Reviews
Part of Photo quizzes: fairness depends on the instruction, not the picture
Seven techniques for taking information away in a hard photo quizzes round
Hard photo quizzes work by taking information away, seven ways of doing it, one technique per round, and a ceiling that reproduction quality quietly sets for you.
A hard picture is not a picture of somebody obscure. It is a picture of somebody the room knows, presented so that recognition has to be earned. The difference matters because obscurity produces silence, and a hard round is supposed to produce leaning, squinting and argument.
This page is a catalog of the ways an image can be made harder, what each of them actually tests, and the ceiling that keeps a hard round from becoming an impossible one.
What to take away
- Difficulty comes from the presentation, not from the choice of person. Pick faces the room knows and take information away from the image.
- Every hardening technique tests something different. Mixing them in one round without saying so makes the round feel arbitrary.
- Somebody in the room should get every image. If nobody does, the image was not hard; it was unanswerable.
Seven ways to take information away
Recognition of a face relies on a small set of cues, and each technique removes one. Knowing which cue you removed tells you what the question now tests.
Seven ways to remove information
Technique
- Tight crop
- Context, hair, outline
- Coarse pixelation
- Fine detail
- Single feature
- Everything but one region
- Inversion
- Upright arrangement
- Era shift
- The known face
- Unusual angle
- Frontal view
- Monochrome low contrast
- Color and skin tone
What it removes
- Tight crop
- Features alone
- Coarse pixelation
- Shape and proportion
- Single feature
- A distinctive eye or mouth
- Inversion
- Piece-by-piece analysis
- Era shift
- Earlier career stages
- Unusual angle
- Familiarity beyond portraits
- Monochrome low contrast
- Structure over surface
What it tests
- Tight crop
- Any face fits
- Coarse pixelation
- Becomes a color field
- Single feature
- Feature not distinctive
- Inversion
- Disorienting, not hard
- Era shift
- Before they were public
- Unusual angle
- Profile looks like others
- Monochrome low contrast
- Reproduction loses contrast
Where it fails
- Tight crop
- Coarse pixelation
- Single feature
- Inversion
- Era shift
- Unusual angle
- Monochrome low contrast
What it removes
- Tight crop
- Context, hair, outline
- Coarse pixelation
- Fine detail
- Single feature
- Everything but one region
- Inversion
- The upright arrangement the brain expects
- Era shift
- The face the room knows
- Unusual angle
- The frontal view most images show
- Monochrome and low contrast
- Color and skin tone cues
What it tests
- Tight crop
- Knowledge of the features alone
- Coarse pixelation
- Recognition of overall shape and proportion
- Single feature
- Memory for a distinctive eye, mouth or profile
- Inversion
- Piece-by-piece analysis rather than instant recognition
- Era shift
- Knowledge of a career's earlier stages
- Unusual angle
- Familiarity beyond the standard portrait
- Monochrome and low contrast
- Structure over surface
Where it fails
- Tight crop
- Crops so tight that any face fits
- Coarse pixelation
- The pixelation is so coarse that the image is a color field
- Single feature
- A feature that is not actually distinctive
- Inversion
- Rooms find it disorienting rather than hard
- Era shift
- An image from before the person was public
- Unusual angle
- A profile can look like several people
- Monochrome and low contrast
- Reproduction, which was already losing contrast
Inversion deserves a note because it works for a reason the room will enjoy hearing. Faces are processed as wholes. Turning one upside down defeats that processing, so gross distortions go unnoticed. This is the Thatcher effect.
An inverted face is hard even when the person is extremely well known. That is exactly what a hard round wants.
Keep the technique consistent within a round
A room can cope with any of the seven, but not with not knowing which one it is looking at.
Mixed versus consistent round
Mixed techniques
- Table effort
- Work out each type
- Round feel
- Reads as random
- Difficulty
- A lottery
- Curve
- None
One technique
- Table effort
- Answer straight away
- Round feel
- Reads as fair
- Difficulty
- Tunable across images
- Curve
- Generous to savage
If image one is a tight crop, image two is pixelated, and image three is upside down, tables must work out each question type before they can answer. The round reads as random.
So build a hard round around one technique, or at most two announced in advance. "All eight images are cropped to the eyes" is a fair hard round. Eight images each obscured differently is a puzzle about the host.
Consistency also makes difficulty tunable. With one technique you can adjust the severity across the round, from a generous crop on image one to a savage one on image eight. The round acquires a curve rather than a lottery.
The reveal format
The strongest structure for a hard picture round is progressive: show the hardest version first, then a less obscured one, then the full image, with points falling at each step. It has the same virtue as a descending-clue question. Tables decide when to commit, the decision carries risk, and every image is eventually answerable by everyone.
Three-stage progressive reveal
- Stage onehardest version, few tables get it
- Stage twoless obscured, most tables get it
- Stage threefull image, everyone can answer
- Points fall at each step
- Run on a screen, not on paper
Three stages is enough. A first stage that only a few tables will get, a second that most will, and a full reveal. Run it on a screen rather than on paper, since paper cannot reveal.
The ceiling
A hard question has one defining property: somebody in the room gets it. Not the host, who knows the answer. Somebody playing.
If, across the whole room, no table identifies an image, the image has failed. It measured nothing, it cost a slot, and it told the tables that the round is not worth trying at. Two such images in a row and the room stops looking at the pictures.
So test the round on people before the night, at the severity and in the reproduction you intend to use, and cut anything that nobody gets. The bar for keeping an image is not that it was hard. It is that it was hard and somebody got it.
The same test catches the opposite failure. An image every test player identifies instantly is not hard, whatever you did to it, and belongs in an easier round.
Reproduction changes the difficulty
Every technique on the list interacts with the projector, the printer or the phone screen the room will actually use. Pixelation applied on a laptop and then projected has been pixelated twice. A low-contrast image on a photocopier becomes a blank rectangle. A tight crop printed small becomes a thumbnail of a texture.
Reproduction makes it harder
Technique
- Pixelation
- Pixelated twice when projected
- Low contrast
- Blank rectangle on a photocopier
- Tight crop
- Thumbnail of a texture when printed small
Reproduction effect
- Pixelation
- Low contrast
- Tight crop
Set the severity in the room, on the equipment, from the back. A hard round that was calibrated on a bright screen at arm's length is a different round in the venue, and the difference is always in the direction of harder.
Where this connects
Base checks for any picture on a sheet, from verifying what the image shows to numbering it so a fold cannot lose it, are set out under photo quizzes. Hardening an image does not reduce that need.
Cropping and context as difficulty tools in a face round are described from the recognition side in guess the celebrity picture quiz.
If the images are grouped by field, the fairness problems in guess the celebrity by profession apply on top, since a hard round is even more sensitive to a table's uneven knowledge. Whatever the technique, the person in the picture still has to be identified by the sheet correctly, which is the claim examined in guess the celebrity.
Common questions
How many images should a hard round hold?
Eight, with a progressive reveal. More than that and the squinting exhausts the room; fewer and the round is decided by one lucky recognition.
Is it fair to use an image from early in someone's career?
Yes, if the round says so. An era shift is a legitimate technique when announced and a trick when it is not.
Should hard images score more?
No. The progressive reveal already rewards early recognition. Adding a multiplier on top makes the round decide the night, which a picture round should not.
What if a table identifies the image from the background rather than the face?
Give them the point, then crop tighter next time. Recognizing where a picture was taken is knowledge too, and the leak was yours.
Can generated or altered images be used to make a round harder?
Not as the question. An image that is not a photograph of the person has no correct answer, and a hard round built on it is a round with a hidden trick.
