A Forecasting Model That Correctly Refuses to Forecast

It beat the baseline by 98.7% on one series and lost to repeating the last value on another. That contrast is the most useful single fact about it. Part one covered what TimesFM 3.0 is: a stateless n