Betting, Data Comparisons

August UAH: Temperature up.

35 Comments

Roy has spoken! The August UAH temperature anomaly was 0.34C. That’s up slightly from July. For trend watchers, the trends since January 1980 remains below “0.2C/decade”. Temperature anomalies along with best fit trends and uncertainty intervals computed several ways are shown below. (Note: The Arima is the one that gives the largest uncertainty intervals of any choice of AR(p,0,q) with p<=4 and q<=4. The best fit arima would give smaller intervals. )

Now, it’s time to see who won the quatloos!
This month, win place and show went to AMac, PaulSk and RobB. Amac was only off by 0.001C. Good going!

Here are the rest of the bets!

Winnings in Quatloos for UAH TTL August, 2012 Predictions.
Rank Name Prediction (C) Bet Won
Gross Net
— Observed +0.34 (C)
1 AMac 0.341 3 51.899 48.899
2 PaulSk 0.35 3 41.52 38.52
3 RobB 0.32 5 55.359 50.359
4 JohnF.Pittman 0.32 5 44.287 39.287
5 Skeptikal 0.32 4 28.344 24.344
6 DavidE. 0.36 1 5.669 4.669
7 Pieter 0.314 5 2.702 -2.298
8 Rotondo 0.31 5 0 -5
9 Ray 0.31 5 0 -5
10 sHx 0.3 5 0 -5
11 ArfurBryant 0.293 5 0 -5
12 Owen 0.39 4 0 -4
13 Boris 0.39 5 0 -5
14 ivp0 0.29 5 0 -5
15 EarleWilliams 0.29 5 0 -5
16 BobW 0.287 3 0 -3
17 PaulS 0.28 5 0 -5
18 lucia 0.28 5 0 -5
19 MichaelP 0.275 4 0 -4
20 Ben 0.267 5 0 -5
21 TimTheToolMan 0.26 5 0 -5
22 pdjakow 0.26 5 0 -5
23 EdS 0.25 5 0 -5
24 SimonS 0.25 5 0 -5
25 TerryMN 0.25 5 0 -5
26 lance 0.25 5 0 -5
27 PaulButler 0.25 4 0 -4
28 Pdm 0.241 5 0 -5
29 AnthonyV 0.24 5 0 -5
30 LouSkannen 0.24 3 0 -3
31 Rick 0.24 5 0 -5
32 ScottBasinger 0.24 5 0 -5
33 MikeP 0.24 5 0 -5
34 LeoG 0.234 3 0 -3
35 DonB 0.23 4 0 -4
36 BobKoss 0.23 5 0 -5
37 denny 0.225 3 0 -3
38 Jefff 0.22 4 0 -4
39 redc 0.22 5 0 -5
40 SteveF 0.217 5 0 -5
41 AndrewKennett 0.21 4.78 0 -4.78
42 SteveT 0.2 1 0 -1
43 Tamara 0.2 5 0 -5
44 EdForbes 0.2 5 0 -5
45 Freezedried 0.185 3 0 -3
46 Genghis 0.17 5 0 -5
47 DiegoCruz 0.14 5 0 -5
48 March 0.12 5 0 -5
49 Hal 0.111 5 0 -5
50 JohnNorris 0.11 5 0 -5
51 KreKristiansen 0.1 4 0 -4
52 RobertLeyland -0.013 4 0 -4

The net winnings for each member of the ensemble will be added to their accounts.

35 thoughts on “August UAH: Temperature up.”

  4. “The August UAH temperature anomaly was 0.34C. That’s up slightly from August.”
    ______

    And that’s a neat trick.

  8. Odd, I don’t remember putting a – sign in, but I was way off even if flipped. Better spin the wheel again!

  9. > wonders if AMac has done a piece of upside-down Mann trickery

    Heh, heh. All your satellite sensor are belong to us.

    I’ll have to speak sternly to my RND() function about that third decimal place!

  13. RE:
    SteveF (Comment #102712)
    September 6th, 2012 at 6:28 pm

    “This is a really hard game.”
    #######################################
    It’s not so hard Steve. Much like wearing a blindfold and throwing darts at the wall or playing Vegas slot machines with a better chance of winning.

    Global weather is fickle. 🙂

    Bravo AMac, Nice shooting!

