Block vs Random – When Improvement Isn’t Really Improvement

How do you structure your training sessions? For most coaches it probably goes something like “do <thing> for 10 minutes. Then do the <next thing>”. Which is a pretty logical way to train, and you can see people improve at <thing> over the course of the 10 minutes. Which is why it would make a really interesting article if I told you that the common sense on this topic was wrong. 🙂

To start with, let’s differentiate two terms I’ll be using.

  • Performance: How well someone performs a skill at a given point in time.
  • Learning: The change in someone’s ability to do the skill. This can not be evaluated directly, but inferred by looking at the change in someone’s performance over time.

In this plot with totally real and not made up numbers you see that the power level dots are measurements of performance in each practice, and the dashed line is an estimate of the learning based on the performance.

Around practice 90 things start to get interesting.

Measuring Learning

So, given that learning is the ability to improve over time, and we can measure performance over time, it makes sense that we would just have someone train and measure how much better they are at the end of every practice. In the plot below you can see someone attempting the same task multiple times in a row and getting better at it every time.

Though the trend towards better performance is clear, this also highlights something important: non-linearity. Even though we see an overall improvement

Miss-Measuring Learning

You show up to class. You have the students to drill. You offer corrections, and they do the drill better. Class ends, and you are happy that everyone got better at the skill than when they started. Another successful training session in the books!

This is also how learning used to be measured. In the previous section the graph shows the same task attempted multiple times in a row and the performance improvement. But is that really what we want to know? My old kayak coach had a saying “If you can really do it, I can come wake you up in the middle of the night and you can do it. If not, you can’t really do it.” 

While I’m not advocating crepuscular break-and-entry as a coaching methodology, it raises an important point. You aren’t training to do the skill at the end of class. You are training to be able to perform the skill any time it may be required.

The following experiment is a great example of the difference between performance and real learning. Participants were trained and evaluated on their ability to load exactly 30% of their body weight onto a scale, and were evaluated based on their distance from the target weight. They practiced for one day and were evaluated on their performance at the end, and then they were evaluated again two days later. The experiment had three groups, each of which received different types of feedback during practice:

  • Concurrent: Could see a scale output of how much weight they were applying, and could practice with real time feedback.
  • KR-1: Could not see the scale output, and only got the result after they stepped off.
  • KR-5: Could not see the scale output, and only got their results after every 5 attempts.

The results showed that the participants that got concurrent feedback had a far smaller margin of error when tested after practice. Which would seem to indicate that they had much better learning.

KR stands for “Knowledge of Results”, if you were wondering.

While the Concurrent feedback group was WAY better after the practice, they were by far the worst when their learning was better. This highlights something that is seen over and over in different research studies:

Your improvement in a practice has little correlation to your actual learning.

Which, as a coach, is a little bit of a gut punch. As coaches we like to see people improve, and having something “click” is extremely satisfying for both of you. But frequently the things that look like the most concrete and immediate examples of improvement are not representative of real learning. And the things that don’t look helpful are actually useful in the long term.

Blocked vs Random

And so we arrive at the headline of the article: blocked vs random. And some more definitions,

  • Blocked Practice: Doing the same thing many times in a row..
  • Random Practice: Doing something different every time.

Now, how random does something have to be to qualify as “random”. Exact definitions may differ, but the difference will be something like so:

And how is this difference important? One might assume that doing a different cut at random robs you of focus, and prevents you from really dialing down and improving the cut you are trying to improve. But, given how this article has been going so far, you have probably guessed that the reality is very counter-intuitive.

For an example, look at the following badminton study. The participants (who had never played badminton) were evaluated on their ability to perform three different serves, from the right side of the court. The blocked group trained one serve a day, rotating through the three different serves every three days. The random group did a different type, randomly, for every serve. They were evaluated after every practice for 18 practices, took a day off, and then evaluated ‘cold’ with no practice. 

(Color added by me)

The blocked group seems to have a bit of an edge in the initial learning, so if you practiced blocked training every day you would probably look at the results and conclude that there is good learning happening. But if asked to perform the skill after a day of rest (say, after traveling to a competition) they were significantly worse.

Even more interesting is the fact that the researchers performed the transfer test. This was evaluating their ability to serve from the left side of the court, which they had never done before. The block training group was stumped, performing even worse than the retention test. But the random group had no trouble at all with this switch!

Block vs Variability?

So what does this badminton study mean for HEMA? First of all, this study, on its own, shouldn’t be trusted. You can find a single study to support just about any claim you want to make, and the value in the scientific method is about repeatability.

But fortunately this result is highly repeatable, and there are mountains of experiments which show the same result across a plethora of different activities. The blocked vs random question is probably one of the most conclusively answered questions in sport science. Blocked is a far inferior way to train.

Current research into the role of variability has suggested that perhaps the “Blocked vs Random” effect is actually a “Static vs Variable” effect. And that the increase of effectiveness attributed to the random training structure is actually the increase of variability in the training. In the study referenced below, a basketball free-throw was practiced with and without variability in the starting position for the throw.

  • C – Constant Practice: Free-throw shooting from the free-throw line only
  • VFB – Variable practice, Front  Back: Free-throw shots from 2 ft in front of and 2 ft behind the free-throw line (randomly assigned in each practice session)
  • VC – Variable practice, Combination: Free-throw shots from 2 ft in front of and 2 ft behind the free-throw line as well as from the free-throw line (randomly assigned in each practice session).
  • VR – Variable practice, Random: Free-throw shots from the ‘elbow’ to the left and to the right of the key, and from the top of the key (randomly assigned in each practice session).

This shows that adding variability to training the same task in a blocked practice also leads to significant improvement. Is it variability that is actually important, and the success of random training is due to the fact it creates variability by definition? Unfortunately there haven’t really been enough studies pitting them against each other, and they both just spend their time beating up on poor ol’ block training. So for now there is no clear answer. (Other than the answer of “they are both much better than block training.)

So, Blocked Sucks?

Regardless of if it’s being compared against random or variable training methods, doing blocked training is pretty much shown to be the worst way to learn. Which is an extremely counter-intuitive result, because blocked training seems to be the most effective if you’re watching people improve over the course of a practice But doing things which are more unpredictable, and change your focus quickly, 

“Don’t train yourself to do the skill. Train yourself to figure out how to do the skill.”

– me. I made this one up myself.

Why does this happen? Well the explanation for that starts drifting into the territory of Ecological Psychology which is a whole huge ball of worms to get your head around. And it’s a good thing I have that series cued up for next time! Until then, focus on not doing the same thing over and over again in training. No matter how much you think it helps you focus.

If you want to read up more on the technical details of this, Motor Learning and Control by Magill and Anderson is a textbook on the topic which has a lot of different studies to check out. And most importantly you can find older versions for sale second-hand at very low prices.

Avatar photo
About Sean Franklin 119 Articles
Sean has a Bachelor's Degree in Mechatronic Systems Engineering, and is currently employed as a Controls Engineer. He is passionate about developing more analytical ways to view sword fighting, wishing to develop evidence based standards for protective gear and rule sets informed by tournament statistics. His martial arts history includes competitive success, medaling in international competitions for Longsword, Messer, Grappling, Rapier, and Cutting. In addition to competition Sean has been invited to instruct at a number of events across North America and Europe. For non-STEMey coaching topics Sean posts on