Insights

The mistakes that show up most in real training footage.

Across 161 video sessions analyzed between August 2, 2026 to September 22, 2026, the same handful of faults keep repeating. These are the counts, not opinions. Every number below is an aggregate, with no athlete, clip or gym identified.

Last updated

161 sessions counted

Boxing: the five most common faults.

161 findings were recorded on boxing footage in this window. Share of sessions is the percentage of all boxing sessions in which the fault appeared at least once.

1. Guard drops after striking

30.4%

49 of 161 sessions · guard

Hands drop below chin level after throwing a strike, leaving the head exposed.

Why it happens: Focus shifts to offense and the return path of the hand is neglected.

The fix: Treat the return path as part of the punch. The hand snaps back to the chin before the next action starts, and the drill is scored on return speed rather than power.

Drill: Rubber band returns. Shadowbox 3 rounds of 2 minutes, snapping each hand back to guard like a rubber band after every punch.

Read the technique breakdown →

2. Chin lifts when stepping in

5%

8 of 161 sessions · defense

Chin rises above shoulder line when closing distance to engage.

Why it happens: Eagerness to engage overrides the habit of staying tucked.

The fix: Make the chin tuck part of the step, not part of the punch. Every forward step starts with the chin pressed toward the chest and the lead shoulder high.

Drill: Chin-to-chest entries. Mirror work, 3 rounds of 2 minutes, exaggerating the chin tuck on every entry and jabbing from the tucked position.

Read the technique breakdown →

3. Stance widens under fatigue

3.7%

6 of 161 sessions · stance

Feet drift wider than shoulder width as rounds progress, reducing mobility and punch power.

Why it happens: Fatigue or anxiety narrows focus to the hands and neglects footwork fundamentals.

The fix: Reset the stance deliberately after every combination instead of letting the base spread over the round.

Drill: Wall shadow rounds. Stand an arm's length from a wall, throw combinations for 3 rounds of 2 minutes, and check stance width on every reset.

Read the technique breakdown →

4. Guard drops after throwing

1.9%

3 of 161 sessions · guard

Lead hand stays low after the jab instead of snapping back to guard position.

Why it happens: Overemphasis on reaching with the jab at the expense of the defensive return.

The fix: Score the jab on the return, not the reach. The lead hand touches the chin again before anything else happens.

Drill: Snap-back jab drill. Mirror, 3 rounds of 2 minutes of single jabs, watching the hand return path rather than the impact.

Read the technique breakdown →

5. Stance collapses under pressure

1.2%

2 of 161 sessions · stance

Stance narrows, feet come together, or the base disappears when pressured.

Why it happens: Pressure causes retreat without maintaining structural integrity.

The fix: Practise giving ground while keeping the base. Circle off at an angle instead of walking straight back with the feet coming together.

Drill: Pressure shadow rounds. Shadowbox backward and laterally for 3 rounds of 3 minutes, checking stance width every ten seconds.

Read the technique breakdown →

Method

How these numbers were counted.

Every mmaGPT analysis writes its findings against a fixed technique taxonomy, so the same fault is labelled the same way across every session. To build this page, findings from August 2, 2026 to September 22, 2026 were grouped by discipline and by fault, then counted. Share of sessions is the number of distinct sessions containing that fault, divided by all sessions in the discipline.

Only disciplines with at least 100 recorded findings are published, which is why boxing is the only discipline here so far. Below the threshold in this window: MMA 89 findings, Muay Thai 82, kickboxing 39, Taekwondo 16, BJJ 13, wrestling 6, karate 5, judo 3, Krav Maga 2. Those will be published as their counts grow. Findings the model could not classify against the taxonomy are excluded from the ranking rather than lumped together.

The page is built from a static snapshot. It holds counts only, and contains no athlete names, user ids, video clips or gym names. For the coaching workflow behind these findings, see how the video analysis works and how to analyze sparring footage like a coach.

Find out which of these is costing you.

Every fault on this page was found on somebody's footage. Upload one clip and see which of them is in yours. First session free, no card needed.