Better Than The Models ?

[Snowmad79]

Thanks for that Roger

I'll spend some quality time reading futher into this including your thread on the other fourm. Seems a hell of alot to take onboard at first glance.

[Timmytour]

By showing that his method holds water consistently?

Surely, as I think he is quite clear in saying, it's not worth getting into his methods if he can't prove his forecasts are accurate?

I think his approach is admirable....he thinks he's onto something....he tests it out which involves two detailed forecasts to date involving relative extreme events, plus one at Christmas which he was asked to make without previously suggesting he could do this.

If this next one comes off he hasn't proved anything, but he's certainly a lot further down the road with the thought that he might be onto something.

He could do this all in private, but I'm glad he's putting it out in public.

[La Bise]

On the contrary, when you claim to hold a method that has the potential to basically predict anything that can happen on Earth, you better lay down a few cards rather than some gnomic utterances about the fourth dimension and ancient wisdom. That could enable others to conduct their own experiments incidentally.

Open knowledge and all that...

[topher41]

New user here , Now I dont post on here i just observe as I have next to no knowledge of weather forcasting

One thing i dont understand is why MB bothered to put out his forcast.

if this is the response I thought i would recieve whith trying to share something I wouldnt

why cant you just let whatever method MB uses prove or disprove itself befor the dissection

[k19]

New user here , Now I dont post on here i just observe as I have next to no knowledge of weather forcasting

One thing i dont understand is why MB bothered to put out his forcast.

if this is the response I thought i would recieve whith trying to share something I wouldnt

why cant you just let whatever method MB uses prove or disprove itself befor the dissection

Its not that we are trying to dissect the theory, MB should keep doing forecasts in my opinion- especially if this one comes off, however if no information is given out, and perhaps, at his own admission, his relatively new knowledge of how the weather works may cause an upset here and there, thus altering the forecast, possibly causing it to be wrong. If some minor details were given out then people like roger and BFTP might be able to help refine the method before it's dismissed. It would be a huge loss to our understanding if he gets this forecast wrong, and then gives up when it may be that a few slight details- which could be provided by people here- were missing.

[Interitus]

MB,

I am interested to see your 'forecast' but if you won't discuss your method, your workings, or your data sources then you have little credibility in my eyes.

If you now refuse to post any more 'forecasts' unless we stop asking you about 'your method' then that further reinforces my doubt.

I'm sorry, but I'm not willing to worship your ego until you can show me some science.

Pieman.

[PhD in Meteorology and lifetime weather observer]

If he is consistently on the right lines then that would be credibility enough; how it is done is irrelevant, even if he read it in tea-leaves.

[topher41]

Fair point well made K.100

could make for and against all night .Myself im just going to wait proof is in the pudding as they say

[MurcieBoy]

I go away for a couple of hours.....

Right first things first, many thanks for the supportive posts, they are very much appreciated. Thanks also to Coast for his reply to my last post.

As I have always said, and I think we all agree here, any method for weather prediction should be judged by its results and not by anything else. That is why my forecasts are very detailed, precise and clear. I don’t want the focus at this stage to be on me or any method I may be using, I want the focus to be on the forecasts. I want the forecasts to do all the talking.

Also, as I have always said, if/when I am happy the forecasts have proved accurate and consistent, I will start to write the book. I am not writing it now (or talking about it in any detail), as the forecasts have proved nothing. I don’t think anything will change my mind, sorry folks.

As to giving up on the method if I get a major failure, I don't believe that I will ever give up the research (as it is my passion; and has been for a very long time).

[Paul Sherman]

Hi MB

The one nagging question I have is at what point do you call the Forecast a Success, Eg What Span of Time and How Far away do the Synoptics need to be. I am still quite amazed that some on here are calling you 2 for 2. I was very Impressed with the November Forecast although the GFS Did also show a Potent Storm on the 28th October at T360 So Kudos to that Model for showing similar synoptics but not in the same place as your Forecast Namely the GFS Had the Storm between Iceland and Scotland whereas yours was to the North East of Scotland, both you and GFS Showed the LP To be in the 970 to 975 Range when in fact it was between 950 and 955mb

Now the thing that is bugging me is your Xmas Effort was not only a few days out but again Pressure was quite a considerable way out with 1023mb instead of your Predicted 1045mb, also there were no Gales in NW Scotland on Xmas Day Itself, your predicted temps of 5-7c were again around 6-10c Out of Range with temps of 1c in the South and -3c Midlands Northwards.

