On Wednesday, February 12, 2014 10:50:22 PM UTC, David Mitchell wrote:
It would simply be guesswork, not bravery, and such 'forecasts' would have little or no value.



Far from it. What do you learn from predicting a near certainty? ............

Yawn. How do you know it is a near certainty? That's where I differ from others. I can tell you when there is an 80% chance that the models can be trusted at 10 days and my percentage accuracy has risen, *because* I've learned and refined my technique. You would have no idea of those times, yet you happily tell me I only forecast when there is a near certainty.

How on earth do you know when the models are pointing to a near certainty that the outcome in 10 days will be as they show on a particular day. Do tell.

This morning, for instance. Try a forecast at 10 days and see if you are accurate. I say there is not an 80% chance of a particular type of weather this morning.

Yawn. How do you know it is a near certainty? That's where I differ from
others. I can tell you when there is an 80% chance that the models can be
trusted at 10 days and my percentage accuracy has risen, *because* I've
learned and refined my technique. You would have no idea of those times,
yet you happily tell me I only forecast when there is a near certainty.

I would imagine most of us could trust the models at 20 days when we have
this relentless Zonal sh*te


Graham

"David Mitchell" wrote in message
...
It would simply be guesswork, not bravery, and such 'forecasts' would have little or no value.

Far from it. What do you learn from predicting a near certainty? A high percentage of accuracy.
What if you test your theories when certainty is unclear? A high percentage of inaccuracy. But
if you analyse your predictions and learn from that, then you you have a better chance of being
able to predict with a greater certainty in the future.

Fair enough in general terms, if you are trying to refine your knowledge about some field and
testing out alternative hypotheses.

However, underlying Dawlish's approach is the fact that on most days, it is *impossible* to give
a 10-day forecast with a useful degree of confidence.
To give a forecast in those conditions is pointless.

And as I have said on several occasions, the absence of a forecast is actually more useful,
providing information about the uncertainty.

What is more useful? Someone who gives you an answer that is more often wrong than right, or
someone who says truthfully "I don't know (and neither does anyone else)".

On Thursday, February 13, 2014 7:44:48 PM UTC, Graham wrote:
Yawn. How do you know it is a near certainty? That's where I differ from

others. I can tell you when there is an 80% chance that the models can be

trusted at 10 days and my percentage accuracy has risen, *because* I've

learned and refined my technique. You would have no idea of those times,

yet you happily tell me I only forecast when there is a near certainty.



I would imagine most of us could trust the models at 20 days when we have

this relentless Zonal sh*te





Graham

*))