@Brad_Olwin It has a MEAN absolute difference (median error of means) of ±10 BPMs and a 95% confidence interval of around 4 BPMs. Meaning that for most of the 58 tested subjects the estimated value deviates by 10 BPMs and 95% of all tested subjects are in the range of 6-14 BPMs deviation. The data points also seem to be gaussian distributed. So for only less then 5% it has a lower deviation of around 6 BPMs and for some of these 5% it has even a deviation of more then 14 BPMs.
Of course, accuracy is highly subjective…but statistically seen, from the available data that I have, I do not use ZS at all.
For me personally, when I know that on average, even under perfect conditions, the ZS estimated threshold is off by 8-10 BPMs in either direction, it is not usable for me. Even the authors discuss that there set-up is highly artificial and it needs to be seen if it even works under normal run conditions and without exercise until exhaustion…also 10 BPMs above my threshold is for me all-out running for 1km while 10 BPM below, is the HR that I have on the first kilometers of a Marathon. For me that is just way too much deviation to rely on it for proper training control and also dangerous for people that believe they can trust the value and base their training on.
That I am not the only one this applies to, is already reflected by the statistics and the numerous posts from people in this forum that struggle with the value that ZS gives them and are looking for help to understand and interpret it.
Communicating the average accuracy under perfect conditions to the average non-expert consumer, would already help them a lot to interpret the value for themselves.
For the future, I am looking forward to more data and studies and maybe an improved algorithm, as the concept is really nice! But right now, it’s not there yet, in my eyes.
