Thursday, February 4, 2010

Climate bet

On a more relaxed tone.....Our guest blogger Gabrielle Hegerl is willing to accept bets on the evolution of the global mean temperature in the next 7 years. Here are the terms of reference:

[Update:  For Gabi Hegerl,   to win or lose the bet  is not a proof of or against anthropogenic global warming. ]

Definition 1: Event A
-one of the years Y=2010,2011,2012,2013,2014, 2015, 2016 or 2017  is reported to have the highest mean annual global temperature of the period between 1880 and year Y (in other words, the year Y is the warmest on record) in the data set HadCRUTv. or successor data set. Successor data set is the data set  used by  the Hadley Centre to compare hottest years to the media at that time.

Definition 2: accept the bet
Send a comment with nickname to this posting in Die Klimawiebel with the text: 'I accept the bet', and send an email to eduardo.zorita at including the nickname and  real name and real postal address

-if event A does NOT occur, Gabi Hegerl will  send by mail, within year 2018, one nice bottle of wine (US definition = more expensive than $10 ) to each one of the first 10 readers of Die Klimazwiebel accepting this bet. Preference ranking is determined by the blog time-stamp on the comment.

- if event A occurs, each one of the readers of Die Klimazwiebel that accepted this bet will send by mail a bottle of nice wine (US definition = more expensive than $10 ) to Gabi Hegerl, not later than one year after event A was declared to have happened.

All conflicts will be irrevocably resolved, after hearing the parties, by eduardo. All breaches to the terms of this bet will be published in Die Klimazwiebel, including real names.



P Gosselin said...

The bet:
"One of the years Y=2010,2011,2012,2013,2014, 2015, 2016 or 2017 is reported to have the highest mean annual global temperature of the period between 1880 and year Y (in other words, the year Y is the warmest on record) in the data set HadCRUTv. or successor data set."

This does not mean that the warmest year on record will actually occur. It only means that HadCrutv will report that it was the warmest. That's a big difference.

Of course HadCrut will find a statistacal way to make a year the hottest, if not all the years!

Use UAH MSU instead and you'll find people to accept this bet(since 1979). I would.
And even if it ended up being the hottest decade, it would prove nothing accept that earth is getting warmer. It would say nothing about the cause.

MikeR said...

Hmm. What do you-all think of this bet? It confuses me. Given that recent temperatures are very close to the highest since 1880 already, wouldn't we expect one of the next eight to be the highest just by random chance?

If the point is to bet on AGW, I'd suggest that at least one of these years must be at least ___ degrees warmer than any year since 1880. Or better yet, that the mean of all eight should be?

Hans von Storch said...

Folks, do not be shy - this is an offer, which is either accepted or not.

Of course you can offer a counter-bet, but make sure that the conditions are unique. For instance, for P. Gosselin - a bet that in the Hadley Center record all but one years will be hotter than anything before. Or that a certain lower level will be passed. But it must be verifiable in an objective way when the time has run its course.

P Gosselin said...

I will not accept any bet that is decided by Hadley or GISS. But I'd like to kindly make the following generous counter-offer:

I bet that the next 5 years, 2010 -2014, will be cooler on average than than the average of last 10 years 2000 - 2009. For this we will use the data from UAH MSU, Ch 4.

I'm proposing the next five years because the chance of us still being around and healthy are better (I'm not getting younger).

Bet amount: A bottle of wine not to exceed €20 in value. This bet is valid only for the first person who accepts it. The offer then becomes void to all others once it is accepted by a person.

PedroS said...

My bet proposal:

I bet that the trend obtained from linear regression of at least one of the following temperature sets (UAH, HadCRU, GISS) in the 2001-2014 timeframe will not be positive (P<0.05). $10 wine bottle to Gabi Hegerl, otherwise. :-)

eduardo said...

Dear Pedro,
your bet is not well defined. Statistically significant under which null-hypothesis (white noise, red-noise, long-memory...)?

Martin Heimann said...

Reminds me of the time we betted in December whether the Alster will freeze or not. The criterion was if the police would officially open it. In the 80's the odds were about 1:5.