  14. I don’t understand the betting scheme.
    http://rankexploits.com/musings/2009/betting-on-things-quatloo-distribution-schemes/
    ————————————————————————–
    3.The over all winner will be the person whose bet came closest to Roy’s posted value. That person will be awarded the smaller of [(2* their wager) or (the total number of Quatloos in the pool)]. The amount they are awarded will be subtracted from the pool. (The net effect is their account gains the number of Quatloos they bet.)
    4.The person with in second place will be awarded the smaller of [( 2*0.751*their wager) or (the total number of Quatloos remaining in the pool)]. The amount they are awarded will be subtracted from the pool.
    5.The person in the “nth” place will be awarded the smaller of [( 2*0.75(n-1)*their wager) or (the total number of Quatloos in the pool)]. The amount they are awarded will be subtracted from the pool.
    —————————————————————-

    Is this the scheme you are using? It makes no sense to me and doesn’t fit the results.

    Are 5 Quatloos given away to everyone every month or just when you start? I assume our Quatloo assets carry over from previous months, but does that go on forever or do you start over periodically with everyone having a 5 Quatloo balance? What is the difference between Gross and Net Quatloos?

    This was my first bet; I tied at 3rd place, but was ranked at 6 (presumably because I bet last). I only bet 1 of my 5 alloted Quatloos, so according to the formula I should have been debited the one Quatloo I bet and been award 2 * 0.75^5 = 0.475 Quatloos: so I should have net 4.475 Quatloos, but the results show 4.669 net Quatloos. I don’t get it.

    Also, the other results don’t make sense given the scheme I linked to above. The winner should not have more than 5 Qs unless they are carry overs from previous months, but in that case there should be lots of people with negative Quatloo balances that are much greater than -5 Qs. And if you are allowed to carry over and can’t win more than what you bet, the smart bet would be to not bet at all since the most you can win is 10 Qs – a 5 Q debit = 5 Qs.

    What am I missing here?

  15. David E– Hmmm… I must have tweeked it from that… A person who bet 5 quatloos doesn’t win 10 quatloos… I remember way back there is a power involved. I tweaked the “n” to avoid having 1 person win all the quatloos or having everyone get some gross winnings. I’ll have to look at the script to remember the formula. It is a “secret sauce”.

    1) Everyone starts with a 0 Quatloo balance. You will then either be in debt or have a positive balance.
    2) When you bet, you have to give me some quatloos. I trust you, so this is a paper transaction. The pot holds all the Quatloos people handed over.

    3) When you Win, I fish out “Gross” from the pot.

    4) Gross – Bet= Net Quatloos. So, that’s how much your balance was incremented. You could figure this out yourself, but some people like to notice that even though I might have handed you 2 Quatloos, you bet 3. So…. your total winnings declined.

    I’m planning to tally up everyone’s sums next week. I’ll find the “n” for you then.

  16. Here’s my reverse-engineered “secret sauce” for determining how the pot is split.

    1. Start with the pot equal to the sum of all bets.
    2. 1st place wins (grosses) alpha*(his bet), where
    alpha = 3.915 times the average bet. Remove this amount from the pot.
    3. 2nd place wins alpha*(his bet)*0.8, limited to what’s remaining in the pot. Remove this amount from the pot.
    4. Repeat step #3 for the nth place, with factor = 0.8^(n-1), until the pot is depleted.
    .
    So it’s just as Lucia described it as quoted in David E’s post, except that the factor 0.75 has changed to 0.8, and the multiplier 2 is now the alpha given by the formula above. Alpha has been 16.3 to 17.3 for the last few months, as the average bet has been 4.2 to 4.4.

  17. Does Lucia issue 5 Quatloos to us each month or do we produce them from the boundless resources of our own imaginations?

  18. Lucia,
    “I’m planning to tally up everyone’s sums next week. I’ll find the “n” for you then.”
    Are the accounts for the UAH bets kept separate from those for Ice Extent?
    Either way, it sounds like a horrendous accounting exercise.
    I assumed the accounts were fictional.

  19. OK I’m a newbie here, but I must say this temperature betting scheme is way too dependent on luck and far too little on skill. It is because the winner gets an outrageous number of quatloos compared to the runners up. In June Tamara missed the temperature by 0.008C and got ZERO Quatloos while the winner missed by 0.005 and got 41 Quatloos! Really?