So my question has to be what sort of error range both in Days out, Pressure, Winds Speeds etc would you class the Forecast as a Success. I know this is a good 4 weeks out and we cant prophesise anything yet but just Interested in what the range is going to be.

As we all know a Much Loved Forecaster whos name ryhmes with Diers claims success on Forecasts that are not only weeks out but also if the Synoptics affect different countries, even Continents, That Forecaster has predicted the Coldest January since 1740 but somehow I think he will Claim success with December claiming the bragging rights and it being sooo close to January....etc etc....you get my drift I am sure.

Thanks for the Reply By Pm Re The US Tornado Question - Will get the History to you asap

Paul S

[dinger]

Its not that we are trying to dissect the theory, MB should keep doing forecasts in my opinion- especially if this one comes off, however if no information is given out, and perhaps, at his own admission, his relatively new knowledge of how the weather works may cause an upset here and there, thus altering the forecast, possibly causing it to be wrong. If some minor details were given out then people like roger and BFTP might be able to help refine the method before it's dismissed. It would be a huge loss to our understanding if he gets this forecast wrong, and then gives up when it may be that a few slight details- which could be provided by people here- were missing.

I agree that it would be nice to know his methods, curiosity being part of our nature.

However to invite others in to refine or dismiss a method as yet undisclosed is like IBM inviting Bill gates to view revolutionary software prior to applying for the patent.

MB is merely proofing a method of forecasting. He does not need to disclose anything.

All of us with an interest in this will no doubt be asking further questions should the forecast verify and once again I think MB will keep his methods private.

If he continues to be accurate with further forecasts then he will be worthy of any rewards he might receive.

If his forecasts fail miserably then he has lost nothing other than time.

Let's wait and see what happens

[Paul Sherman]

Yes

It is Interesting that there are two camps of People in this thread, those that want to know the Methods and those that are not really Interested in the Methods but just Sceptical at such extreme Events. I am in the camp that says why should he put up the methods, but also agree with Coast in saying why put it out there at all and wait a few years to see how Plotted dates pan out and then post the findings, I guess the latter would have non believers saying "Anyone could have said that" etc etc

I guess we will all find out in 4 weeks time, cant wait to see what will happen if the 3 Main Models start plotting something extreme in the Models at T240 For Instance.

Paul S

[BarometerBunts]

Hi,

This is just an observation. MB, I think I understand where you are coming from in terms of testing your methods and sort of understand why you wish to concentrate on detailed forecasts and not reveal too much at this stage.

However, I can also understand why some on here appear frustrated at the lack of information, albeit some have been more tactful than others.

Whilst I appreciate that you have explained you are very busy, I do think it may be prudent, if you are serious about testing your accuracy, to seriously contemplate more regular forecasts, at regular intervals and maybe not restrict to extreme events. For example, over on the model discussion forum, there is frequent argument about FI, the period of time which is technically outside the realms of reliable forecast. Would it be possible for you to perhaps forecast five/six days ahead. If you chose a set day and time to do this on a regular basis, it could be controlled and you would have the benefit of this forum to monitor and assist in terms of data collection.

This way, your accuracy could be monitored and it would give you more of an idea as to whether your method needed tweaking (or not)

[La Bise]

Excellent comment Paul.

I'm expecting nothing less than a spot-on forecast, the man himself said he was nearly 100% sure about it, as he had lots more control variables to verify. It's all written on this thread, including the much more recent lowering of expectations when faced with a bit of probing. I've given up on having an explanation being given, it's quite clear nothing will be said aside some gnomic utterances which I'm not even convinced are actual clues.

I wish people would stop confusing a desire to understand what seems to be outright quackery (do any of you really believe all events on Earth can be predicted, from a persepctive of the "fouth dimension" thank to some "ancient wisdom"...?) with being close-minded and unreceptive.... Do not confuse "open minded" with "gullible". I'm open minded to new ideas, I find RJS comments about his method very interesting and it's clear he has worked a lot on it, testing it and accumulating material. I'll leave to others to analyse it and declare it good/bad.

Anyhow, stil interested to see how that forecast pans out

[VillagePlank]

I wonder what the price of cotton is going to do this year and whether it could've been predicted in 1967?

Thanks for the negative rep-points, guys

Now, with all this talk of dimensionality and looking for natural patterns, perhaps you'd like to use your favourite search engine, and look up "mandelbrot, cotton prices, 1967" There you will find out about fractals which were first used to study non-Gaussian distributions of non-linear time-series, and the key-point was self-similarity at different scales denoted by a dimension term. I had some limited success with such techniques with 850hPa temps about a decade ago - unfortunately, it failed the normal statistical tests so gave up on it.

