On 7 Apr, 16:01, Dawlish wrote:

The operational run is the one that the computer
has selected as the most likely to achieve outcome and even that is
often hopelessly wrong at 10 days and each single operational run has
a very low probability of achieving outcome.

It's very difficult to interpret the output from the ensembles in the
GFS when they're run at a different resolution to the operational run.
On occasions the operational run can be completely different to the
ensembles which renders the whole thing useless and just highlights
the role of model resolution on forecast accuracy !

Richard