@ Blue Meenie and John Whelan
Those were both thoughtful additions with what you added about TSS/CTL. I tried to steer away from analyzing sirpoc’s training (update: it’s clearly working for him) or redrafting the blueprint for implementing it, and it seems you’re both thinking in that direction too, refreshing!
I’d agree that if you run the experiment where the control is just running easy and the alternative adds in 2-3 quality days, for the same number of hours per week, there would be around a 5% improvement in performance within 6-12 weeks. That would actually reflect my personal experience too. However, what specifically you do (tempo intervals vs continuous tempos vs fartlek), I think, is more a function of what you respond to, both physiologically and psychologically. So that’s where I was going when I said workouts matter but so too does the individual.
Regarding the accuracy of TSS… there are very simple modifications you could make so that the formula is more accurate for runners, particularly for the range of paces that most people run their workouts. Again, if on the faster end, a score of, say, 90 isn’t even achievable because the duration required at the pace is not physically possible, but on the slower end it can be achieved with an easy to moderately paced longer run, that skews any analysis being done. I still think the system is useful though. Similar to how Jack Daniels uses a standardized running economy curve for VDOT, you could do the same with a lactate curve for assigning a stress score. A simple exponential, in my opinion, makes for a much improved equivalency across the pace spectrum of say 70% to 120% FTP.
However, CTL is responsible for tracking the long term training load, and that seems to be the focus of this discussion. It doesn’t surprise me that one notices a fairly accurate linear relationship between their race performance and CTL (particularly when you train as consistently as sirpoc has). You can take ANY positively sloped function though, from linear, polynomial, power, or exponential to represent your stress score. When you run it through an exponentially weighted filter, which CTL is, you end up with the same shaped function. A positively increasing, but at a decaying rate, improvement curve over time. Add additional stimulus to your training and you’ll accrue more CTL. So, yes, it can be used as a predictor (time to update the calculator haha), but consider the accuracy of the input.
I imagine the TSS/CTL approach is popular in the cycling community, because it’s simple and accurate enough. There are plenty of alternative models that exist and which are specifically meant to be performance predictors though. In fact, Andrew Coggan created TSS/CTL by simplifying a more nuanced performance prediction model (Banister Impulse -Response Model).
