The Most Common AI Mistakes Sports Coaches Make

Catapult AI calculates optimal training loads based on movement data and recovery markers, but it cannot see that your athlete is grinding through personal stress or dealing with a nagging pain they have not reported. You end up pushing someone who is already broken.

The fix

Compare Catapult's recommendation against your own observation of the athlete's mood, focus, movement quality, and what they tell you in person before finalising that day's load.

Hudl AI flags movements that statistically correlate with injury in large datasets, but it misses the context that your athlete has always moved that way and plays their best football from that position. You bench a healthy player based on a statistical phantom.

The fix

When Hudl flags an injury risk, watch the video yourself and ask the athlete whether that movement feels normal or wrong to them before restricting their play.

WHOOP shows heart rate variability and sleep data, but an athlete can score poorly and still be mentally sharp and ready to train hard. You miss the chance to push someone through a minor dip and instead coddle them when they need challenge.

The fix

Treat WHOOP scores as one input, not permission or prohibition. Always combine the data with your observation of their sprint speed, decision-making, and what they say about how they feel.

ChatGPT can generate a well-structured periodisation plan, but it does not know your sport's specific demands, your team's fixtures, or your squad's weaknesses. You get a generic plan that looks professional but misses what your athletes actually need.

The fix

Use ChatGPT as a starting template only. Rewrite every session to target the movement patterns, decision-making pressures, and fatigue states your athletes face in real matches.

Second Spectrum shows where players are on the field and how often they occupy certain zones, but it cannot tell you why an athlete made that movement or whether they made the right decision. You mistake location data for tactical understanding.

The fix

Use Second Spectrum to spot patterns you might have missed, then watch the video yourself and ask the player why they moved that way before drawing any coaching conclusion.

Catapult says today is a 'high intensity' day so you program sprints, but the athlete does not know why this matters for their position or upcoming opponent. They lose ownership of their own development and start resenting the workload.

The fix

Tell the athlete the specific reason for today's training before they start. Connect Catapult's data to what they need to improve in actual matches.

You spend five minutes looking at WHOOP scores and Catapult metrics before training, then spend zero minutes noticing the athlete's posture, eye contact, and energy in person. You miss the stuff that actually matters for their performance.

The fix

Spend your first five minutes of each training session watching athletes move and talking to them, before you look at any data.

Hudl AI shows that an athlete had lower 'intensity' in the second half, so you tell them they lack mental toughness. But you did not watch the game and do not know they made two crucial assists that created the play but did not get counted in AI metrics.

The fix

Watch the footage yourself before any conversation about performance. Use data to prompt questions, not to draw conclusions about the athlete's character or effort.

Second Spectrum shows the athlete was 'out of position' on a play where the opposition scored, but your eye saw they made a correct decision given the information they had in real time. You damage trust by citing data that was actually incomplete.

The fix

Ask the athlete about the decision in the moment they made it before consulting positional data. Combine both to understand what went wrong.

You show an athlete their Catapult metrics, they get excited about improving their numbers, and you think they are now internally driven. But six weeks later they are injured and burnt out because they chased metrics, not performance.

The fix

Notice whether the athlete is motivated by the competition and development itself, or just by the number on the screen. If it is the latter, redirect them to what matters.

You have ten years of experience managing fatigue across a season, but ChatGPT can generate a plausible-looking plan in seconds. You second-guess your own judgment and follow the AI instead of trusting what you have learned in the field.

The fix

When ChatGPT output conflicts with your experience, keep your version. Use AI to speed up work you already know how to do, not to replace the judgement you built over years.

Catapult calculates 'high intensity efforts' and 'acceleration load' based on accelerometer data, but these are interpretations, not direct measurement. You treat them as fact and make decisions that affect athlete wellbeing based on algorithm choices you do not understand.

The fix

Learn what each AI metric actually measures and what it cannot measure. Know the margin of error before you use it to make a decision about an athlete's training.

You see an athlete is moving stiffly and probably should not play, but Catapult shows 'recovery score 87' and you think the data knows better. You do not trust your own judgment about what you are seeing.

The fix

Your direct observation of movement quality, pain, and readiness is data too. It is just not quantified. Weight it equally with what the tool shows.

Every athlete responds differently to training load, sleep deprivation, emotional stress, and competition. WHOOP gives you one standard score, but you know Sarah needs less sleep than Marcus and comes back hard after losses. You use the tool to replace that knowledge.

The fix

Build your understanding of each athlete's actual patterns. Use AI to confirm or question what you already know about them, never to replace it.

Second Spectrum shows a defender spent more time in their own box than in midfield, so you conclude they are not pressing high enough. But in your league, defensive shape wins. You misuse positional data and criticise correct play.

The fix

Always interpret AI analysis through the lens of your actual sport and competition. A metric that looks bad in isolation might be exactly right for your game plan.