Tennis coaches love preparation. We watch video, study statistics, build scouting reports and look for patterns. We might tell a player that an opponent serves wide more often at 40–0, attacks second serves from a particular position or struggles when a heavy topspin ball is pushed high into their backhand.
None of that information is necessarily useless. In fact, it can be incredibly valuable.
The problem begins when preparation turns into prediction. We take something that has happened regularly in the past and start treating it as though it tells us what is going to happen next. From there, practice can quickly become a collection of if this happens, do that rules.
I recently contributed to an academic paper examining this exact problem. The paper was written as a critical response to a proposed Preparation–Perception–Action framework, or P-P-A, which argues that preparation should be recognised as an additional stage before perception and action. Our response does not argue that preparation is unimportant. Quite the opposite. Tactical planning, rehearsal, opponent analysis and psychological preparation all matter. The question we raise is whether preparation actually needs to be separated from perception and action as another stage.
From an ecological perspective, I think there is a more useful way for coaches to look at it. Preparation can shape what a player becomes sensitive to, what they intend to achieve and which opportunities they are ready to exploit. But once the match begins, the opponent does not have to follow the script.
That distinction matters enormously for how we coach tennis.
The problem with trying to predict tennis
Imagine you have an enormous dataset showing that Roger Federer historically serves wide most frequently when he is 40–0 ahead. You tell your player this before the match and they prepare themselves to return the wide serve.
So far, so good.
But your player’s preparation now changes their behaviour. Perhaps they shift their return position slightly towards the expected serve. Federer can see that. That new position provides information to Federer, and suddenly the serve down the T becomes more inviting.
The information that helped the returner prepare has changed the very interaction they were preparing for.
That is tennis.
This is why I would rather think about scouting information as something that can guide intention and attention, rather than something that predicts the next action. Knowing that an opponent tends to do something can make us more sensitive to that possibility, but it should not make us blind to everything else.
Human behaviour contains regularities, otherwise analytics and scouting would have little value. But those regularities are not guarantees. Two players can compete at the same venue in similar conditions on consecutive days and produce two very different matches because both performers are continually adapting to each other.
The challenge for the coach is therefore not to make the future more predictable. It is to prepare the player to remain adaptable when the future inevitably fails to follow the plan.
Perception and action are not two separate boxes
One of my biggest breakthroughs when moving towards ecological coaching was understanding that perception-action coupling is not a neat linear sequence.
I do not perceive the ball, stop perceiving, select a forehand from somewhere inside my brain and then execute the action. My movement continually changes what information becomes available, and the information I detect continually influences how I move.
I perceive to act, but I also act to perceive.
That relationship continues throughout the point.
This is one of the central ideas behind perception-action coupling. Contemporary research into anticipation in racket sports has repeatedly raised concerns about studies that separate perception from the action performers would normally make, because asking somebody to verbally say where a serve is going is not necessarily equivalent to actually returning that serve. A 2024 review of 115 racket-sport anticipation studies found that representative perception-action conditions remain underrepresented within the literature.
An older tennis study by Farrow and Abernethy illustrates why this matters. Players performed differently when prediction was coupled to the movement required to return serve compared with conditions where they simply gave a verbal prediction. Expertise was particularly apparent under the more naturally coupled conditions.
For coaches, the implication is relatively straightforward. If the skill eventually requires the player to perceive information while acting, we should be cautious about assuming we can train perception completely separately from the action it is supposed to regulate.
Returning serve is much more than watching the ball
Return of serve gives us perhaps the clearest example.
If somebody is serving at very high speed, waiting until the ball is halfway across the net before beginning to organise yourself is obviously problematic. Skilled players become sensitive to information available before and around contact, while continually updating action as the serve unfolds.
That information might include aspects of the ball toss, racket-arm movement, body orientation, server position and other kinematic relations. Contextual information matters too: previous serves, score, tendencies and court position can influence what becomes relevant.
The crucial distinction is that I would not coach this as a treasure hunt for one magical cue.
“Watch the toss. If it goes there, the serve is wide.”
That creates another rule.
Instead, I want the returner gradually becoming attuned to relationships between information and action. What tends to change when this server goes wide? What do you notice when they use the T serve? What information remains useful when they disguise the serve particularly well?
The academic paper makes the same broader argument. Prior information and anticipation do not need to sit outside perception-action coupling. They can shape attunement and intention while action continues to be regulated by information unfolding in the environment.
