If the atmosphere were self-contained, without outside influences, I could buy that, Matty. Then it would be like a coin tossed at random.

But some of the influences on the atmosphere seem to be a lot less random than the thumb flicking the coin. Some ocean circulations seem to take years, even decades, to complete their cycles. If you believe the effects of volcanic activity and solar activity are important too, then they can also persist for years.

The point I'm trying to make is that you don't have to treat every coming of winter as a fresh, random event the way you do the toss of a coin. There seem to be some influences out there which last longer than 12 months: the hard bit is determining which they are, what their effects are and how long they last.

Fair enough. But "probability based on historical average" is a little more informed than "a coin toss". But it ain't a forecast, I agree on that.

The coin toss was just an analogy. Clearly forecasting the weather doesn't fall into just two categories. The point is that there is currently no higher probability of one type of winter over another. All historical data does is give us our probability figures, no more, no less.

Regarding seasonal forecasting in general, I actually don't think it will ever be possible with any remote degree of accuracy on a consistent basis. Tiny variables all over the planet change entire weather patterns over time. I doubt we will ever have a good enough handle on this to be consistently accurate over such a distance of time. We cannot forecast the weather for tomorrow with 100% accuracy.

 I'm also no fan of percentage probabilities. For a start we simply don't live long enough to see enough winters to verify whether the percentages have worked out. And what difference does it make to central authorities' planning that the chance of another 1963 is 1% or 10%? Surely you should plan for the worst possible scenario regardless.

The best I can do is to look back and see whether there are certain conditions which might be necessary for an extreme winter to occur. For instance, the three epic winter spells post-War in the UK (Feb 1947, Jan 1963, Dec 2010) have all occured during negative phases of the PDO and been preceded by either a very late and weak US hurricane season or a completely absent US hurricane season. We have those two preconditions right now, so this, imo, is a winter in which something epic MIGHT occur, while in most years it very likely wouldn't. How likely is something epic? Dunno. We don't have anywhere near enough years of records, never mind understanding fully the mechanisms.

All imo, of course.

As with any situation that requires planning, it is always the right move to plan for the worst. Standard planning.

Well, classifying a season after it's happened against a long-term average is straightforward. But I feel that calculating whether forecasters got their pre-event probabilities correct needs a lot more than even 100 forecasts and outcomes to determine.

To illustrate: Let's say you have the data for all the 20th century's UK winters. You know what constitutes an average UK winter over that period. Starting in 2001 you start making seasonal forecasts against those averages. You could forecast 33% below-average, 33% approx. average and 33% above-average one year and it turns out above-average. Next year you go 40%, 40%, 20%. It's above average again. Are those forecasts right or wrong? They both included the correct outcome among their probabilities. How many forecasts and outcomes do you need before you can make a robust estimate that your forecasting method is offering percentages which are accurate? Personally I'd want 100 goes at the very minimum, and preferably a lot more than that, But we'll all be dead by then.

You might feel you can tell whether forecasters are on the wrong track with their probabilities with fewer results to check than that. But I think most people's experience of life, investments or supporting football teams is that odd results, ill luck or good fortune can persist for way longer than you could plausibly plan for. 

I'm still waiting for the definative and ultimate winter long range forecast - TomC's Winter Tyre Choice!

The polar Presutti winter 😱

There's your mistake. You've altered your prediction for the following year based on what happened the year before. Don't stray from the hundred year average. Doesn't mean it'll be an average winter, but if you start making future predictions based on small samples of a year or couple of years you're always going to come unstuck.

I'm not sure you're understand that we already have 100 years of data that averages are based against? Obviously this is a rolling contract and as each year of OUR lives passes it will lead to small changes in the data. That's how it works.

No. I wanted to avoid that, so I deliberately offered an example where a forecast ignored what happened the year before. After an above-average winter the probability of such an outcome actually fell in my example from 33% to 20% in the next forecast.

I wanted to make the point that assessing even "clean" probabilities forecasts, where the forecasters haven't panicked in the face of yearly fluctuations, is impossible. 