You'll have to excuse me if I'm wrong here, but wasn't the "method" derived from analysis of the stock markets?

New user here , Now I dont post on here i just observe as I have next to no knowledge of weather forcasting

One thing i dont understand is why MB bothered to put out his forcast.

if this is the response I thought i would recieve whith trying to share something I wouldnt

why cant you just let whatever method MB uses prove or disprove itself befor the dissection

Welcome to NetW!

... because without knowing the "method", it cannot be shown to be the case either way re: standard confidence tests. And if it's not fractals, not some sort of wavelet analysis, nor Fourier analysis, from what I can glean it must be more akin to divination.

Edited January 7, 2011 by VillagePlank

[jethro]

Welcome to NetW!

... because without knowing the "method", it cannot be shown to be the case either way re: standard confidence tests. And if it's not fractals, not some sort of wavelet analysis, nor Fourier analysis, from what I can glean it must be more akin to divination.

Isn't that akin to saying "if it's not displayed in a way that makes sense to me, then it's all bunkum"?

[VillagePlank]

Isn't that akin to saying "if it's not displayed in a way that makes sense to me, then it's all bunkum"?

Not at all; the "method" can't be tested because we don't know it, and from what we do know, there is plenty of (scientific) literature out there that cover the cases, and they've all not done so well.

You can't do confidence tests (chi-squared etc etc) without knowing the method, for example, since we can't determine degrees of freedom. As I've stated elsewhere, zonal systems bombing down the North Sea are hardly uncommon for the UK - so the only exceptional part of the forecast is the >100mph winds (etc) or the extremity of the forecast.

In this case how to do we test the forecast? Do we say it is a hit since low pressure is in more or less the regions stated? Do we allow for, say, 10% pressure reading error? Do we measure gust speed in the favoured locations? Do we do all of the above and if any fail, then the forecast is a dud? Actually, if MB could supply us with the parameters for testing the forecast, it might allay fears, somewhat, but, unfortunately, he's gone for a near 100% ...

Surely it's easier with normal common or garden statistics that are known to be effective?

Edited January 7, 2011 by VillagePlank

[jethro]

I'd have thought in these circumstances it comes down to whether or not it happens. It could be argued that SW winds and Atlantic driven rain like the weather we're experiencing today happens all the time, but we don't actually get it all the time - someone accurately predicting that kind of weather on any given day is good enough for me.

If a judgement of method for testing purposes is to be had then a simple comparison with a benchmark system like the METO forecast is probably good enough - what margin of error do they work to in percentage terms of accuracy of forecast?

How the forecast is reached, what is studied to enable a forecast to be made is irrelevant IMO; accuracy of predicted events happening is the only measure needed and that can only be done retrospectively.

[VillagePlank]

If a judgement of method for testing purposes is to be had then a simple comparison with a benchmark system like the METO forecast is probably good enough - what margin of error do they work to in percentage terms of accuracy of forecast?

Well, colloquially, the MetO will say it can't be done, and, therefore, any hit from a distance of four weeks or more is significant. I think it can be done, and therefore normal scientific parameters for testing should be used.

How do we measure a forecasts accuracy? What is the domain of enquiry? The terms of reference? In the view of the experimenter what constitutes a successful forecast? Or a bad one? Is it independent? Crucially, is the experiment falsifiable? The conditions for success or failure, especially in the absence of "the method" should be issued at the same time as the forecast in my view - and it's fairer, since an error margin can be assigned (spacially, or temporally) given the evidence accrued by previous forecasts (so tweaks to the "method" can be quantified, properly)

Retrospectively looking at this chart or that chart and subjectively determining that they kind of look the same, in my view, is simply not enough to verify nor validate a forecast. For sure, if it's windy people will prop up the local hostelry and go "wow, that was close" or "hey, I think he's on to something" but importantly, a failed forecast is not enough to show that he isn't, and, respectively, a successful forecast is not enough to show he his.

Edited January 7, 2011 by VillagePlank

[MurcieBoy]

Paul (#159)

I hear what you say. Seriously, I would not claim a “hit†if it was not there, no point. I have no commercial angle on this, my interest is in getting it right and failures need to be met head on.

By the way, the Xmas forecast posted on here (#15) is the one first one I posted up on ukww, there was another issued a few days later with charts etc. It is covered in the #1 Video.

Paul, have you had a squint at the #2 Video which is the candid appraisal of the Xmas forecast? There is also an interesting comparison with the GFS model charts which started 27 days after my forecast.

