(Im)perfect Practice Makes Perfect Preparation

A cliche I grew up on was that “practice makes perfect”. A really basic, yet probably useful, idiom that was used to motivate my less enthusiastic peers to get to footy training and put in the hard yards. As I got a bit older and training standards were held to a higher level, this eventually evolved into the idea that it was actually “perfect practice makes perfect preparation”. This was something that generally only came up when training was going 'bad'ly; the nights where, for no obvious reason, basic skills weren’t as crispy, or we had just come off a loss on the weekend and the coach had decided to turn the disciplinarian dial up a few clicks. A drill might get stopped because we weren’t doing it well enough, followed by some fitness, another drill, or the obligatory reminder that everyone needed to pull their heads in and focus.

When nobody made mistakes, on the other hand, things were inherently seen as good. This was the idealised version of “perfect practice” in motion. Coaches were happy, players were happy, and everyone left training reasonably confident that things would go well on the weekend.

I’ve become increasingly interested in challenging this idea because I think it carries two assumptions that aren’t necessarily helpful.

1. The first is that the goal of practice should be perfection. By perfection, I mean the way it is commonly represented in coaching: high rates of successful execution, minimal mistakes, and athletes performing skills and tactics in the way the coach intended.

2. The second assumption is that this version of perfect practice is what best prepares athletes for competition.

There are a few reasons I wanted to write about this. I recently attended a coaching development course and one of the coaches there announced, in front of a room full of people, that my blog posts were one of his favourite resources on training. I haven’t written anything in about a year, so shout outs Vas, I hope you’re reading. I’ve also barely written anything recently without the help of AI, so a more regular creative outlet probably isn’t a bad way of slowing down the inevitable brainrot as LLMs increasingly take over the creative and critical parts of our lives.

The more important reason, though, is related to coaching. Over roughly the last six months I’ve become a lot more comfortable allowing athletes to fail without immediately feeling the need to intervene. I’ve always enjoyed solving problems, and it’s probably one of the major reasons I became a coach in the first place. If an athlete couldn’t do something, I wanted to understand why. If a movement didn’t look right, I wanted to fix it. If an exercise was too difficult, I’d find a regression. If somebody asked for feedback after a repetition, I’d give it to them. At the time, this felt like coaching. The athlete had a problem, I solved it, they completed the task successfully, and we moved on.

I’ve become less convinced that this process was always helping as much as I thought it was. By continually identifying the problem, interpreting what caused it and then providing the solution, I was sometimes completing a fairly significant part of the learning process for the athlete. I was setting them up for successful practice performance without necessarily giving them much opportunity to deal with the discomfort of not immediately knowing the answer. I would leave training pleased that I’d found a way to make everything work, then arrive at competition frustrated that the same athletes weren’t particularly resilient when things stopped working. That contradiction is really where this article starts.

Competition is inherently imperfect

The goal of practice is ultimately to prepare athletes for competition, which means coaches need a pretty strong understanding of both what competition demands and how training can prepare athletes to meet those demands. Sports performance is often broken down into physical, technical and tactical qualities, although I think psychology and the environment need to sit alongside those as factors that influence how those qualities are eventually expressed. The balance obviously changes from sport to sport, but performance itself is almost always an interaction between multiple things rather than a neat expression of one isolated quality.

Competition also contains uncertainty. There is variability, pressure, fatigue, environmental change, unfamiliar situations and, in many sports, an opponent actively trying to stop you from succeeding. Even in events where the perceptual and strategic demands are relatively low, the athlete still has to express performance in a context that is never completely predictable.

This is where I think the traditional interpretation of perfect practice starts to become problematic. If we systematically remove disturbances from training because they reduce the likelihood of successful execution, athletes can become very good at performing under conditions that rarely exist in competition. That doesn’t mean training should deliberately be messy or that mistakes are automatically valuable. There is nothing particularly clever about making training harder for the sake of it.

It does mean that errors, variability and difficulty shouldn’t automatically be interpreted as evidence that training is going poorly, because sometimes the thing that makes a practice task less successful in the short term is the same thing that makes it more useful later.

