On Monday, November 17, 2014 6:39:21 PM UTC, Col wrote:
Dawlish wrote:
On Sunday, November 16, 2014 7:20:04 PM UTC, Dave Cornwell wrote:
--------------------------------------------------------------
One of the few when the models have not worked out to be right. I
know you've suggested I do it but I'm too lazy :-) - but for
statistical sake I would love you to try and make a forecast (or
guess, which it would be with no agreement, say 6 times a year, to
see if it is worse or better than the ones with the model
consistency. Be interesting to see exactly how far from 20% it would
be. My guess is 50% as it is random - but then of course the next
question is how does one assess accuracy?

My guess would be around 50% too. I feel that there are only
particular times when forecasting accurately is possible. If someone
wants to verify that, be my guest. My work has certainly shown that
when particular parameters are met, forecasting at 10 days can be
usefully accurate.

Although this was of course one of the 20% that got away
Of course even a failure tells us something interesting.
10 days ago there was no sign of the current transition to a more
settled regime, you wouldn't have made that forecast if there had
been. Around 7 days ago the models began to pick up on it and
although it has ebbed and flowed somewhat has stuck with that
theme ever since.

Although this forecast wasn't a complete disaster we did get the
low pressure dominating after all, the 'type' wasm't correct.
As you only assess on a righ/wrong basis, what are your criteria
for evaluation?
There must be a fair few 'marginal' situations.
--
Col

Bolton, Lancashire
160m asl
Snow videos:
http://www.youtube.com/channel/UC3QvmL4UWBmHFMKWiwYm_gg

You'd be surprised Col. I'm my own harshest critic on judging a forecast right, or wrong and thus, I don't see the marginal ones as close. I'm either just about spot on (correct) or not (wrong). Although the low pressure dominance is there, there is no way this is Atlantic air. A case, for me, of too wrong to be right.