Why body shape decides what gets eaten
A fish does not swallow length. It swallows cross-section, through a mouth, and both halves of that are properties of shape. This guide explains what AquariumHQ's compatibility engine now does instead of the length rule, why the change can only ever add warnings and never remove them, and how to apply the same reasoning to a pairing the catalogue does not hold.
Short answer
The hobby's rule is that a fish will attempt anything under about 40% of its own length. That is a statement about one shape of predator and one shape of prey, and it is wrong in both directions for everything else: it over-warns about deep-bodied fish and badly under-warns about slender ones.
The rule, and the two ways it fails
A 12 cm angelfish and a 12 cm silver hatchetfish are the same length. They are not remotely the same mouthful: the angelfish is a disc nearly half as deep as it is long, and the hatchetfish is a keeled wedge. A predator that can manage one may be entirely unable to manage the other.
The failure that matters more runs the other way. A 10 cm kuhli loach is about a centimetre through. A predator that could never swallow a 10 cm tetra will take the loach without difficulty, and the length rule says the tetra is at risk and the loach is not. Elongate fish — kuhli loaches, spiny eels, ropefish, pencilfish, glass catfish, halfbeaks — are the group the rule misses most badly, and they are exactly the group sold as peaceful community animals.
The predator's half is no better. A 30 cm giant snakehead and a 30 cm bristlenose pleco have the same length and gapes that differ by a factor the length rule represents as nothing at all. One takes prey most of its own length; the other takes none.
What replaces it
Two shape classes per animal — how wide the mouth opens as a fraction of the animal's own length, and how deep the prey is through its widest cross-section — produce a threshold instead of a constant. The coefficients are pinned so that the ordinary case, a moderate mouth against a torpedo-shaped prey, reproduces the familiar 0.4 exactly. Every departure from 0.4 elsewhere on this site is a departure that shape caused, which is what makes the change checkable rather than merely plausible.
The classes are assigned per family, with individual species overridden where the family default would be confidently wrong: Osphronemidae covers the honey gourami and a 70 cm giant gourami, and Cichlidae covers the discus and the jaguar cichlid. Every assignment records whether it was examined for that species or inherited from its family, and the site never averages those two together.
Shape can widen the net and cannot narrow it
Run purely as geometry, this model moves thousands of pairings towards a worse verdict and rather more thousands towards a better one. The second number is the problem. A shape model that removes several thousand predation warnings has done something indistinguishable from relaxing a threshold to quieten an audit, and no argument about the geometry being correct makes that a safe way to ship it.
It is also not obviously right. A recorded predation trait is a statement about what an animal DOES; geometry is a model of what it could swallow whole, and the two come apart. A red-bellied piranha takes bites out of fish far too deep to ingest. So both thresholds are evaluated and the stricter outcome wins. Where shape says a prey is harder to swallow than its length suggests, that appears as context inside the warning rather than as a downgrade of it.
Applying it to a pairing this site does not hold
- Look at the prey end-on, not side-on. The question is the widest cross-section, which for a depressed catfish is across its pectoral spines and for a puffer is inflated.
- Look at the predator's mouth when it is open, not closed. An ambush predator's gape is frequently the width of its own head and nothing about the closed mouth suggests it.
- Slender prey is riskier than its length says. If you keep anything eel-like with anything predatory, assume the length rule is flattering you.
- A grazing or sucking mouth cannot take a fish at all, and in those pairings the risk usually runs the other way — towards the grazer being harassed rather than the fish being eaten.
- None of this describes armour, spines or behaviour. A fish that fits inside a mouth may still be un-swallowable, and a fish that does not fit can still be attacked.