Signal-to-background ratio
The ratio of expected signal events to background events in a selection.
A high ratio makes signals robust against background uncertainties. In counting experiments the achievable significance scales roughly as S/√B for large B. Optimising cuts trades signal efficiency against background rejection.
In plain words
If you are listening for a friend's voice in a crowded room, what matters is not only how loud your friend is, but how loud the room is. The signal-to-background ratio compares the number of events from the process you want with the number of look-alike events from everything else. A ratio of ten means the signal dominates; a ratio of one in a hundred means it is buried.
In detail
For a counting experiment with S expected signal events and B expected background events, the significance of an excess is approximately S/√B when B is large, because the background fluctuates by about √B. When B is small this approximation overstates the significance, and a more accurate expression based on the likelihood ratio is used.
The two formulas illustrate why both quantities matter. Ten signal events over a background of 100 give Z ≈ 1, hardly noticeable. Ten signal events over a background of one give Z ≈ 5.7 with the full formula. The ratio S/B also controls how strongly a result depends on systematic uncertainties: if B is ten times larger than S, a 10% error in the background estimate is as large as the whole signal.
Example
In 2011 T2K reported six electron-neutrino candidates where about 1.5 background events were expected, the first indication of νμ → νe appearance and of a non-zero θ13. The small count was meaningful because the background was small and well understood. Today, rare-event searches such as neutrinoless double beta decay aim for a background-free regime, where fewer than one background event is expected in the whole exposure.
How it is optimised
Every selection cut removes some signal and some background. Analysts choose cuts that maximise the expected sensitivity, usually with simulation and control samples, and fix them before looking at the signal region in a blind analysis. Background reduction is often more valuable than more mass: lowering B by a factor of four improves S/√B as much as quadrupling the exposure.
Frequently asked questions
What is a good signal-to-background ratio?
It depends on how well the background is known. Precision oscillation measurements work with S/B above one, while some searches succeed with S/B far below one if the background is measured very precisely.
Is S/√B the same as the sigma of a discovery?
Approximately, for large B and no systematic uncertainties. Published significances use full likelihood fits and include systematics and the look-elsewhere effect.
Related terms
Mentioned in 6 pages
Last updated: · Editorial standard
Cite this page
Text
Neutrino Atlas (2026). “Signal-to-background ratio”. Retrieved 22 November 2026, from https://neutrino-atlas.com/terms/signal-to-background/BibTeX
@misc{neutrinoatlas2026termssignaltobackground,
title = {Signal-to-background ratio},
author = {{Neutrino Atlas}},
year = {2026},
url = {https://neutrino-atlas.com/terms/signal-to-background/},
note = {Accessed 2026-11-22},
language = {english}
}