 42 day set.
So, you'll see references to: cim 42d, cim 90d, eug 42d and eug 90d, the 4 data sets.
The first two graphs are histograms that record the total time spent running at various paces. In these graphs the cim and eug data are displayed on the same graph in different colors.
On the first graph we have the 90 day set, we can see the green color (eugene) has a peak at 9.25 min/miles (aka 9:15-9:30m/m), where as the blue (cim) is at 9.00 min/miles). A lot of my base miles were a touch faster during the CIM buildup. I also didn't have as much time at 8.5 m/m paces for eug (blue bar has another 100 minutes of that).
On the other hand, at paces above 8.25, there is more green: about 90 min of 8:00 pace and 50 min of 7.5 m/m. So instead of 8.5 m/m stuff Jill had me do some long intervals at various faster paces.
This next graph shows the same type of data, but for the 42 day set. Looks pretty much the same except, you can see that now I spend less time at slower recovery paces (10 ++) than I did during the cim buildup.
Now we move to a different data set: this is a histogram of miles spent at various heart rates from 100 to 150. I also have some overall stats on the title line...miles about the same, HRs too, but pace a touch faster average for eug.
You can see the green peak bars at 120 to 125 BPM...they correspond roughly to all that base pace running at 9:15 to 9:30 m/m. You can also see more blue color at the very low recovery HRs (110 or so)..that's that cim recovery running I'll bet.
Same data again but now 42 day...it looks similar but the average pace for this period is 9:11 (eug) vs 9:38 (cim). Give the ave HRs are about the same you can see I have gotten fitter than cim. I like to convert the pace/HR data into something called yards-per-beat, as you know...a measure of overall running economy. It's 1.51 vs 1.45 during CIM which is a pretty whopping 4% better.
Now a very colorful graph showing each run for the last 42d for cim (newer runs on the right, older on the left). The HR mix during the run is show by the color stack in each bar. You can see my last three long runs (19, 20, 22 miles) clearly sticking up. Between the last two long runs is a 14 mile run which is almost all orange (80% working heart rate). This is my 13.1 mile time trial run.
Same graph for eug, fewer runs (but 3 during my taper are not entered yet). Less red/black going on here...(never hit the really high HRs as not doing many hill sprints on this cycle)
Now the same two graphs again but for 90 day periods. You can see the total mileage is higher for eugene (and about 14 miles is not entered to boot). The long runs for eugene have more green and less higher HRs it seems to me.
Now we get into some really interesting stuff. AFAIK, I'm the only person that tracks this stuff and I find it very interesting. Here I am plotting the yards-per-beat for all runs that were better then 1.45 (weeds out the hilly runs). Same as the above graphs, new runs to the right) For cim it's pretty flat over time:
For eugene, you can see it has improved during the last 42 days. We already knew that my y-p-b was getting better from the graphs above but here you can see a best-fit line plotted against the run samples.
Ok same data again, but now for the whole 90 days. And look at this!? During CIM i seemed to get less efficient overall. I'm not sure why this is, perhaps because I was extremely tired or because of my developing AT.
This eugene cycle shows a clear trend of getting more efficient. 4 runs that are better than 1.55 on the right side are dragging up the average I'm sure. This seems to have started around 30 days ago. It could be due to my training mix, shoe lift fixing leg strength imbalance or ???
And now another graph I like, this plots averae pace vs HR for each run with a best-fit line. I print the slope of the line in the bottom comment.
13.1 means that it takes a 13.1 bpm increase in my HR to speed up my pace by 1:00 min/mile. Amazingly this function is always linear in the 'normal' running paces. The intercept number says what HR would be required to run infinitely fast 8) 8). Actually this number lets you compare to graphs to see if the line has an up-or-down offset...lower is better.
You can see this data is pretty consistently saying the slope is about 12-13 bpm per m/m. The intercept is lowest for the 42day eugene data. i.e. now.
Kudos to you if you've gotten this far! If you spent more then 5 minutes looking at this stuff you are a nerd for SURE 8-P
Objectively, it looks I am in better shape on this buildup than I was for CIM. I already new this, but being a nerd I like to see the facts tell me so. I also like having the data for future training comparisons.
So, instead of 9:02 average pace I will attempt 8:55. (8:50-ish on the Garmin)...this would be about a 3:53 marathon.
Weather still looking good for race day!
I LOVED this post. I wish that I had the ability/time to compile my numbers. I followed and enjoyed your data; so I must be a nerd.ReplyDelete
The last chart is the best IMHO.
Well done and cudos.
Holy freaking hell, if you don't frame those graphs, I will!!! LOVE THEM!ReplyDelete
Ur training plans in living color 8)Delete
Cool Paul....very cool. By the way, what do you estimate to be your maximum heart rate just out of curiosity..?ReplyDelete
And I agree...you do seem to be much more fit, at least according to the data presented.
My max HR is about 168-170. This was tested by running me into the ground on a treadmill at 12% incline in the doctors office 8)
The charts and data is awesome!!!ReplyDelete
Wow this is intense! :) Good luck in Eugene, can't wait to read the race report.ReplyDelete
Analysis aside, Eugene is built for a negative split. Especially if you have the right strategy. Best of luck. Run well!ReplyDelete
Hello Paul, I am also doing various stats on my run results, and also discovered weird randomness in the Efficiency plots (yard/beat). Now I know it is due to the fact that it is not a constant value, but it varies with the pace. Even if one day you run at 130bpm and another day at 140bpm, this is already a difference large enough to mess your efficiency plot up.ReplyDelete
What I find more reliable, is to use one, fixed Pace_vs_bpm curve (red line from your last figure) and compare all runs to it. That is, always calculate how many second ahead you are in respect to that curve at a given heart rate. Once a year or so you can revise the reference curve, cause it will change as you become more/less fit.