I take Gabi's bet - under the condition that we drink the wine together.

As to UAH MSU, Ch 4 as criterion: That can be fouled, too, e.g. by deteriorating sensor or faulty satellite (or yet the detection of a new processing software error).

Anonymous said...

Seems to me that this is a very cleverly designed bet.

Gabi has restricted her payout in the event she loses to 10 bottles, worth $100 at a price per bottle of $10.

On the other hand, if she wins, and there are 100 participants, then she wins 100 bottles.


P Gosselin said...

No takers!?
UAH MSU has started with +0.735 or something! I'm giving you a head start.

eduardo said...

@ 8

No, the number of participants is limited to 10 in both cases

PedroS said...

Dear eduardo,

in my bet proposal, I meant the simplest linear regression possible: no auto-correlation, etc. Just take the annual global temperature and find the best straight line fit
y=ax + b.

Then, find confidence intervals for the slope (a), as oulined in, e.g.

In my bet, I mean that in at least one of the temperature sets, the null hypothesis: trend (2001-2014)<=0 cannot be rejected at the P<0.05 level.

Hans von Storch said...

Concerning the Hegerl-bet, I am rather disappointed by our categorical nay-sayers. None of them (Martin Heimann is not part of that group) has accepted the bet, meaning that they all consider it not only possible but eben probable that one the next 7 years is warmer than what we have on the thermometer record so far. Some doleful participants try to excuse themselves by claiming that the defined reference, HadCRUTv, would be and will be systematically flawed. The reference to HadCRUTv was put it to make sure that we eventually will not run into technical problems when we want to decide the bet.

Some suggest modifications, such as Pedro S. But his version is rife for technical discussions, when he claims: a significant trend in 2001-2014. He does not say which assumptions he wants to apply for the serial dependence of the data (the reference for testing significance of trend is assuming "The Y values are independent.", which is not the case in case of global mean temperature). Why not, Pedro S - just relating the 2001-14 trend to that given critical trend of X deg/year, which you want to assign to the 5% risk. I guess to detect a trend with a risk of less than 5% with serial correlation, as they prevail in such data, would result is a rather big trend. Which would also spell out which trend, namely one at 10% risk level, you would find plausible. Or even better - since 2001-2009 is already known, how big would the trend in the remaining 5 years have to be? In x deg/year? Also, Pedro S, say, which data.

The trick is to formulate a bet, which employs quantities, which will only be known in the futures, and to define watertight technical terms for verification. Gabi Hegerl's bet is of that sort.

The bottom line is: nobody of our critical friends dares, in particular not those pathetical anonymous participants - why? It is just a little sporty exercise!

For the sake of fairness, I should tell you that there will be a press release after a while, telling the public about our bet - including the response of our critical fellows, just in case that none of them is brave enough to accept the bet.

P Gosselin said...

Okay, Prof von Storch, I accept the bet. It is, after all, only one 10-dollar bottle of wine. And hopefully HadCrut will have gotten its house in order by then.
I will send my addess shortly.

Anonymous said...

Bjorn says:

@Hans and Eduardo,

Gabi's bet ist contradicting your own spirit of research. When some people try to contradict AGW with one cool summer, the reaction always is "Oh, climate is a matter of decades, not of individual years!" Rightfully so! Hence, I would rather bet on a five or ten year average not being warmer than a SINGLE year, say, 2000.

One single hot year in Gabi's bet has a probability of around 50%, I would guess. This is quite a sporty bet, but I take it. And I would like to double. Gabi will have a good bottle of St. Emilion in case she wins. In extension to Martin Heiman's idea, I propose a joint party at the loser's cost with good red wine and food, and a possibility to meet and discuss in person. I'll offer to have this party at my home (close to Frankfurt), but any other proposal is welcome. See you in 2018!

eduardo said...

I feel that the 'climate' is pretty tense these days. A bottle of wine can elicit such laden responses. Perhaps the wine has completely evaporated in 2018 under greenhouse warming.

I just came from shopping . The employee in the supermarket told me (she doesnt know me personally):'... these climate researchers.. all swindlers !', so I paid my sausages and evaporated as well.

