A boxer throws ten thousand jabs and nobody worries. Put a loaded bar overhead through a similar range and it lands on every list of exercises to avoid. The range is comparable. The verdict is opposite. Which tells you the list was never really about the movement.
Every few months a list circulates. Exercises you should never do. Upright rows, behind-the-neck presses, deep squats, sit-ups, leg extensions, good mornings. The list barely changes and it is never quite right, because a list of movements cannot answer a question about a person.
I have coached boxing and athletics since I was eighteen. In that time I have watched people get hurt doing things nobody would call dangerous, and watched others do everything on the forbidden list for decades without incident. The difference was never the exercise.
Take the comparison in the title seriously for a moment.
A boxer’s jab takes the shoulder into a fast, repeated, near-full-range extension under load — the load being the arm itself plus the acceleration. A competitive amateur throws hundreds in a session and tens of thousands across a career. Nobody circulates a list warning against it.
An overhead press takes the shoulder through comparable range, more slowly, under an external load. It appears on every avoid-this list ever written.
The difference is not the joint angle. It is that the boxer arrived at that range through years of graded exposure, with the surrounding tissue adapted to it, controlling the movement themselves. The person under the bar may have arrived at the same range this afternoon, borrowing a rep count from somebody on a screen.
The movement is the same. The event is not. And the event is what injures people.
Racket sports make the same point at the other end. A club tennis player lunges deeper, faster and more often than almost anyone does in a gym, off balance, repeatedly, for two hours. If lunge depth were inherently hazardous, tennis would be a public health problem.
You will have heard that the knees must not travel past the toes in a squat. It is probably the single most repeated coaching cue in existence.
A 2023 review in the Journal of Clinical Medicine traced the rule to studies from the 1970s and 80s and described its survival as a myth. The problem is what deliberately restricting anterior knee travel actually does: it shifts substantial peak torque onto the hip and lumbar spine, which the original work largely ignored.
Their conclusion is unusually direct. Restricting that natural movement is likely not an effective strategy for healthy trained individuals, and outside knee rehabilitation it should not be applied as a general rule. In other words: a cue designed to protect one joint has been quietly loading two others for fifty years.
And restricting range costs you something measurable. A 2021 meta-analysis found full range of motion training produced greater strength gains and substantially greater lower-limb hypertrophy than partial range — effect sizes of 0.56 and 0.88 respectively. The safety cue has an adaptation price attached, and nobody mentions it.
I want to be careful not to overcorrect, because that is how coaching fashions work — a rule is debunked and the opposite becomes gospel.
Squatting deeper does increase knee flexion moment, and a 2013 study found that the load reductions people typically make when going deeper are not enough to offset the increase. Depth is not free. It is a load decision.
Which is the actual point. Not forbidden, not mandatory — a variable to be managed against the person in front of you. That is a harder thing to put on a list, which is precisely why lists persist.
They fail for the same reason movement screening fails as injury prediction, and I have written about that at some length after discovering it undermined a position I had held for decades.
A meta-analysis of thirty prospective studies covering 3,404 runners tested over a hundred biomechanical and musculoskeletal variables and found essentially nothing predictive. Screening tools do not predict injury either. What does predict it, consistently, is previous injury — a fact about the person, not about the exercise.
A forbidden-exercise list is the same category error wearing different clothes. It describes the movement and ignores everything that determines whether the movement is a problem: training history, tissue tolerance, recent load, technical control, and what else happened that week.
Elite powerlifters squatting at 90% of one-rep max generate average peak patellofemoral joint contact forces of around 26.7 times bodyweight, and tibiofemoral forces above 23 times bodyweight. Those joints tolerate it because two decades of graded loading built tissue that does.
The same movement, the same joint angles, performed by someone who sat down for eleven years, is not a smaller version of that event. It is a different event with the same name.
Here is the mechanism that actually produces the injuries the lists are trying to prevent.
A conditioned twenty-four-year-old athlete demonstrates something impressive. The movement is genuinely appropriate for them — they have earned the range, the load and the speed over years. It is filmed in ninety seconds with no history attached.
A deconditioned enthusiast attempts it on Tuesday. When it goes wrong, the exercise gets blamed and joins the list.
Nobody involved has done anything unreasonable. The athlete is not lying. The viewer is not stupid. What is missing is the only variable that mattered, which is the twelve years of preparation that made the demonstration unremarkable for the person demonstrating.
Not a list of banned movements. Observable markers, during the movement, in this person, today.
The bent-over row is the clearest example of that fourth point. It is a back exercise. Watch closely and a great many people are performing a biceps curl with the elbow drifting forward, the shoulder blade never moving, and the spine gradually rounding to make up the difference. The name on the programme says one thing. The movement being executed is another. That is not a dangerous exercise — it is a different exercise being performed under a false name, and the risk sits in the mismatch.
Most of them self-select out.
A deconditioned person cannot get into a deep overhead squat. The range is not available to them, so the loaded version of the problem never arises. They will not perform a heavy behind-the-neck press because they cannot reach the position. The exercises that genuinely hurt beginners are almost never the exotic ones on the list — they are ordinary movements performed with more load, more volume or more fatigue than the person had earned.
And people train what they are already good at. This is the quiet reason so many programmes drift: we gravitate toward movements that feel competent and away from the ones that expose us. The exercise somebody avoids is usually more informative than the one they are worried about.
The observable markers above are thirty-seven years of coaching, not a validated screen. I have no trial showing that watching for them reduces injury, and given what the prospective literature has done to movement screening generally, I would be surprised if one existed.
What the evidence does support is the negative claim — that forbidden-exercise lists and biomechanical screens do not predict who gets hurt. That is a reason to stop using lists. It is not, by itself, proof that my alternative works.
And genuine contraindications exist. Uncontrolled hypertension, acute injury, certain post-surgical windows, some connective tissue conditions, retinal problems with heavy straining, and pregnancy-specific considerations. Those are medical, they are individual, and they are the reason to ask rather than to consult a list.
There are almost no dangerous exercises. There are dangerous mismatches — between what a movement demands and what a person has currently earned the right to do.
Which is worse news than a list, because a list can be memorised and a mismatch has to be assessed. But it is the true version, and it explains why the boxer is fine and the person copying a video is not.
Educational content, not medical advice. If you have a diagnosed condition, an existing injury, or you are pregnant or recently post-surgical, get individual advice before starting or changing loaded exercise — those are the situations where a genuine contraindication may apply.
Including what you have actually earned the right to do. That is an assessment, not a list.
Online movement coaching → Book a discovery call