The Blackboard is now taking bets on the August UAH reading that will be posted by Roy in September. Note that in July, Roy switched to 2 significant figures. This should lead to more ties and favors early betting. Also, to spread the values bet out, I think I should the betting scripts out early. (I may need to post the September bets in final week of August. Tell me what you think of that idea.
Meanwhile, this months bets close on Midnight August 11.
[sockulator(../musings/wp-content/plugins/BettingScripts/UAHBets5.php?Metric=UAH TTL?Units=C?cutOffMonth=8?cutOffDay=12?cutOffYear=2012?DateMetric=August, 2012?)sockulator]
Bets Close 8/11/2012
Personally I am in favour of closing the bets once the actual UAH figure for the previous month is announced.
This would minimise the advantage (if any) gained by those (including myself) who use actual AQUA CH5 numbers as a guide for the start of the month.
As you say, I think this would have the effect of causing a greater spread of bets, since they would be based on the bettors judgment of how temperatures were likely to change, than if actual figures are used for a large part of the month.
Interestingly, using AQUA CH5 doesn’t seem to help in betting too much. Policlimate seems to do better for that, but it’s not updating lately, as far as I see. If we just use CH5, August is going to be another drop. But one would bet that even before the listing of July, since CH5 has been running quite cool for awhile now.
To remove much of the luck involved in the contest, I think it would be fun to also keep track of entrants accumulated average prediction errors for a series of months: e.g. posting a running average with quatloos awarded quarterly and annually. This could be scored by the RMS error of the predictions. You might also use a quasi-mean error by dividing the sum of squares by n-1 monthly guesses instead of n, thereby penalizing those that forget or purposely withhold an occasional guess – while still leaving them a chance at the annual race – if they are good.
I think this would be a better indicator of who has the best insight into LT temperature trends; afterall, the best golfer is considered to be the player with the lowest score not the player with the most holes-in-one.
Ged, I think that using AQUA CH5 does help to get the predicted UAH at least within the likely range of possibilities and avoid making a bet which is entirely impossible, as would be the case without it.
Of course, global temperatures “normally” fall in August, but that doesn’t necessarily mean a fall in the anomaly, if the fall in temperatures is lower than average.
On the other hand, it is possible that CH5 temperatures may actually rise in August, as they did in 2004, which resulted in a rise in the UAH anomaly from -0.19c to -0.06c.
I think that having the actual CH5 figures for the first 1-2 weeks is a definite advantage over having no figures at all for the month.
Having to bet that early would make the betting a bit more of a head-scratching event, but I’m not entirely sure it would have the desired effect of spreading out the values bet on. Considering the shift to 2 digits is something new for the betting, maybe we could do that next month as a trial and see how it works out.
Or you could contact Roy Spencer and tell him that going to 2 digits has messed up your betting and you need him to go back to 3 digits. 😛
Skeptical
I could but either:
1) This was a fluke or
2) People have been giving him grief over the 3 digits.
If (1) it will correct itself. If (2) those giving him grief have probably been annoying him more than I ever will so he’s not going to go back to 3 digits.
I think what I’ll do is something like open bets the last week of the month and then close them the 4th day. That gives people 10 days to get their bets in and will nearly always close after UAH is posted. To remember to do this, I need to “pre-blog” the betting (which is easy.)
re: “I think what I’ll do is something like open bets the last week of the month …”
What?
John,
I mean that I’ll open bets for Sept UAH on August 25th or so. Then I can close them Sept 3.
Shortening the cutoff date to the 3rd or 4th of the month? That’s madness.
I’ve always found it easier when you can bet on a horse half way through the race (July UAH closed on the 16th), I might have to take my quatloos elsewhere.
redc,
Why not wait until the race is finished, then everyone can bet on the winner?
Is the point to guess what the temperature will be, or to replicate the UAH calculations?
Of course, that gives and edge to the people who are watching UAH. 🙂
But with only 2 digits I think I have to stop doing that.
I’d rather submit my prediction for low summertime Northern Arctic Ice extent and mass……. Any Bets it will be lower than 2007?
cyclone–
We already did bet on extent and mass. I’m pretty sure some bets are it will be lower than 2007. I’ll runs something on that tomorrow.
MASIE and JAXA NH ice extent have been diverging sharply the last few days with MASIE having a much lower rate of extent loss. I certainly wouldn’t bet real money on the outcome.
Okay I get it now. I like the idea of betting on or about the 3rd, I typically bet early anyway. Too much analysis seems to foul up my process.
re redc: “… I might have to take my quatloos elsewhere.”
No worries with me Lucia, my quatloos are staying right here where I earned ’em! I am holding out for “The Blackboard” gift catalog to see what I can buy.
I think using Roy’s chart as a dart board while blind folded would give me the better result. 🙂
MASIE uses multi sensor reports aided by humans. It’s delayed substantially from others s I recalll
cyclone–
We already did bet on extent and mass. I’m pretty sure some bets are it will be lower than 2007. I’ll runs something on that tomorrow.
Ya’ll betting Quatloos or Beer?