    Lucia, how about tweaking the scheme too spread around the Quatloos – put more skill in the game?

  20. Lucia,

    Here is my betting scheme suggestion: humbly submitted, for your consideration.

    —
    (I don’t know how this editor will handle
    my table with its tabs etc. Hopefully its not scrambled)
    —

    ————————————————-
    Payout vs. Prediction Error:
    Nominal
    Less Than or Greater Than: Payout.
    Equal to: Bet multiple:
    0.07 0.05 1.5
    0.05 0.04 2.0
    0.04 0.03 2.5
    0.03 0.02 3.0
    0.02 0.01 3.5
    0.01 0.00 4.0
    if =0.000 4.5
    ———————————————-

    –
    Then the nominal payouts are normalized to the pot size as follows:
    –
    (Total Pot Size) / (Total Nominal Payout) * (Nominal Payout) = Actual Payout

    –
    So for example, suppose 50 people participate and contribute a total of 200 quatloos to the pot and the pots nominal total payout turns out to be 220 quatloos going to 20 people.

    –
    You are one of the 20 and have bet 5 quatloos that the anomaly would be 0.35 C; the actual turned out to be 0.40 C so your error is 0.050 C. You would have nominally won 2.0 * 5 = 10 Qs. However the payout has to be normalized to fit the available quatloos in the pot so your actual winnings would be:
    200/220 * 10 = 9.1 Quatloos. Likewise if the pot had been larger than the nominal total payout you would have won more than 10 Quatloos.

    –
    Looking over the last 5 months on average about a third
    of participants would win some Quatloos in a month. On average the total Nominal Payouts would be very roughly equal to the total actual payouts, but this would vary quite a bit month to month.

    –
    The advantages over the current system:
    1. Simpler, more transparent and predictable payout.
    2. Close guesses always get rewarded and with more consistent payouts.
    3. It better rewards skill in the long run.
    4. Difficult months to predict (where fewer people are in the winnings) the actual payout will be significantly greater than the nominal after normalizing; conversely easy months to guess you will have a smaller payout – but will always get paid for close guesses.

  23. DavidE– If I did that, I’d smooth it and code it. But I’m not going to set aside time to code it. (I’ve promised people I’d sum up and I haven’t even done that yet.)

  24. Lucia,

    I guess smoothing would make it easier to code. If I offered to handle the accounting would that be helpful? I can set it up in Excel, could grab the bets after you publish them and send the results to you after the UAH comes out. I’d understand if you feel that was too intrusive or just too much of a hassle. If your interested let me know, if not I’ll drop it.

    I noticed a weakness in my scheme: the lack of a provision for no winners or just a few winners. To avoid the accounting hassle of carrying pots forward and inordinantly high payouts, these occasions could be handled with a rule that the entire pot must be payed out but the maximum payout must not exceeed something like 3* the nominal payout. The range of winning prediction errors would be expanded until there are sufficient winners to produce a maximum payout of 3* the nominal payout.

  25. DavidE–

    I’d understand if you feel that was too intrusive or just too much of a hassle. If your interested let me know, if not I’ll drop it.

    This doesn’t reduce the work. I would then have to
    a) accept your excel spreadsheet each month.
    b) write a script to read your results into my database.
    c) record the wins and losses for each person.

    To avoid the accounting hassle of carrying pots forward and inordinantly high payouts, these occasions could be handled with a rule that the entire pot must be payed out but the maximum payout must not exceeed something like 3* the nominal payout. The range of winning prediction errors would be expanded until there are sufficient winners to produce a maximum payout of 3* the nominal payout.

    Yep. If/then/else’s in the code. 🙂

    I’m not convinced your ways is better than mine. But that’s possibly because I’m not really spending time thinking about the relative advantages to different pay out methods. In your method, the person who gets closest doesn’t necessarily win any more than a person who is almost as close. On the other hand, people who are very close but still come in fifth might not get shut out– as far as winnings they might get as much as the first place winner.

    My bets are more like horseracing. If you horse comes in first, by a nose, you win. If another horse beats you by a nose, the other horse beats you. I don’t know that either method of deciding how much to pay out is “better” than another. I like your method if this was grades on tests in course work where we want people to get A’s, B’s and C’s based on mastery. To a large extent we would like to eliminate luck in that instance. If, by some fluke we got all A students in a class, hypothetically, they should all be given A’s. Likewise for E’s. Same standard holds for ribbons given to 4-Hers in county fairs.

