When creating forecasts for time series which have step changes we create a model of conditions under which we expect the time series to step, specifically the values at and interval between steps, based on historical data. This is a probabilistic model so we run a number of roll outs to estimate an expected value and distribution. We have seen this misbehaving when the forecast time series value is too far from the values for which we have a reasonable characterisation of this distribution. It would be more appropriate to be cautious in such cases: at the moment the behaviour depends on the choice of characterisation of distribution tail values. This issue covers the work to detect and avoid stepping the time series in such cases.
When creating forecasts for time series which have step changes we create a model of conditions under which we expect the time series to step, specifically the values at and interval between steps, based on historical data. This is a probabilistic model so we run a number of roll outs to estimate an expected value and distribution. We have seen this misbehaving when the forecast time series value is too far from the values for which we have a reasonable characterisation of this distribution. It would be more appropriate to be cautious in such cases: at the moment the behaviour depends on the choice of characterisation of distribution tail values. This issue covers the work to detect and avoid stepping the time series in such cases.