Interestingly, a systematic review of tennis first-serve anticipation found that under the most representative task conditions, evidence did not show expert players consistently initiating lateral movement towards the eventual serve direction before racket contact. The authors instead highlighted regulation from information unfolding around the server’s action and the first moments of ball flight. That does not make anticipation irrelevant. It challenges an overly simple picture of expert players merely guessing correctly earlier.
A ball machine can vary the ball without varying the information
This has significant implications for practice design.
Take a ball machine. Modern ball machines are becoming incredibly sophisticated. They can change speed, spin, height, placement and sequence. Some newer systems can even move around the court.
But ask a different question: where is the information that normally creates the shot?
The ball still generally emerges from a machine. There is no opponent organising themselves around the previous ball. There is no racket preparation, changing balance, body orientation or contact point for me to become sensitive to.
That changes my perception-action coupling.
Research cited in our paper compared tennis returns against a ball machine and a real server and found differences in movement initiation and swing timing. Removing opponent-derived information altered how players organised their actions.
That does not mean a ball machine is useless. I have said this many times in my other articles about practice design: the question is not whether learning can happen. The question is what we expect that learning to transfer to.
You can get lots of useful hitting volume from a ball machine. You might explore racket movement, calibrate after an injury, experience a particular ball or bring attention to something very specific. What I would challenge is the assumption that hitting 100 machine-fed returns is equivalent to developing the perceptual skill required to return 100 serves generated by another human.
They are different informational problems.
The missing player at the other end matters
There is another layer here that I think coaches often overlook.
Imagine I hit a forehand that completely jams my opponent. They stumble backwards, make contact late and produce a defensive ball.
My first action has changed my opponent.
That change now provides information for my next action. Their balance, contact point, position and movement create new opportunities for me.
Perhaps I can move forward.
Perhaps space has opened on the other side.
Perhaps their compromised contact means I can take more risk.
This is what I mean when I talk about nested affordances. One opportunity for action can open another. My behaviour helps create the information that subsequently guides me.
A ball machine cannot really give me that relationship. Even a highly sophisticated machine can send the next ball somewhere different, but it has not become off balance because I jammed it. It has not been pulled three metres outside the singles line by my previous forehand. It has not panicked because I suddenly moved forward.
The opponent is not simply another object on the court.
They are part of the information.
This is why I increasingly think of tennis as an ongoing process of perception and adaptation. Every adaptation changes what becomes possible to perceive next.
Preparation can operate over seconds, days and years
This brings us to one of the heavier academic arguments in the paper: timescales.
It sounds complicated, but coaches experience it constantly.
Some perception-action relationships unfold incredibly quickly. I pick up information from the opponent, ball and environment and reorganise my movement over fractions of a second.
Others develop much more slowly.
I might watch an opponent on Monday and notice something about how they defend the backhand corner. On Tuesday, I practise against somebody recreating aspects of that problem. On Friday, I finally compete against the opponent.
Where exactly did preparation finish and perception begin?
That is the problem with trying to turn skilled behaviour into three neat boxes of prepare → perceive → act.
Our paper argues that preparation can instead be understood as part of the organisation of perception-action relations across multiple timescales. Film study, tactical rehearsal and previous competition might gradually change what the player becomes attuned to. The real-time action on Friday is occurring over seconds and milliseconds, but it is being influenced by a history that has developed over much longer periods.
Personally, I find perception and adaptation a useful way of thinking about this. I perceive something and adapt. That adaptation changes the environment, which gives me something new to perceive. Multiple processes can be unfolding across different timescales rather than one perception-action loop beginning and ending every time I hit a forehand.
Scouting reports are useful — until they become instructions
This changes how I think about opponent analysis.
Consider the famous tactical battle between Rafael Nadal and Roger Federer. For periods of their rivalry, Nadal’s heavy left-handed topspin into Federer’s one-handed backhand created an obvious tactical problem. That historical information would be extremely useful when preparing another player.
We might develop an intention to expose the same area.
We might start searching for opportunities to get the ball high onto the backhand.
But Federer was not a static problem.
He adapted.
That is the bit a scouting report can never completely solve for us.
If the opponent starts taking the ball earlier, changing position or finding a different response, blindly following the scouting report becomes a disadvantage. Preparation should help the player recognise opportunities, not stop them recognising that the opportunity has disappeared.