I prefer halfway through the month for closing the bets. Yes, I know, that’s like half cheating, but what can you do? Make betting less popular?
Much that I like the Blackboard, I am more likely to visit here for betting purposes than for what Zeke has to say about anything.
From blog PR/popularity perspective, the later the bets close the better it is. When more people get to bet, more people come back to see if they have won anything.
Closing the bets early will only make the whole exercise less entertaining. The purpose of betting on monthly UAH figures is entertainment. The betting is not about winning quatloos though it’s always nice to have a few quatloos for rainy days for what it’s worth.
sHx
I don’t really understand why closing the bets earlier will make betting less popular or less entertaining.
What is your reasoning behind that belief?
Ray,
I wonder how much has to do with people understanding the range of predictability from Aqua and whatever.
Though it would probably make Lucia’s life too complicated, I could see a scenario where the bets were left open all the way until the last day of the month, but continuously downweighted.
BillC–
I’d need to figure out the algorithm for downweighting. Would the winning value be the minimum of (difference/time) with time in days of the months left? Or what?
Technically, it’s not difficult. But should it be time^-1? time^-1/2? time^-2 etc. I don’t think there is a “right” answer.
Ray (Comment #101136),
It will probably make no difference for people who take guidance from tea leaves or tarot cards, but it will make all the difference for those who take guidance from AQUA. The entertainment is in watching AQUA for the first week or more of the month and then guessing which way it will go for the rest of the month. By taking away that guidance, people will be betting completely blind… where’s the fun in that?
Skeptical–
That’s what I think. Some people enjoy watching AQUA thinking it will help them refine their bet. It does help somewhat.
If I weren’t too busy with other non-betting non-climate stuff, I might pull out a list of “dif” vs “time before end of month” data (where “dif” is the error in the guess), plot that data and see which value of ‘a’ gives the best fit for
dif= time^a.
Presumably, “a” will be negative– people who bet latter do better because they have more information from AQUA. But I’m not certain of that and certainly the rule will be violated on the first day of betting because no one can bet earlier.
But… dang… now I’m tempted to do it. (Better not. I need to do other things!)
Lucia,
The problem I can see with downweighting is that it will no longer be a case of whoever is closest wins. It will become a lot more complex and off-putting to people who don’t like or don’t understand the weighting system. It might be worth waiting until bets close this month, have a look at the spread and then decide if anything needs changing.
While betting on 2 digits means we only have a tenth of the possibilities of a 3 digit bet, people are (generally) still only going to bet within a band of numbers… I can’t see that changing the entire betting system will do anything to change the size of the band that the bets will fall into.
Lucia,
Your blog, your call ;).
NEW MYSTERY HOCKEYSTICK GRAPH
Since this is a betting forum, anyone who is interested can check out my new mystery hockeystick. I promise you this – it’s real data with no special manipulation. The only hint the graph gives you is to tell you which curve is in which axis and thus the scale of the units. I am surprised I haven’t seen this in the blogosphere yet, but I am guessing maybe it surfaced a long time ago. Mosher might know.
Skeptical
I agree. At best, I would pick the “optimum” rate of down-weighting based on time. But it gets complicated, and…. that starts to make things less fun.
Sure. but the purpose is for people to have fun. I think it might be less fun if everyone is trying to gauge how much difference the time effect makes.
Re: Steven Mosher (Aug 7 20:42),
Not as far as I can tell. I get very good correlation with JAXA daily and NOAA monthly for over a full year at lag 0. There have been enough wiggles that a significant lag would be obvious. The big problem is that they only post the last 30 days. So for longer term, you have to create your own archive. JAXA recovered faster from the minimum last year than MASIE, but the current difference is much larger than that and in the other direction.
David E.
As someone who has “hole in one’d” a couple of times, I agree. The most consistent/accurate methods are more valuable.
I’m playing a betting against-the-grain strategy, which is not really a useful scientific endeavor, but is good game-theory!
cheers,
Robert
Skeptical,
Personally I would have thought that the more of the month you are guessing, the more entertaining/fun it will be.
dewitt, there was a nice fight about Massie on nevens blog last year.
cant recall all the details..
Here is my analysis of the betting this month:
NO. OF BETS 52
MAX 0.390
MIN -0.013
MEAN 0.246
MEDIAN 0.250
STD DEV 0.076
MEAN 1-26 0.243
MEAN 27-52 0.249
MEAN PLUS 1 SD 0.322
MEAN MINUS 1 SD 0.171
WITHIN +/- 1 SD (%) 76.92
ABOVE MEAN 27
BELOW MEAN 25
MEAN V JULY -0.034
BETS ABOVE JULY 16
BETS BELOW JULY 34
There seems to be a strong consensus that the figure will be between 0.17c and 0.32c, with the betting fairly evenly balanced between those who are above the mean and those below.
As a result, the mean and median figures are very close.
The mean figure this month is slightly lower than the July figure but the number of bets lower than last month, outnumbers those above by almost 2:1.
This month’s S.D. figure is almost the same as last month’s but the number of bets between +/- 1 SD has gone from 68% to 77%.