Regarding this portion of the description of the Forecasting Task: " because a portion of the seasonal water supply may be known (for full months that have passed before the issue date, see “Antecedent monthly naturalized flow” section below), those forecasts will have an unknown portion that is only the residual (remaining) naturalized flow."
Does this mean that it is legal to use the test monthly naturalized flow values for months prior to the current month of the issue date? For example, in July we could use the test monthly naturalized flow values and then add them to a prediction for the monthly naturalized flow value for July?
Hi @oshbocker. I have made this message public. Please ask general questions as a public topic instead of as a private message.
Does this mean that it is legal to use the test monthly naturalized flow values for months prior to the current month of the issue date?
Yes, it is okay to use the test monthly naturalized flow values for months prior to the current issue date when you are overlapping with the forecast season.
For example, in July we could use the test monthly naturalized flow values and then add them to a prediction for the monthly naturalized flow value for July?
Yes. For an issue date in July, you can use the known test monthly naturalized flow values through June, forecast the monthly value for July (the residual/remaining unknown amount), and add those together for your prediction.
Could we assume that on July 1, we always got the data up to date (June 30) for the monthly naturalized flow? Well, my code could handle the missing data but the performance will be not as good as it should be. So should we throw an exception so that we could re-run it or should we still handle the missing values?
You should handle the missing values. We will download the data for the 2024-07-01 issue date once, and if June’s monthly naturalized flow is not included, it will just not be available.
For any direct API access data sources, deliberately throwing exceptions to wait to rerun at a later date with more data is abuse of the opportunity to fix your code and may lead to disqualification.