Perfect practice may require imperfect design

When an athlete fails at something there is a pretty understandable urge to make the next repetition better. Coaches can simplify the task, provide another cue, demonstrate the movement, regress the exercise, give the athlete more information or just tell them exactly what went wrong. All of those things can be useful, and I’m not particularly interested in arguing that coaches should suddenly stop coaching. The question is more whether intervention is always necessary, particularly when the athlete may be capable of finding a useful solution themselves.

If every error is immediately followed by a correction, the coach gradually becomes part of the athlete’s regulatory system. The athlete performs, the coach interprets what happened, provides the solution, and the athlete tries again. Competition removes most of that support. Athletes have to recognise when something isn’t working, interpret what they are feeling or seeing, and make adjustments while performance continues. That ability is difficult to develop if practice rarely asks them to do it.

This is probably the biggest change I’ve made in my own coaching. I don’t mean deliberately setting athletes impossible tasks or watching them repeat rubbish movement because I’ve suddenly decided errors are good. There is nothing inherently developmental about being bad at something. The value comes when the athlete can experience the consequence of an unsuccessful solution and has enough useful information available to search for something better. Sometimes I should intervene immediately. Sometimes the better coaching decision is probably to let the problem survive another repetition or two.

To me, the difference between providing an athlete the solution and challenging the athlete to provide the solution themselves becomes the difference between 'commentating' and 'coaching'. Particularly as a track coach, in which it can be easy to set yourself up stationary next to the track and watch your athletes run, I've found myself standing idle commentating athletes runs to them. Perhaps this has been helpful, perhaps not. Ultimately, it matters less that I have my own understanding of an athlete's performance than the athlete becoming in-tune with how their body is moving and evaluating what that means for performance outcomes.

Representative learning

Representative learning design gives some theoretical backing to this idea. At its most basic level, representative practice tries to preserve some of the important information and constraints that exist in competition. Athletes don’t simply learn movements in isolation. In most sports, they learn how to organise movement in response to information.

This is easiest to see in team sport. A young footballer might become exceptionally good at dribbling through a predetermined series of cones. The execution can become incredibly clean, the movement pattern is known before the repetition starts and mistakes may eventually become quite rare. Once defenders and teammates are added, however, the technical action still matters but the problem changes completely. The player now has to detect space, perceive pressure, monitor the movement of others and select an appropriate action. The movement is being regulated by information that wasn’t present in the cone drill.

This was something I spent quite a bit of time thinking about in an assessment during my Masters on perceptual expertise. One of the central ideas was that experts aren’t necessarily processing more information than everyone else. Through experience, they become better at recognising which information matters. They become attuned to useful cues, recognise familiar patterns earlier and filter out information that is less relevant.

That obviously relies on exposure, but it also relies on the training environment containing some of the information you eventually want the athlete to become sensitive to. If most of practice is predetermined, coach-directed and organised in a way where the solution is known before the repetition begins, the athlete has much less reason to develop that perceptual component. The movement might improve, but the relationship between information and action is being trained to a much smaller degree.

I created the framework below for this assessment to illustrate these ideas and found it useful when thinking about the different kinds of practice we use.

	Low information	High information
High action	Technical practice: isolated skill repetition, predetermined drills, running drills. Movement is present but decision information is limited.	Representative practice: small-sided games, live scenarios, match simulation. Perception and action are tightly coupled.
Low action	Low representativeness: lecture-style coaching, static instruction, demonstration.	Perceptual practice: video decision training, tactical analysis, pattern recognition tasks.

The point of the framework isn’t that everything should sit in the top-right box. Technical drills can be useful. Video work can be useful. Removing complexity can be useful. What matters is understanding what has been removed when a task is simplified, and whether that matters for the quality you are trying to develop.

There is also a relationship here with how expertise develops more broadly. Sporting situations rarely repeat exactly, so expert performers don’t appear to simply memorise thousands of individual scenarios and retrieve an identical answer when one appears again. They become sensitive to families of patterns and learn which relationships or cues remain useful as the surface details change. If training continually manufactures the same problem in exactly the same way, athletes can become very good at solving that version of the problem, but competition is generally less accommodating.

