First of all, AACC #599 & #600 have both been bangers. Not just in my humble opinion but judging from metrics. Especially interesting that TR’s ‘Big Data’ showed that training impulses (like big-volume club/team training camps) can be detrimental to training progress just like zero-volume weeks can be detrimental to training progress. In fact, MORE detrimental if you look at the charts.
Everybody go listen to the episode. It is worth your time.
But this #600 episode is DRIVING ME CRAZY. Here is how TR says to calculate the metric:
Ok. Average training hours/week over the last 12 weeks. Calculate the standard deviation for that series. Divide standard deviation by the mean & that’s your consistency score.
I feel like I understand that but then there is this slide:
What!? I can’t get any of those numbers. Ok, the average looks like it’s over 13 weeks instead of 12. Fair enough. But no standard deviation calculation gets me the consistency scores listed.
How am I screwing this up? Or maybe it’s close enough just to follow the 12wk stdev/12wk avg & just forget about the rest…
I thought it would be a bad number because in that time frame there were two events & training weeks previous to the events were light…plus I traveled for one so that introduced further inconsistency. But, really, it’s a pretty forgiving metric considering the positive benefits accrued.
Definitely makes you want to geek out with the data set. Is 7 days the best period to track? Is standard deviation really the best metric to put in the numerator. Those TR employees are spoiled for data, for sure. I can see where you’d be tempted to spend the whole day poking around the data & not get any work done.
I made a spreadsheet to calculate my consistency score. In one column I input the hours for each week and at the bottom of that column it spits out the score. I formatted the score cell to be green when under 35, yellow when between 35-80, and red when over 80.
…then I thought, what if I calculated my score using weekly TSS instead of hours? So I made a second column using TSS instead. My TSS consistency score was better than my hours score. I feel like TSS is a better metric to use than training hours. Does anyone have input on this?Training Consistency Calculator. Hours or TSS.
Yeah definitely. what would be interesting would be to see what the margin of above and below the mean weekly hours/tss etc, was to ensure highest gains and personalise this according to user. So each person can say “right I’ve got an extra x number of hours this week, Does doing that constitute a deleterious increase?”
Do younger riders have great absorption capacity for variability?.
Also need to take into account that a declining, albeit consistent, mean volume/tss etc, is likely to result in negative effects.
Now clearly if you are following progressive overload and recovery cycles; you will get variation. If you are able to vary your weekly hours and loads. If your hours and load remain perfectly static, that has a 0 coefficient of variation.
It looks good just based on this metric. Indeed your weekly hours are consistent. But there is no progressive overload and recovery.
Neither the mean nor standard deviation can be negative for this data. You cannot have negative weekly hours nor load. The coefficient of variation will therefore also be positive.
I’ve had a look at my data over the last 12 months (Uploaded a file from Intervals.icu) and got 40 for my hours and 44 for my data. The thing is, the reason I train is to have regular big rides throughout the year with my friends. I realise I could train better by not doing the rides that I train for, which wouldn’t make much sense
I do think it would be a neat feature to have this added in. I know its been mentioned elsewhere on the forum before but I think a general compliance score/page would be helpful as well.
We all think that we are performing well and following the plan, but sometimes seeing what is actually going on can be helpful.
The message seems to be simple. Consistency leads to the best improvement. It would be good to see how we are doing in that regard.
I know, but it might be ok to have 3 weeks in a row of declining hours. If the decrease is say only 10% preceding mean volume, whilst still raising ftp.
You are describing a 3 week taper. Essentially a slight decrease in fitness, but much bigger reduction in fatigue. You are better able to express the fitness you have.
Interpretation: A CV of about 14% indicates your weekly training volume is fairly consistent. The main source of variation is the 22:28-hour week, which is substantially higher than the rest. Without that week, the CV would drop to roughly 8–9%, indicating very consistent training volume week-to-week.
I guess I’m pretty consistent with very little rest!
Did mine the old fashioned way, and got 26—surprising, because the last couple of months have felt wonky/uneven to me, but it’s all relative.
I submitted a Q to the podcast, though: if a low CV (which is what Training Variability is) is important for long-term progress, the advice that used to be standard regarding long/big events—“you don’t have to train long rides to do a long race”—might be a bit off. That is, getting closer to that high-volume week with high-volume training puts you in a better position to recover and get back to your standard week, right?
I think many folks on the forum who do long/big rides have said this already, but as I get ready for big things, I do kindof disregard TR’s volume reccos and do my own thing. Long rides, esp. two long days back to back, bring up a red day on the TR calendar, but they really do help me feel ready for an epic.
I would say TR has moved away from the idea that long rides aren’t needed. But they can speak for themselves. Maybe it’s better to say you don’t have to do long rides to be successful at longer events, but if you do add long rides to your training you’ll have a greater chance of success at your longer events. You just find out stuff on long rides that you don’t find out on shorter rides. Nutrition stuff, bike fit stuff, equipment stuff, battery stuff…all kinds of stuff you don’t want to discover for the first time on race day.
I’d assume it has to be 7 days weeks to make any sense with most people’s schedule & plan. It would be really interesting to have a tool to shift the start point of the “week” and see if that uncovers hidden trends in consistency.
Standard TR calendar looks at Monday-Sunday, I’d be interested in how looking at Saturday-Friday might expose big swings in training volume/intensity that we actively cover by adjusting the length of Sunday endurance ride after looking at the week stats.
Or a breakdown on Weekend vs Weekday training would be cool.
Ran your numbers through another LLM and it gives different values for the sample stdev and mean. Can’t be bothered to go to PC to get actual values. I don’t think it would affect the general conclusion as it’d still be in that ball park.
It also highlights the simple fact if you have much higher average weekly hours, you can also have a bigger deviation in those hours, than say someone only averaging 5 hours a week, and still have a lower coefficient of variance.
That would make no difference if the periods being examined are still 7 days. The sample stdev and mean would still be the same (give or take minor differences as you are sampling a slightly shifted set of data) looking across 12 lots of consecutive 7 day periods.
