Ready, set, go?
Some brief thoughts on readiness scores
Alan Couzens, M.Sc.(Sports Science)
One of the core concepts of my coaching practice is using readiness scores to help to guide training decisions. So, I figured it was time to dive into this concept in a little more depth so that you, the athlete, understands all that lies behind this very important training metric. Let's start at the beginning...
What is a 'readiness score' & how do I use it to guide the training?Simply, a readiness score is a way of assessing the athlete's current overall "state" by combining their HRV, resting heart rate, sleep, life stress, mood, soreness and fatigue into one ~0-100 "readiness score" that I can reference before determining the training prescription for the week. Each of these inputs is plugged into a neural network (see below for a simplified visual) and an estimate of the athlete's readiness for a given session is made. When readiness for hard training is low, I will plan recovery. When readiness for hard training is high, I will plan the key "loading" workouts.
This "adaptive" approach of taking recovery when the body needs it runs counter to & is, in my opinion, vastly superior to more traditional 'fixed patterns' of loading e.g. 3 weeks on, 1 week off etc.
So, why do I use a composite readiness score instead of just looking at HRV?
Simply because, when done right, a combined readiness score is a better predictor of both how the athlete will feel given a particular session/workload *and* how much fitness improvement the athlete will get from that session/workload!
Given sufficient data, the more useful features we can include in a predictive model, generally, the better the model performs. By including a simple morning wellness questionnaire coupled with HRV & RHR data over a long period of time, we have access to a lot of useful context on the relationship between the athlete's health and wellness parameters and their performance. The next step is to leverage that information in the right way to make better decisions!
So, just what does "doing it right" entail?
First and foremost, the scores must be relative to the athlete's normal values. In other words, if I have 2 athletes, one whose heart rate variability regularly swings by 50ms per day & another whose HRV stays within 10ms, it is completely pointless to set an arbitrary cut off of, say a change of 20ms, to determine when the athlete is tired. The scores must be tailored to both the athlete's average score and how much deviation is typical around that score. The easy solution for this is to use z-scores, where 0 represents an a average value, +1 is one standard deviation above the mean, -1 is one standard deviation below the mean etc. Scaling shifts in HRV and the other measure this way normalizes the score to each athlete &, for subjective measures like mood, eliminates the issue of the type of athlete who always ranks their mood as fantastic or always ranks themselves as tired. Z-scores zero the measure to the athlete's individual norm.
Similarly, readiness must be normalized to the relative intensity and duration of the planned training session. When we say an athlete is "ready", ready for what?! Ready for an active recovery walk? Ready for a personal best Ironman? It should go without saying that any readiness assessment should also take into account the nature of the task that we want the athlete to be ready for!
Finally, each of the measures that make up the readiness score will have different weightings for each individual athlete when it comes to what shifts in the score mean on a practical level. For example, some athletes will have a lot of variability in their sleep patterns due to life circumstances-maybe sleeping 6-7 hrs during the week & catching up on sleep with 9-10 hrs on the weekend. While not ideal, these shifts may not be indicative of fatigue & may not affect the training. OTOH, for some athletes with very stable patterns of 9 hours every night, a shift to 7hrs of sleep may be more practically meaningful of something that will have a potentially negative bearing on the athlete's ability to handle and respond to a given type of training. For this athlete, their sleep deserves a higher weighting in determining readiness.
And, just how do we determine these weightings? How do we determine what's important?
In addition to collecting these morning metrics, we collect data on how the athlete's fitness is improving along with how they are feeling during each type of session. By comparing the readiness scores with these two factors, we can 'calibrate' the weights of each of the factors - sleep, sleep quality, mood, soreness, HRV, resting heart rate, fatigue to determine which are significant "warning signs" of impending doom.
That is, we can take each of these parameters, apply an individualized weighting to it & predict how the athlete is likely to respond to a given session, on an ~0-100 scale, with zero indicating not even motivated enough to attempt and 100 indicating one of their best performances.
Importantly, this individualization is, at some level, costly & computationally demanding, at least enough to deter wearable manufacturers from taking this approach. For this reason, the "readiness scores" that you see on most of the wearables are not individualized, are not calibrated to the individual athlete in and way & are therefore, for all intents and purposes, practically useless in guiding the training decisions for that individual athlete.
This approach of utilizing individualized readiness scores to guide the training has been, in my opinion, one of the most useful revelations to come out of my foray into machine learning. Since employing it, I see less incidence of fitness plateaus among the team, I see greater training response & I see less incidence of injury and illness. In short, this is a key paradigm shift that I see myself making use of in my coaching practice from here forward.
Hopefully this brief post provides a little more insight to the athletes following my training plans on just what is going on behind the scenes the next time you see "Numbers are indicating some significant fatigue in the system (Readiness: 0)" on your Training Peaks account :)