The table above was inspired by watching football (soccer) private coaches teaching children (often less than 10 years old) footballing skills in pre-determined ways; dribble through cones, no defensive pressure, etc. It's a very pretty training set up. On the contrary, if we were to imagine the talent pathways of more successful footballing nations, beyond the prevalence of academy systems, would we see more representative training with developmental athletes? Kicking a raggedy ball barefoot in a game on the streets of Rio de Janeiro may not be 'coaching' but the environment is set to coach athletes to deal with the demands of the sport in all facets compared to what I often see in modern, privatised sports training environments.

“I cannot teach anybody anything. I can only make them think.” - Socrates

Sprinting is an interesting test case

Most of this argument is relatively easy to make in football, rugby or basketball because the informational demands are obvious. The 100m is different. Sprinting is a relatively closed skill with comparatively low perceptual and tactical demands. Once the gun goes, the physical objective is extremely stable.

I still think sprinting provides one of the more interesting examples of this problem because a closed skill isn’t necessarily performed in a completely closed environment. Sprint training can become incredibly sterile, and there are good reasons for that. Speed development depends heavily on producing very high outputs, so coaches naturally control variables that allow athletes to run fast. Recovery is long enough to preserve quality, surfaces are selected carefully, technical feedback is available and biomechanical analysis can become quite detailed.

None of that is inherently problematic. If the goal of the session is maximum velocity, the athlete actually needs to run very fast. Reducing recovery just to make the session more difficult would probably compromise the quality you were trying to develop. Controlled environments allow us to isolate and overload physical qualities, which is a pretty important part of training.

The potential issue is when athletes become dependent on those conditions to express them. Competition might involve an unfamiliar track, different blocks, poor weather, wind, a disrupted warm-up or a starter whose rhythm feels different to the one the athlete experiences at training. Championships add call rooms, delays, multiple rounds and uncertain recovery durations. Seven other athletes are moving alongside you and a false start can alter the rhythm of the race before it has even begun. The physical problem is broadly the same, but the context in which the athlete has to solve it is not.

I would point to the 2024 Olympic Men's 100m Final as the perfect example of this idea. The pinnacle of the sport. The 8 fastest men in the world. 4 years until this opportunity presents itself again. Thousands of hours spent training. Millions of dollars invested in the process by national sporting bodies, athletes, sponsors, etc. All these factors are compressed into 9 seconds of sprinting in front of a full stadium.

3 minutes are taken for the athletes to run out in front of the crowd. This is followed by 3 minutes of a stadium DJ playing tense electronic music. The athletes have never experienced this before a race - certainly not for a race of this magnitude. "We have never waited this long before the start of the race", says the commentator. Another 2 minutes go by until the gun finally goes and Noah Lyles ultimately wins by thousandths of a second. I believe he won not just due to the manner in which he raced, but in the manner he handled the previous 8 minutes. This was not in the script of what a 100m race should be like, yet it was the ultimate test for the sport's greatest athletes.

https://www.youtube.com/watch?v=7Xnr805bm4E&t=710s

This has made me increasingly interested in the idea of robustness. Improving the highest level of performance an athlete can produce is obviously important, but so is improving the reliability with which that performance can be expressed when the conditions aren’t ideal.

The practical question that follows from this is probably the most useful one in the whole article: how much instability can we introduce while preserving the physical quality we are trying to train?

I don’t think instability should be maximised. If a maximum velocity session is manipulated so heavily that nobody can run particularly fast, the variability has become more important than the quality the session was supposed to develop. The opposite can also be true. If every variable is controlled to the point where an athlete only performs well under one exact set of circumstances, the environment may have become overly sterile.

I increasingly think of this as a continuum: Sterility ← controlled instability → chaos

At one end, the environment is highly controlled and optimised for successful performance. At the other, there is so much uncertainty, fatigue or disturbance that the target quality disappears altogether. Somewhere in between is a level of instability that allows the athlete to continue expressing the physical quality while being exposed to a slightly broader range of conditions. Where that point sits is obviously going to depend on the athlete, the training phase and the purpose of the session.