Hans von Storch said...

P Gosselin/13 & Bjorn/14 - Exactly, it is just 10$ - and likely a lot of fun! Danke für den Sportsgeist.

Werner Krauss said...

to Eduardo:
the employee was not exactly right: climate researchers are not swindlers, but they are gamblers.

eduardo said...

I have included a disclaimer that does not change the terms of the bet

Unknown said...

The employee in the supermarket told me (she doesnt know me personally):'... these climate researchers.. all swindlers !', so I paid my sausages ...

you paid? Oh man... trying to be honest is a mean trick by you, you, you, climate researcher, you. Actually, that's trick that "skeptics" never try. Last Lie: NASA deliberately crashed its CO2 measurement satellite. Oh je...

@P Gosselin
just try to download some data (the data that is allowed to be published freely) and some source code of CRUTEM3

may be you find the fraud... oh je.

Tobias W said...

After having thought long and hard - and having lived through the coldest winter since I was a little toddler - I have also decided to take the bet. May the best climate astrologist win:-)!

wflamme said...

Confused ... looking at the update - to win or to loose about guessing whether there might be an outlier in the Hadley yearly series or not will not prove anything about GW/AGW.

Confused ... Gabrielle limiting her losses about not being right to ~100$? I accepted similar risks buying a RAID at ebay yesterday.

Confused ... bets aren't changing but background information is.

Confused ... Hans saying this is nothing but pure sportsgeist. Hans saying there will probably be a paper about this. So there will probably be a paper about a ~100$ bet by Gabrielle Hegerl about yearly temperature series outliers and the sportsgeist attitude of anonymous commentators at Klimazwiebel?

Don't drink and drive.

Rob said...

I take the bet. I'm quite impressed about so much objection raisers amongst the contributors. Is this typical german? Indeed, they/we have a special term for this: Bedenkenträger.
Let us take it sportive. I'm basing my hope to obtain a bottle of wine on Latif's MDO analysis as well as the very low count of sunspots, perhaps low enough to have not too much bias in the rising interval of this cycle.

P Gosselin said...

It's a tie!

Now what? Does Gabi think she is owed a bottle of wine?

eduardo said...

Our target is HadCRUT3, so Gabi is still on hock

eduardo said...

For some unknown reason that this comment has not been published:

James Annan has left a new comment on your post "Climate bet":

"the data set HadCRUTv. or successor data set"

Well of course HadCRUT4 is out and shows 2010 to beat 1998. So the bet is already a win for Gabi!

How many takers were there in the end, and has there been any progress towards a settlement?

eduardo said...


we also have been thinking about this bet, and I think it is not that clear-cut, at least for me. The data set HadCRUT3 is still being issued - the last update that I am aware of is for December 2012 and has been uploaded on January 18th, 2013.

In my understanding of the terms , we would switch to the successor data set once HadCRUT3 ceases to be updated, although I agree that there may be other interpretations. This may be too subtle for non-native English speakers like most of us here. In any case, Gabi has not claimed to have won the bet so far.

Hans von Storch said...

As for the technical process - the post of James went into moderation queue, because comments older than 30 days do so. When I saw it, I let it pass, but I did not check if it really was published. Seemingly it was not. If I made an error, I do not know; if so, I apologize.

Anonymous said...

am I getting this right... the bet is/was about who would have won the champions league if only there had been a goal camera or if only the referee had seen that foul in mid-field or if only the most important player of team X had not been ill that day?

Isn´t there a metric not bound to change after the fact, to avoid what Eduardo pointed at in #25?

eduardo said...


I think your examples are not quite realistic. I would rather compare this situation with a bet on the unemployment in particular country. Nobody counts exactly the number of unemployed people. This number is estimated by different methods, which may yield different results, and they usually do in practice.

In this case, a new method was introduced in the mean time, with numbers obtained with the old method still being estimated.

Probably, it is our fault for not having foreseen this situation. On the other hand, this is assumed to be a bet between , ehem, gentlemen, who should be able to agree on what has actually happened.

@ReinerGrundmann said...

Mike Hulme has some interesting observations about the different ways of determining what is "normal"