The paper describes this ecologically as preparation shaping attunement, calibration, intention and sensitivity to affordances, rather than functioning as a command generated before the performance begins.
That is how I would use video with a player. I am not trying to tell them what will happen. I am trying to broaden what they might become sensitive to when it does.
Deception shows why tennis is an interaction
I saw a fascinating example of this courtside at Wimbledon in 2026.
Daniil Medvedev was playing big-serving German Jan-Lennard Struff in the third round. Struff eventually won the match in straight sets.
What fascinated me was the return game. Medvedev was dealing with an enormous serve and at times appeared to be searching for a way of getting a better read on it. One behaviour I noticed was Medvedev initially showing one return position and then changing his position as the service action developed.
Think about what that does.
The returner is no longer simply trying to read the server. He is giving the server information.
Leave a large gap towards one side and perhaps the server perceives an opportunity to attack it. Move as the toss develops and suddenly the affordance changes. Now the server may have to maintain the original action, disguise the intention or adapt the serve very late.
At the same time, the server is trying to stop Medvedev getting useful information too.
Both players are perceiving.
Both players are acting.
Both players are trying to influence what the other person perceives.
That is a much richer explanation of anticipation than one player simply trying to guess where a serve will land.
It also helps explain skilled intentionality. The server can have an intention, but that intention does not necessarily have to be a rigid command that is executed regardless of what happens. Highly skilled performers can adapt movement as the informational landscape changes.
A plan is useful until you get punched in the face
I often use a boxing analogy when discussing tactics.
All great plans are great until you get punched in the face.
Tennis coaching is full of beautifully organised plans. If the ball lands here, attack there. If they serve here, return there. If you receive this ball, play this pattern.
Then the opponent arrives.
And unfortunately, the opponent has not read your practice plan.
They are actively trying to stop you playing your preferred forehand. They want to disrupt your timing, move you into uncomfortable positions and take away the opportunities you have practised.
This is where knowing what to do and being skilled enough to actually do it become very different things.
Our paper makes this distinction when discussing tactical preparation. Knowing an opponent frequently attacks a particular channel is useful, but possessing that knowledge is not equivalent to being perceptually attuned and physically capable of regulating action successfully when the situation appears.
In coaching, that means I am increasingly reluctant to give players highly specific patterns that prescribe exactly what should happen next.
Instead, I want them experiencing multiple solutions to similar problems.
How else could you attack this second serve?
What changes if they recover more quickly?
What if the ball arrives higher?
What if they take away the space you expected?
The aim is not to abandon tactics.
It is to develop tactics that remain adaptable.
For me, it is not about having a plan to win. It is about finding a way to win.
The problem with “if this, then that” tennis
Years ago, I worked with a coach who used something called the Ninja Rules. He is a good friend of mine, so if he ever reads this, apologies in advance.
The idea was grounded in sensible high-percentage tennis. If a ball landed deep cross-court, for example, the statistically safer response might be another deep cross-court ball. Court geometry, net height and recovery position could all support that decision.
The problem appears when a useful probability becomes a law.
Ball lands here = you must hit there.
Two balls can land in almost exactly the same location and afford completely different actions. On one, I am off balance and defending. On another, I arrive early with my opponent badly positioned and have the action capacity to attack down the line.
Same bounce location.
Different problem.
This is where I think we need to be very careful with prediction-based tactical coaching. High-percentage information can educate intention, but it should not overwrite perception.
Highly skilled players regularly exploit opportunities that would violate a simplistic tactical rule because the opportunity available to that player in that moment is different.
The ecology matters.
Don’t train players to see less
This has probably been one of the biggest changes in my own coaching.
Traditional tennis instruction often narrows perception. Watch the ball. Watch the spin. Watch the height. Look at where it is going to bounce.
Those things matter, but the tennis player is interacting with much more than a ball.
There is the opponent.
Your position.
Their position.
The court surface.
Wind.
Score.
Time available.
Space.
Your own action capabilities.
The changing relationships between all of them.
From an ecological perspective, expertise is not simply becoming better at consciously processing a longer checklist. It is becoming more attuned to useful information and informational relationships that regulate action.
So my practice design increasingly tries to develop greater sensitivity rather than reduce the game to one cue.
I may ask, “What did you notice about your opponent?” or “What changed before you decided to move forward?” But I am not interested in turning every rally into an oral examination where players shout out cues as they play.
