Same face. Different score.

Two things distort a jaw-ratio or symmetry read before anything about your face changes:

• Angle — center your face at eye level, straight-on. Up, down, or off to one side skews the proportions.
• Distance — keep it consistent between attempts. Closer or farther changes apparent width-to-height ratios.

• Even, diffused light — daylight facing a window, or a soft front light — reads more consistently than a single harsh overhead source.
• Overhead-only light casts shadows under the brow, nose, and jaw that a landmark model can read as structure that isn't there.
• Jaw-related measurements are usually the most light-sensitive of the group — worth getting right first.

• A relaxed, neutral expression — mouth closed, eyes open and level — gives the most reproducible read.
• Smiling, squinting, or turning the head mid-capture introduces asymmetry unrelated to your baseline geometry.
• If a score swings a lot between attempts, rule out expression drift before assuming the setup is bad.

Mogalytics reads six camera-visible geometry ratios and settles them into a weighted composite over a 30-frame window, trimming the top and bottom 10% of samples so a single blink or turn doesn't skew the result:

• Eye line — canthal tilt, the angle from inner to outer eye corner.
• Jaw ratio — jaw height relative to face height.
• Frame symmetry — bilateral landmark variance, left versus right.
• Midface — eye line to lip distance relative to interpupillary width.
• Cheek ratio — cheek width relative to jaw width.
• Bridge — nose bridge height relative to width.

None of these change because you bought anything. They change because the camera, pose, light, or expression changed — see the full breakdown of why scores swing between scans — which is exactly why a repeatable setup is worth finding before you're live.

Related terms discussed alongside these six, each sharing an axis with one of the measurements above but not computed as a separate number:

The point isn't one lucky frame:

• Save a baseline reading, change one variable at a time — distance, then light, then angle — and save again.
• Every saved reading stays local, so you can compare session to session and read the trend.
• Once stable, check what tier that composite lands in, from Molecule to Slayer — a single-scan reading, distinct from how Omoggle's separate ELO rating works across many duels.
• Then walk into a live duel with a setup you've already tested — see is Omoggle safe? — the honest answer first, or why Omoggle isn't working if the queue itself won't load or scan.