How to read a compatibility verdict
AquariumHQ publishes verdicts for 4,088 pairings. Ninety-one of them carry documented behavioural evidence; the rest are derived. This guide explains what the difference means, what the three qualifiers on every documented relationship are for, and why a chart with one cell per pair is wrong about roughly half the pairings it covers.
Short answer
Most compatibility verdicts anywhere, including here, are arithmetic on stored traits rather than reports of what actually happened. That arithmetic is sound and it is not the same claim, and a system that presents the two identically is overstating the first — precisely on the pairings nobody has tried.
Derived and documented are different claims
A derived verdict says: these two animals want compatible water, neither has a trait in our records suggesting conflict, and the arithmetic comes out favourable. That is a statement about AquariumHQ's own data.
A documented verdict says: this pairing is reported, consistently or inconsistently, and here is what the reports say. That is a statement about the world.
The first cannot substitute for the second, and the gap between them is not small. Across 4,088 published pairings and nineteen assessed dimensions each, 0.3% of the individual factors rest on a documented observation. The rest are derived from stored traits or honestly recorded as unknown.
That number is published rather than buried because it is the most important thing to know about any compatibility system, including the ones that do not tell you.
The three qualifiers, and why a binary chart cannot hold them
Every documented relationship here carries three dependency ratings, because three things routinely decide an outcome that a single verdict cannot express.
- TANK SIZE. Most conditional pairings are conditional on dimensions. Tiger barbs and angelfish in a 60 cm tank with five barbs is a different proposition from a 120 cm tank with twelve, and a chart with one cell has to pick one and be wrong about the other.
- LIFE STAGE. A pairing that works for six months and fails when one animal reaches adult size or starts spawning is not a compatible pairing. It is an incompatible one with a delay, and the delay is long enough to be mistaken for success.
- INDIVIDUAL VARIABILITY. Some species are genuinely unpredictable between individuals — bettas above all. Reporting a species-level verdict for an animal whose behaviour is decided per fish is false precision, and it is why so many keepers find the charts wrong.
The shapes that keep recurring
Read across ninety-one documented relationships, three patterns dominate, and they cross families that have nothing else in common.
The pairing fails on a delay. Angelfish and cardinal tetras, guppies and angelfish, oscars and almost anything, a black ghost knifefish and a community tank. Every one works for months and then does not.
The tank decides, not the species. Rift-lake cichlids, large Central Americans, every territorial pairing. The outcome flips on footprint and on how much structure breaks the sightlines, and the species names barely move it.
The animals interbreed, which is not an incompatibility at all. Platy and swordtail, guppy and Endler's, Malawi mbuna across genera. Nothing is harmed and nothing fights. What is lost is the identity of the fish, permanently and invisibly, because the first hybrid generation looks like its parents. No compatibility engine has a dimension for this and none can, because it is not a welfare question.
Where the evidence is genuinely thin
Most of the 96,141 possible pairings in this catalogue have never been tried by anybody who wrote it down. For those, the honest rendering is a derived verdict labelled as derived, and that is what they get — a pair with no documented entry is not presented as though somebody had checked.
This is also why AquariumHQ does not publish a single compatibility score. A number implies a measurement. What exists is a set of separate assessments on separate dimensions, some derived, some documented, some unknown, and collapsing them into one figure would destroy exactly the information a reader needs to decide how much to trust it.