The environment should do most of the teaching.
The question is whether I have designed a task in which being sensitive to the relevant information actually helps the player succeed.
How should coaches practise anticipation?
This is where Representative Learning Design becomes extremely important.
If I want a returner to become sensitive to information from a server, I need a server.
That sounds almost ridiculously obvious, but much of sport practice and anticipation research has historically removed the action component or altered the information available. The 2024 racket-sport review I mentioned earlier concluded that representative perception-action conditions remain comparatively underused in anticipation research.
The same issue applies to coaching.
Showing a player isolated video clips and asking them where the serve is going may test something interesting, but it is not identical to standing on a baseline with a racket and having to regulate your movement against a live serve.
Video can still be useful.
Statistics can still be useful.
Scouting can still be useful.
The ecological question is what happens afterwards.
Can I bring that information into a representative environment where the player has to detect it while acting?
If an opponent disguises serve particularly well, perhaps I expose my player to multiple servers. Maybe I ask servers to vary position, intention and disguise. Perhaps I manipulate the returner’s starting position. Maybe we add scoring consequences that encourage servers to exploit space and returners to adapt.
Now preparation is helping reorganise perception-action rather than simply giving the player another fact to remember.
Three things I would change in tennis preparation tomorrow
The first thing I would change is how we use predictions. Do not abandon analytics or opponent tendencies, but stop presenting probabilities as certainties. “This player often serves wide here” should start a search, not end one. Allow that information to guide intention and attention while keeping the player sensitive to what is actually unfolding.
The second change would be to design for adaptability. Instead of rehearsing one pattern until the player knows exactly what comes next, let similar tennis problems emerge in different ways. Vary opponents, positioning, scoring and informational constraints. Give players reasons to find more than one solution.
The third would be to design practice so players receive more useful information, not less. If the opponent provides critical information for the skill, think very carefully before removing them. If the player eventually needs to perceive and act simultaneously, preserve that coupling wherever the learning objective allows it.
This does not mean every practice has to be a full tennis match. As I have written in previous articles on Representative Learning Design, representativeness is a scale. Coaches can simplify tasks. We can reduce complexity and occasionally decouple parts of a skill when there is a clear reason to do so.
The important question is what information we remove, why we remove it and how quickly we intend to put it back.
Preparation matters — but perhaps not in the way we thought
The academic paper that inspired this article does not argue that athletes should stop preparing. It argues almost the opposite.
Preparation matters because it can change the performer.
Watching video can alter what becomes relevant. Practising against certain tactical problems can change attunement. Previous matches can recalibrate action. A scouting report can orient intention. Years of experience can shape what information a player becomes sensitive to in fractions of a second.
What our paper questions is whether all of that requires a separate Preparation → Perception → Action stage. The ecological argument is that many of these phenomena can already be understood through the organisation of perception-action coupling across different timescales. The paper concludes that the proposed P-P-A framework does not yet clearly provide distinct mechanisms, predictions or applied advantages beyond those ecological explanations.
For most tennis coaches, however, the academic disagreement is probably not the most important part.
The practical question is.
Are you preparing players to reproduce what you think will happen?
Or are you preparing them to adapt to what actually happens?
Those are very different things.
The best tennis players do not simply execute a collection of pre-programmed responses. They continually become sensitive to changing information, act upon opportunities, change the environment through their own actions and adapt again when the opponent responds.
Sometimes they perceive in order to act.
Sometimes they act to create something new to perceive.
So when designing your next tennis session, I would stop asking only, “What pattern do I want my player to learn?”
Ask instead:
“What information does my player need to become sensitive to, and does my practice actually give them the opportunity to use it?”
That question will probably tell you far more about whether your preparation is preparing them for tennis.
Practical next step
This is also where my Tennis Practice Assessment Scorecard can help. Take one practice you currently use for return, anticipation or tactical preparation and assess what information is retained. If there is no opponent, no consequence, no relationship between one action and the next, or the player already knows exactly what ball is coming, you do not necessarily have to throw the practice away.
Change one thing.
Introduce an opponent. Make the ball emerge from a real action. Add a meaningful decision. Allow the opponent to adapt.
Then watch how your player changes.
The goal is not a perfect score. It is creating practice environments with enough relevant information to give learning a better chance of transferring to the match court.
