Choosing by preference versus choosing by protocol

Do not submit the photo you like best if you want a comparable series: a favourite is chosen for flattering light and angle, not for matching your other submissions, so it overstates the result. Most people pick their favourite from a folder of similar shots, and that instinct is exactly backwards.

The selection bias in picking your favourite

A favourite photo is chosen for how it makes the subject look, which usually means the most flattering angle, the most forgiving light, the moment the framing happened to land well. That is a selection process, and selection processes systematically bias toward the high end of whatever range the photos span. Out of a set of otherwise similar shots, the one that looks best is disproportionately likely to be the one where lighting or angle happened to drift in a favourable direction - not because it is more representative, but because favourable drift is what makes a photo somebody's favourite.

This is the same statistical shape as choosing the best of several rating attempts instead of the median: selecting for the outcome you want, from a set of otherwise similar draws, reliably produces a number that overstates the underlying reality, whether the selection happens at the photo stage or at the results stage.

Why it destroys comparability

A series is only comparable if every entry in it was chosen by the same rule. "Whichever one I liked best that day" is not a rule that produces consistency, because "liked best" implicitly varies with whatever conditions happened to be favourable that particular session - different light, a different angle you didn't notice drifting, a different framing. Two series built by picking favourites are really two piles of the most-flattering outliers from two different sessions, not two comparable data points.

Why this is harder than it sounds

Knowing the rule does not make it easy to follow, because the pull toward the flattering shot is not a lapse in discipline, it is what looking at your own photos does to anyone. Epley and Whitchurch (2008) found that people were more likely to pick an attractively enhanced version of their own face out of a lineup as the real one, an enhancement bias the authors link to automatic rather than deliberate processing. Nobody scrolls a folder of near-identical shots and picks the second-best one on principle. The practical fix is to remove the choice rather than resist it: decide before shooting which frame in a sequence counts - always the third shot, always the one at the marked timer interval - so there is no moment where a folder of options sits in front of you inviting a pick.

The protocol answer

Submit by conditions, not by preference: take the shot the way the protocol specifies - fixed distance, fixed angle, fixed light, matching whatever the baseline was set up to be - and submit that one, regardless of whether it happens to be your favourite of the batch. If several exposures were taken under identical settings as a burst, submitting the first or a fixed position in the sequence, rather than scrolling through and picking the best-looking, keeps the selection rule constant across every session.

This applies the same way whether the destination is a model or a human recipient: sending your most flattering shot to a person is a legitimate choice if flattery is the actual goal, but it is a different goal from building a record that means the same thing each time you add to it, which is what this site is generally arguing for when it treats score comparability as worth more than any single high number. It is also the same discipline behind a physical measurement log: the reading taken is the reading recorded, not the best of several taken until one looked right. And it is unrelated to how much of the frame the model actually reads - the fix is upstream of the model entirely, in which photo gets uploaded at all.

The file worth sending to a rating tool is the one your protocol produced, not the one your eye picked out of the batch afterward. That distinction is worth writing into the protocol itself, as a line that says which frame counts before any photo exists to be tempted by, so the decision is made once, in advance, rather than fresh and biased every single session. Write the rule down where you will see it before you shoot, not after, because a rule you only remember once the folder is already full of tempting options has already lost.