Video feedback: how often to watch your swing

The trade: better practice, worse retention test

Two things get measured in a feedback experiment and they rarely agree: how well the practice session went, and what survives to a test a day or two later with the feedback switched off. Frequent feedback flatters the first and can shrink the second. That pattern has a name, the guidance hypothesis, credited to Winstein and Schmidt in 1990: cues after every attempt crowd out the error detection a learner would otherwise do alone.

The clearest golf test is old and small. Butki and Hoffman put 78 novices through 96 putts each under three conditions: one group watched every putt the whole way, a second saw the ball's path on half its putts, a third never saw the path and learned only where the ball finished. During practice the group that saw everything putted best. At the retention tests both deprived groups had passed it, and the one that had seen least led both.

The one golf video study, read twice

14 daysWhen the video group's advantage appeared, and not before

Guadagnoli, Holcomb and Davis gave 30 golfers four 90-minute sessions and split them three ways: practice with video, with a teacher talking, or alone. Everyone hit 15 balls with a 7-iron in a test straight afterwards and again two weeks later. Their own summary is blunt. On the first post-test both taught groups performed worse than the self-guided one. On the second both were better, with video best.

A 2024 systematic review reads that experiment measure by measure and gets a quieter story: error distance showed no group differences at all, and the video group's gain turned up in total distance, only at the second retention test. The softer account is the one with the measures attached. That review covered 52 randomised golf-learning trials, more than half underpowered, and concluded that for augmented feedback no single method can be considered the most favourable.

Reduced, faded and bandwidth are three different dials

66%The feedback frequency that beat both 100 and 33 percent

Researchers did not simply turn feedback down. Reduced schedules give it on a fixed fraction of attempts; faded schedules give it often at first and less as practice goes on; bandwidth schedules give it only when the error falls outside a tolerance, so an attempt inside that tolerance meets silence.

Less is not always better. Nunes and colleagues had 60 adults aged 60 to 69 practise a basketball free throw, 90 attempts across three sessions, with feedback on their movement pattern after 100, 66 or 33 percent of them. Sixty-six percent was superior on both the movement score and the points score, and held at the 24-hour retention and transfer tests. Frequency is not the only dial, either. Wulf and colleagues gave children a soccer throw-in and found feedback after every trial beat feedback after every third, when it pointed at the movement's effect rather than at the body.

Bandwidth has been tried on golfers. Smith and colleagues split 48 novices across three chipping conditions, one fed after every chip and two fed only when the error exceeded a 5 or 10 percent tolerance. The review's table and its own text disagree about which bandwidth group improved, so hold that result loosely. Neither version names the group fed after every ball.

Left alone, golfers ask for about one look in four

27%Trials on which learners asked to see their own video

Aiken, Fairbrother and Post let learners call for video of their own basketball set shot whenever they wanted it. They asked on 27 percent of trials overall, 33 percent in the first block dropping to 19 percent in the last, and scored higher on form at transfer. Two other self-chosen rates sit in the same paper: 5.8 percent in a study by Wulf and colleagues, and 44.7 percent falling to 28 percent in one by Chiviacowsky and Wulf.

Golfers do the same. A 2026 putting experiment gave 28 adults 50 practice putts and let half decide when to see the result; they asked 4.9 times on average, under one putt in ten, and beat their yoked partners a day later. Post and colleagues found the equivalent transfer advantage in chipping, while a putting study by de Souza Nunes and colleagues came out level. No chosen rate landed near every swing, though, and every group that chose tapered.

A rhythm for one bucket, and the swings between

Where the numbers converge is a fraction between one swing in ten and one in three, weighted toward the start. On a 60-ball bucket that is a dozen looks, most in the first half, the gap widening as you go. Watch the first swing, then hit six or eight before the next look; that sits closer to every schedule above than watching in pairs.

The swings in between are the point rather than the filler. What the guidance hypothesis describes is a screen so prompt that it replaces a judgement which would otherwise have been made, so form the judgement first. Before the clip opens, commit to what the club did. In that putting experiment, how well a golfer monitored their own performance predicted improvement only among the players whose feedback timing was decided for them (r = 0.56, p < 0.05).

None of this was measured on a golfer alone with a phone. Every schedule above was set by an experimenter, on novices, and the golf review notes that most of its trials used simple putting tasks. Pebble Swing returns joint angles and a backswing-to-downswing tempo ratio from one clip, which argues for spending clips carefully rather than filming everything.