It’s far too early at this stage to see the upside of the “methodâ€; as it is being “tested†and hopefully refined. You can however see the worst sides to it (at its current level of evolution).

Re error range: it has to happen on the day forecasted. None of this airy-fairy nonsense “it can happen ± 3½ days from the day forecastedâ€. If it has not happened on the day forecasted it’s a fail! As it’s rightly pointed out, the “method†has been derived from a need to trade in Dow futures, where timing is everything.

[jethro]

Well, colloquially, the MetO will say it can't be done, and, therefore, any hit from a distance of four weeks or more is significant. I think it can be done, and therefore normal scientific parameters for testing should be used.

How do we measure a forecasts accuracy? What is the domain of enquiry? The terms of reference? In the view of the experimenter what constitutes a successful forecast? Or a bad one? Is it independent? Crucially, is the experiment falsifiable? The conditions for success or failure, especially in the absence of "the method" should be issued at the same time as the forecast in my view - and it's fairer, since an error margin can be assigned (spacially, or temporally) given the evidence accrued by previous forecasts (so tweaks to the "method" can be quantified, properly)

Retrospectively looking at this chart or that chart and subjectively determining that they kind of look the same, in my view, is simply not enough to verify nor validate a forecast. For sure, if it's windy people will prop up the local hostelry and go "wow, that was close" or "hey, I think he's on to something" but importantly, a failed forecast is not enough to show that he isn't, and, respectively, a successful forecast is not enough to show he his.

With respect VP, what utter tosh.

We're not talking about looking at charts here, we're talking about a verifiable forecast - it either happens or it doesn't.

The only reason why it would need to be a falsifiable experiment is if other people/organisations were trying to forecast weather via the same methods - why would they need to? If MB or anyone else comes up with a system which is as accurate, or more accurate than established methods of meteorology then my advice would be "keep schtum" on the methods involved. Everyone's got to earn a living, why on earth would you dilute your earnings potential by sharing the information?

I can understand that it may be fascinating to understand and try to pin a scientific explanation for predicting something as unpredictable as the weather but not having access to that knowledge doesn't mean the prediction can't be done.

[nking]

Well, colloquially, the MetO will say it can't be done, and, therefore, any hit from a distance of four weeks or more is significant. I think it can be done, and therefore normal scientific parameters for testing should be used.

A bold statement which I'm sure you will back up with some "proof of concept" yourself..... a forecast perhaps.

[VillagePlank]

With respect VP, what utter tosh.

We're not talking about looking at charts here, we're talking about a verifiable forecast - it either happens or it doesn't.

The only reason why it would need to be a falsifiable experiment is if other people/organisations were trying to forecast weather via the same methods - why would they need to? If MB or anyone else comes up with a system which is as accurate, or more accurate than established methods of meteorology then my advice would be "keep schtum" on the methods involved. Everyone's got to earn a living, why on earth would you dilute your earnings potential by sharing the information?

I can understand that it may be fascinating to understand and try to pin a scientific explanation for predicting something as unpredictable as the weather but not having access to that knowledge doesn't mean the prediction can't be done.

A lot of people are banding about whether 'it happens or it doesn't'

My point is ... whether what happens or doesn't? How can we tell? I've watched the videos and, for sure, it's a weather forecast alright; and as far as I can see the only test possible at this stage is to overly the chart for each day on the chart forecast. The chances of that happening are absolutely miniscule.

So, in the interest of fairness, how much lee-way do you want to give? 0.5 degree long/lat for the centre of each LP, 10% +/- max gust speed for location's mentioned? What about temperature? How about broad synoptic pattern correct? (HP going SW, with GIN corridor become hyper-active? - which is probably the fairest affirmation that we could do, actually)

All experiments should be falsifiable: it is about constructing a null-hypothesis with the alternate: and we can only do one half of the hypothesis, here.

I take onboard your point about earnings, but as I understand it, this is not an exercise in making money, since the same method works on the DOW futures markets, and if he really wanted to do that he would restrict it to there - I think I am right in stating that this is about saving lives.

How about some testable parameters four weeks in advance? I can't really believe that you want a direct overlay of the charts for the day against the forecast charts. For sure the strictest test possible, but also, probably, the unfairest as, I suspect, that even the MetO don't get that some 4/5 days in advance.

A bold statement which I'm sure you will back up with some "proof of concept" yourself..... a forecast perhaps.

