Part of the Maths & Statistics suite · 235 calculators

R Squared Calculator

How much of the variation your model explains — R² from observed and predicted values, with the adjusted version that penalises extra predictors.

R² = 1 − SS_residual ÷ SS_total.

Results update as you type
Results
90%
Adjusted R²
Residual sum of squares
Total sum of squares
Root mean squared error
Mean absolute error
Reading
Observations
Reviewed September 2026. Pure mathematics: the result does not depend on where you are. Terminology follows the Australian Curriculum (maths, brackets, decimal point).
No account required · Google Analytics off unless allowedCalculator arithmetic runs in your browserResults update as you type
All calculations run 100% in your browser. The calculator code does not submit your figures to GlobalCalc to obtain a result.
About r squared

How the r squared calculator works

R² = 1 − SS_residual ÷ SS_total. It compares your model against the simplest possible one — always predicting the mean. R² = 0 means you did no better than that; R² = 1 means the predictions are exact.

Adjusted R² subtracts a penalty for each predictor, because plain R² can only rise when you add variables, even useless ones. On out-of-sample data R² can go negative, which simply means the model is worse than the mean.

Formula: R² = 1 − SSres/SStot; adj = 1 − (1−R²)(n−1)/(n−k−1)

Worked examples

InputsNote
A good linear fit90%R² about 96%
Perfect predictions100%R² = 100%
Predicting the mean0%R² = 0

Frequently asked questions

What does R² actually measure?

How much better your model is than always predicting the mean, in squared-error terms.

Is a high R² always good?

No. Time-series data can produce R² above 0.99 from pure trend, and a model can fit training data perfectly and predict nothing.

Can R² be negative?

Yes, on data the model was not fitted to. It means the model does worse than the mean — a genuinely useful warning.

When should I use adjusted R²?

Whenever comparing models with different numbers of predictors. Plain R² can only go up when you add variables, so it always favours the bigger model.

What is a good R² value?

Entirely field-dependent. Physics experiments expect 0.99; human behaviour research is pleased with 0.3.

Where these figures come from

Last checked: September 2026. Formulas are fixed by mathematics and do not change with tax years or regulations.