Function evaluate 

pub fn evaluate(
    series: &BTreeMap<(usize, usize), Vec<f64>>,
    accumulator_columns: &[usize],
    train_fraction: f64,
) -> PredictionReport

Fit and evaluate a one-step predictor over remembered series.

accumulator_columns is what the observer believes accumulates: either told by the mirror, or worked out from behaviour. It changes which baseline each cell is scored against, so getting it wrong is penalised.

The series are split in time: the first train_fraction fits, the rest scores. No cell is ever scored on a sample used to fit it.