See LI thread in archives. Let me know what you think ... am currently rewriting for my blog with new stuff (and more accurate forecasts) but will be some time before it is in any sensible form. It's a theory, and back in 2008 it predicted that about now the CET would plummet whilst global temperatures kept going up. It's all archived away somewhere, including all of my errors, misjudgements, the theory, the maths etc etc etc. It passes historical modelling statistical tests, as I have described here, but since it works on annual data, there is not a big enough sample set (three years) to derive any useful test of it's predictive power - it needs to go monthly, as it were - since I will too old to care, otherwise.

Edited January 7, 2011 by VillagePlank

[jethro]

A lot of people are banding about whether 'it happens or it doesn't'

My point is ... whether what happens or doesn't? How can we tell? I've watched the videos and, for sure, it's a weather forecast alright; and as far as I can see the only test possible at this stage is to overly the chart for each day on the chart forecast. The chances of that happening are are absolutely miniscule.

So, in the interest of fairness, how much lee-way do you want to give? 0.5 degree long/lat for the centre of each LP, 10% +/- max gust speed for location's mentioned? What about temperature? How about broad synoptic pattern correct? (HP going SW, with GIN corridor become hyper-active? - which is probably the fairest affirmation that we could do, actually)

All experiments should be falsifiable: it is about constructing a null-hypothesis with the alternate: and we can only do one half of the hypothesis, here.

I take onboard your point about earnings, but as I understand it, this is not an exercise in making money, since the same method works on the DOW futures markets, and if he really wanted to do that he would restrict it to there - I think I am right in stating that this is about saving lives.

How about some testable parameters four weeks in advance? I can't really believe that you want a direct overlay of the charts for the day against the forecast charts. For sure the strictest test possible, but also, probably, the unfairest as, I suspect, that even the MetO don't get that some 4/5 days in advance.

See LI thread in archives. Let me know what you think ... am currently rewriting for my blog with new stuff (and more accurate forecasts) but will be some time before it is in any sensible form. It's a theory, and back in 2008 it predicted that about now the CET would plummet whilst global temperatures kept going up. It's all archived away somewhere, including all of my errors, misjudgements, the theory, the maths etc etc etc.

I'm not a model watcher but I do know there are various models, throwing out various runs, numerous times a day - rarely, if ever do they accurately predict future weather down to the teeniest detail. Whatever margin of error they are afforded is, to my mind, an acceptable margin of error allowed universally, for all predictions. Including this one from MB.

You say in your opinion weather CAN accurately be predicted; I know you've put a great deal of effort into the LI thread over a substantial period of time, you wouldn't perchance be wanting to pin MB down on his methods so that you may add the additional information to augment your own understanding for your LI theory?

[VillagePlank]

You say in your opinion weather CAN accurately be predicted; I know you've put a great deal of effort into the LI thread over a substantial period of time, you wouldn't perchance be wanting to pin MB down on his methods so that you may add the additional information to augment your own understanding for your LI theory?

Alas, natural cycles (ie those that can discerned distinctly, and extrapolated forward - which, I presume, is what MB is doing here) fail dismally in the LI energy exchange. I didn't think you'd take me for a cynic, J?

I'm not a model watcher but I do know there are various models, throwing out various runs, numerous times a day - rarely, if ever do they accurately predict future weather down to the teeniest detail. Whatever margin of error they are afforded is, to my mind, an acceptable margin of error allowed universally, for all predictions. Including this one from MB.

OK - but I've no idea what the universal margin of error is ....

Edited January 7, 2011 by VillagePlank

[Paul Sherman]

Paul (#159)

I hear what you say. Seriously, I would not claim a “hit†if it was not there, no point. I have no commercial angle on this, my interest is in getting it right and failures need to be met head on.

By the way, the Xmas forecast posted on here (#15) is the one first one I posted up on ukww, there was another issued a few days later with charts etc. It is covered in the #1 Video.

Paul, have you had a squint at the #2 Video which is the candid appraisal of the Xmas forecast? There is also an interesting comparison with the GFS model charts which started 27 days after my forecast.

It’s far too early at this stage to see the upside of the “methodâ€; as it is being “tested†and hopefully refined. You can however see the worst sides to it (at its current level of evolution).

Re error range: it has to happen on the day forecasted. None of this airy-fairy nonsense “it can happen ± 3½ days from the day forecastedâ€. If it has not happened on the day forecasted it’s a fail! As it’s rightly pointed out, the “method†has been derived from a need to trade in Dow futures, where timing is everything.

Hi MB

Thanks for the reply, no I Did not see the 2nd Video so that clears things up nicely.

And thanks for clearing up the error range, that gives us a much clearer idea of where expect forecast verifications to settle.

All eyes on the next few weeks