Find a readable equation for your data — something like
2*x^2 - 3*x + 1, not a black box with a million weights.
Load a CSV or try a sample, pick the shape you expect, and press Run.
Under the hood it breeds random equations against your data and keeps the ones that fit — evolution, on algebra.
Runs entirely in your browser — nothing is ever uploaded. Fits one input to
one output (y = f(x)); if your CSV has more columns you choose
which is x and the rest are ignored.
Fewer is better, and by a lot. Each function you add
explodes the search space, so tick only what could plausibly be in your data.
Redundant choices (tan is just sin/cos) are struck through and dropped
automatically; div is the worst offender — it lets the search
fake any curve with a ratio of polynomials and never find the real law.
| fit (R²) | — |
| generation | — |
| equations tried | — |
| seed | — |
Every run uses a different seed, so running again searches somewhere else — the same data can have more than one good answer.
The best-fitting equation is nearly always the biggest one allowed — extra terms always shave off a little more error, leaving you a perfect R² and an unreadable answer. So here is the best at each size; the marked one (pinned top) is the simplest that gives up essentially nothing. Click any row to plot that equation.
The same trade-off in full, sorted simplest-first. Click any one to plot it above.