i could make up any old email too.
still no proof.
links please.

On 15/02/2011 9:34 PM, Dawlish wrote:


Out of those 31 forecasts - all of which I'm happy to email to you -
26 have been correct and 5 incorrect. You can easily access them all
in the archive on here and see that.

You can happily question my methods and I'll happily defend them,
though it appears you would rather discuss in the 3rd person, which
makes discussion rather difficult, but *don't* question my honesty.
It's not in question.

On Feb 15, 6:20*pm, Dawlish wrote:

There aren't any like for like stats at 10 days. Which "forecasting
houses" (whatever they are) produce those stats? You've just made that
up, haven't you? If there are stats; link to them. I'd be really
interested to see them, but there aren't any at 10 days, which is why
I know you've just made that up.

Are you suggesting that forecasting companies do not perform regular
verification of 10-day (indeed all) forecasts? I'd be very surprised
if that were the case.

Stephen.

On Feb 16, 2:25*pm, Stephen Davenport wrote:
On Feb 15, 6:20*pm, Dawlish wrote:

There aren't any like for like stats at 10 days. Which "forecasting
houses" (whatever they are) produce those stats? You've just made that
up, haven't you? If there are stats; link to them. I'd be really
interested to see them, but there aren't any at 10 days, which is why
I know you've just made that up.

Are you suggesting that forecasting companies do not perform regular
verification of 10-day (indeed all) forecasts? I'd be very surprised
if that were the case.

Stephen.

It's not made public and therefore it's impossible to verify.

On Feb 16, 2:46*pm, Dawlish wrote:
On Feb 16, 2:25*pm, Stephen Davenport wrote:

On Feb 15, 6:20*pm, Dawlish wrote:

There aren't any like for like stats at 10 days. Which "forecasting
houses" (whatever they are) produce those stats? You've just made that
up, haven't you? If there are stats; link to them. I'd be really
interested to see them, but there aren't any at 10 days, which is why
I know you've just made that up.

Are you suggesting that forecasting companies do not perform regular
verification of 10-day (indeed all) forecasts? I'd be very surprised
if that were the case.

Stephen.

It's not made public and therefore it's impossible to verify.


Why would verifications be made public? Except the Met Office's,
perhaps.

Stephen.

On Feb 16, 4:05*pm, Stephen Davenport wrote:
That's about 1,000
possible daily forecasts, and around one in ten. So if 70-something
per cent of that one in ten is correct and the 90 per cent not
attempted are "wrong", then that's a "correct" rate of just under 8
per cent. We do a lot better than that.

I've tried the same maths with Paul, but it unfortunately falls on
deaf ears, Stephen.

You could take an even more evasive forecasting method of say,
forecasting when 7 consecutive GFS/EC runs that point to the same
forecast rather than 3, publish once every 3 months, and claim a 95%
accuracy rate. Huzzah.

Richard

On Feb 16, 4:05*pm, Stephen Davenport wrote:

3. As I understand it, nobody would expect a daily forecast from a non-
professional (and a degree of kudos to anyone that tries) but I think
you have made 100 forecasts in, what, three years? That's about 1,000
possible daily forecasts, and around one in ten. So if 70-something
per cent of that one in ten is correct and the 90 per cent not
attempted are "wrong", then that's a "correct" rate of just under 8
per cent. We do a lot better than that.

p.s. I'm quite glad that a professional who is doing actual day-to-day
medium range forecasting is able to throw his two penneth in to this
discussion.

Richard

On Feb 16, 5:01*pm, Richard Dixon wrote:
On Feb 16, 4:05*pm, Stephen Davenport wrote:

3. As I understand it, nobody would expect a daily forecast from a non-
professional (and a degree of kudos to anyone that tries) but I think
you have made 100 forecasts in, what, three years? That's about 1,000
possible daily forecasts, and around one in ten. So if 70-something
per cent of that one in ten is correct and the 90 per cent not
attempted are "wrong", then that's a "correct" rate of just under 8
per cent. We do a lot better than that.

p.s. I'm quite glad that a professional who is doing actual day-to-day
medium range forecasting is able to throw his two penneth in to this
discussion.

Richard

So am I.

On Feb 16, 7:07*pm, Dawlish wrote:

Show me any organisation that forecasts with any decent degree of
accuracy at 10 days, including yours and I'll believe you. I really
would like to think an organisation that can do this. Why you feel you
have to withdraw, after your reaction to a request for accuracy
statistics from you being "Don't be daft". I really don't know.
Nothing is discourteous about questioning your accuracy, when you
don't provide *any* evidence for your implied assertion that you are
accurate. I don't think you (which organisation are you defending
anyway?) can forecast with accuracy at 10 days - say 60+% accuracy, in
answer to your question. If you can; demonstrate it. You were very
quick to jump in and criticise my forecasts, which have demonstrable
accuracy in what I do.

I did not jump on the accuracy of your forecasts - in fact my very
word was "kudos" to anyone who tries it. My question is regarding the
practical usefulness of forecasts issued on such a sporadic basis.

How about taking my word for the fact that I/we have a reasonable
accuracy at ten days? I haven't for one second questioned the 70-80%
accuracy you state for the forecasts you produce. What is "daft" is to
expect me to produce privileged data on an open discussion board. I
have nothing I feel that I need to prove so whether you believe it or
not does not matter but I will defend myself and forecasters in
general.

By baldly stating that you simply "do not think" that I/we can
possible forecast with any semblance of accuracy you are impugning my
integrity and professionalism, and that is more than discourteous, and
underlines why I am withdrawing from a discussion which I should
perhaps have had more sense than to enter in the first place.

Stephen.

On 16/02/11 15:09, Stephen Davenport wrote:
On Feb 16, 2:46 pm, wrote:
On Feb 16, 2:25 pm, Stephen wrote:

On Feb 15, 6:20 pm, wrote:

There aren't any like for like stats at 10 days. Which "forecasting
houses" (whatever they are) produce those stats? You've just made that
up, haven't you? If there are stats; link to them. I'd be really
interested to see them, but there aren't any at 10 days, which is why
I know you've just made that up.

Are you suggesting that forecasting companies do not perform regular
verification of 10-day (indeed all) forecasts? I'd be very surprised
if that were the case.

Stephen.

It's not made public and therefore it's impossible to verify.


Why would verifications be made public? Except the Met Office's,
perhaps.

Stephen.

1. To demonstrate the skill of their forecasts to the public.

2. To demonstrate that they have nothing to hide by being open with
their forecast skill.

"Stephen Davenport" wrote in message
...
... but I think
you have made 100 forecasts in, what, three years? That's about 1,000
possible daily forecasts, and around one in ten. So if 70-something
per cent of that one in ten is correct and the 90 per cent not
attempted are "wrong", then that's a "correct" rate of just under 8
per cent.

But there's a flaw in that calculation.
You cannot assume that the 90% 'not attempted' would have been wrong
- a certain proportion would have been correct simply by chance.
(just like the 'professional forecasts )