Unless you make one outcome a 0% chance, no individual probabilities forecast can ever be labelled "wrong". The question is how many outcomes do you need to see before you can label it "good" or "bad"? Too many to make them of practical benefit if you're talking about seasonal forecasts, I think, as you only get one result a year.

I've always liked the loaded dice analogy... imagine sticking miniscule bits of blu-tak onto one side of a dice. The chance of landing on the blu-tak side increases, but all those other faces remain perfectly achievable too.

Of course that makes definitive long range forecasting that verifies very closely a feat that may well lie beyond human capability (we have to program the computers in the first place) for the forseeable future.

Yup, that's the best we'll get.

Certain global weather conditions in the lead up to any given winter will slightly increase the likelihood of a prevailing general set-up occurring, but nowhere near to the extent that any sort of accurate forecast could be made. Not least because even if a certain synoptic set up dominated a winter, it may be in the 'wrong' position/orientation to impact the UK in the way many would expect.

Good debate guys, interesting and enjoyable reading, thats what makes these forums and winter such a great place to see out the dark months.

Ta, kenny. I was worried I was going all obsessive and off-topic.

I really can understand Matty's view that you might as well just stick to the historical averages when making a seasonal forecast. In practical terms I wouldn't be surprised if you turned out to be correct more often than anyone attempting anything more detailed. If nothing else, three months is in itself a long term, and even winters which include extreme spells will trend back to the average in the final analysis.

But ... you'd have to be terribly stone-hearted not to be at least curious whether something more precise than that might be feasible. Matty himself is imo, or he wouldn't be on this thread. And ultimately I can't believe god plays dice with the atmosphere.

Yes, but statistical probability and the real world are 2 very different things.  

As the Monte Carlo Fallacy can prove....but then i'm sure you wouldn't like 25 odd cold winters in row Matt ;)  

I would - that's why it's called Winter... 

Agreed , you'd be happy if this was the norm

http://modeles.meteociel.fr/modeles/cfs/run4/cfsnh-0-2658.png?18

It's been doing that for weeks , very few charts with a snowfest scenario

One model though blah blah blah

Being fair he only tweeted that an hour ago , this a day after he posted

To conclude the discussion I'd say that on balance the chances of a close to average or colder than average winter are slightly higher than the chances of a milder than average winter. However, remember that developments in the weather during the rest of autumn could well mean the TWO winter forecast will reach a significantly different conclusion.

Doesn't mean he will be right of course , just one opinion .

Is one to assume that by 'developments in the weather during the rest of the Autumn' Brian meant 'during the next day or so' ? 

So it really begs the question - why issue an ultra LRF one day and then issue statements that it probably isn't going to be right the next? Illustrates how fickle the long-range world can be. Scientific and reasoned argument for and against the probability of certain patterns I agree with, but if you're foolish enough to actually make a forecast please don't change it - whether it be by re-issue or tweet or rumour or smoke signal - every day or so otherwise it makes the original forecast completely useless.

Brian's initial outlook of fairly cool/cold, but not that much in the way of snow may be a good one if the AO tends to be positive, but the Atlantic Sea Surface Temps act against that promoting blocking.  In that scenario we'll probably see a fair amount of high pressure, but without the AO playing ball, it'll be hard to get the cold air advection to the UK.  Even in that scenario there's always the chance of a week long arctic/siberian type blast.

Matty's post above makes a good point though 

So...mild winter it is then.

All joking aside, as Matty says time and time and time again, we just don't know what's going to happen in winter 2013-14. SO many variables and not to mention the systems that can cause a sudden cool down, such as the fabled SSW etc etc.

Have to add though, this October has felt quite normal to me, very Autumnal in the last week, although maybe lacking a few gales.

The zonal train has fired up in a way it hasn't really done in recent Octobers, and that feeds my mind into thinking we will get a bog-standard "normal" winter....That's just guess work of course, but we can't keep having "The coldest winter in 100 years" as the Express bleat on about EVERY year.

Trouble is, they won't sell many papers with headlines such as: "Britain facing bog standard cool to mild winter with rain."