Starts provide a simple example. If every acceleration uses the same coach, the same cadence and roughly the same set duration, the stimulus can become fairly predictable. Varying the hold time changes almost nothing about the physical quality being trained, but removes some predictability. Racing another athlete does something similar. The task is still acceleration, but now there is peripheral movement and potentially a different level of competitive arousal. Different tracks, different lanes, environmental conditions, competition-style waiting periods and occasional variation in recovery can all introduce small amounts of instability without fundamentally changing the physical task.

The important thing is that these manipulations don’t need to be dramatic.

'Old School' coaches have this element of coaching mastered with a million tricks up their sleeve in a way which didn't put themselves at the centre of learning and instead continued to challenge thei athlete's ability to perform. Changing block settings behind athletes' backs; shit-talking between repetitions, coughing between 'set' and 'go'; telling an athlete they false started when they actually did not and certainly not cancelling training or competitions when it's wet or head winds forecasted. Furthermore, 'Old School' coaches understand the value of integrating races as a part of the training process as it exposes not only the technical or physiological facets that training needs to continue to address but also an athlete's ability to handle the environmental and psychological stressors that only racing as the ability to expose.

The aim isn’t to make training harder. It is to broaden the range of conditions under which the athlete can still access the quality being developed.

Feedback can make practice sterile too

Coach feedback fits into this discussion as well. Immediate feedback can be extremely useful, particularly when an athlete is learning something new or when the coach has access to information the athlete can’t reasonably identify themselves. The issue, again, is whether it becomes the default response to every repetition.

If every rep is followed by external feedback, the athlete has relatively little reason to develop their own interpretation of performance. This is something I’ve deliberately experimented with more recently. Rather than immediately telling an athlete what I saw, I might allow several repetitions to occur first, ask what they experienced, or simply leave something alone for long enough to see whether they adjust without intervention.

Sometimes they fix it, sometimes they don’t, and sometimes their interpretation is completely different from mine. All three outcomes give me useful information. For me, this has probably been one of the harder coaching habits to change because intervention feels productive. Giving feedback feels like coaching and finding the right regression feels like coaching. Allowing somebody enough space to work something out can feel like doing less, even when it might be the better decision.

Not every session needs to resemble competition

There is an easy way to take all of this too far, because not every training session needs uncertainty and not every drill needs to contain an opponent, fatigue or competitive pressure. There are perfectly good reasons to simplify practice. Some sessions exist to develop physical capacity, some expose athletes to maximal outputs, some refine a technical quality and others provide variability or competitive pressure. Controlled and representative training aren’t opposing philosophies. The programme as a whole needs to prepare the athlete for competition, rather than every individual exercise needing to replicate it.

Where I think coaches can get into trouble is when successful practice performance becomes confused with successful preparation. A technically clean drill can transfer poorly. A messy session can still be useful. An athlete can complete something successfully because the environment has made the answer obvious, while another can fail at something and still leave training better prepared for the next time they encounter the same problem.

This is probably the biggest thing I’ve changed my mind about. For a long time I judged too much of my coaching by what training looked like while I was standing there. If the athletes were successful, the session felt successful, and if they weren’t I felt responsible for fixing it. I’m now more interested in what happens later, particularly whether the qualities we are developing remain accessible without exactly the same cues, environment and feedback.

I don’t think “perfect practice makes perfect preparation” necessarily needs to disappear. I think the definition of perfect practice needs to change.

Sometimes the best environment will be tightly controlled because the athlete needs access to a very specific physical or technical stimulus. At other times it should contain more variability, less feedback and enough uncertainty that the athlete has to adapt. The difficult part is understanding which one the athlete actually needs.

From that perspective, imperfection isn’t something that needs to be eliminated from training. It is another variable that can be manipulated.

Perfect practice isn’t the absence of errors or variability, but the deliberate design of training that develops performance while also giving the athlete enough exposure to the kinds of disturbances that may affect how that performance is eventually expressed.

That, to me, feels like a more useful interpretation of perfect preparation.

(Im)perfect practice makes perfect preparation.

Leave a comment on my instagram (@jacked.speed) if you have any thoughts on this blog post and thanks for reading.

Cheers,

Jack