    But I’m not sure that system is better for horseracing. (Actually, both academics and 4H use a mixed system. Blue, Red and Green ribbons are A,B,C in 4H. But the purple ribbon is sort of like the valedictorian. It’s for the best one.)

  26. “My bets are more like horseracing. If you horse comes in first, by a nose, you win. If another horse beats you by a nose, the other horse beats you. ”

    The difference is you win money in a horse race; quatloos have no value if they merely represent blind luck.

    .

  27. DavidE.
    I don’t see how payouts being in $$ v. Quatloos affects whether winning is due to blind luck or skill. I understand you don’t like the payout method. That said, there hasn’t been a general outcry against the method used. Coding effort would be required to change the entire system. I just don’t see the need. Anyone who wants to set up a betting system with a different pay out method is free to run their own system on any betting site they host. If people find some rival sites betting system more attractive, I suppose they’ll go there to bet.

  28. “I don’t see how payouts being in $$ v. Quatloos affects whether winning is due to blind luck or skill.”
    ——————————————————-

    It doesn’t. Money has it’s obvious value, where counting Quatloos here can only have worth by bringing validation or doubt to ones personal choice of climate models, but then only if skill is a highly significant factor in how one accumulates Quatloos. Isn’t climate modeling a central theme to your blog? All I am saying is that there is an unnecessary amount of additional luck involved in Quatloo accumulation: by virtue of the betting scheme. Since UAH LT temp is inherently noisy, anything to reduce noise or luck in the betting scheme adds psychological worth and cache’ to the Quatloos while still leaving plenty of surprise people enjoy in games. Which just makes it more fun IMO.
    –
    If you prefer the ranking scheme and/or avoiding any recoding; – I don’t know your algorithm but – I should think just tweaking some variables in the formula ought to spread Quatloos to more people with near misses and de-emphasize the payout differentials – which would have the cumulative effect of increasing the ratio of skill to luck over the course of a year.
    –
    Apparently, I am one of a few or perhaps the only one who cares how it’s done so I can understand why you don’t want to bother with it. It’s no big deal. The only reason I keep arguing the point is not because I am still trying to convince you to recode the scheme, but rather I don’t think you understood my central objection. And I like to argue. LOL.

  29. DavidE.

    I don’t think you understood my central objection. And I like to argue. LOL.

    Your central objection seems to be that you think your payout scheme somehow reduced the amount of luck in some way or another. Am I mistaken?

    All I am saying is that there is an unnecessary amount of additional luck involved in Quatloo accumulation… anything to reduce noise or luck in the betting scheme adds psychological worth and cache’ to the Quatloos

    Everyone agrees there is lots of luck involved in the monthly outcomes for UAH. I don’t think your payout scheme affects the amount of luck one iota. It stays the same.

    Any system where we predict short term weather will be strongly influenced by luck. But some people do win more often than others. Quite often, the reason is that they pay attention watch channel 5 until just shortly before the bets close and make bets based on the most recent possible information. These people do win more often.
    Others just look at last months weather and make a wild guess.

    I don’t think the payout scheme affects this much.

    If you think spreading the winnings out more deeply will reduce the standard deviation of the amount won/lost by individuals over time– of course. But this is gambling. Reducing the standard deviation of winning is not what makes gambling fun. The whole point is that the winners win a bunch. If we wanted to reduce the standard deviation, we’d just give everyone back exactly what they bet at the end of each month. There’s no fun in that.

  30. Lucia,

    ————————————————————–
    “Your central objection seems to be that you think your payout scheme somehow reduced the amount of luck in some way or another. Am I mistaken? ”
    ————————————————————————
    You are not mistaken.

    ———————————————————————
    “I don’t think your payout scheme affects the amount of luck one iota. It stays the same. ”

    ———————————————————————
    –
    That certainly is not the case.
    –
    Case in point: the June results.
    http://rankexploits.com/musings/page/6/
    Tamara’s estimate (ranked 4th out of 36 entries) had an error of 0.008 C; Rotondo’s estimate (ranked 1st) had an error of 0.004 C. Both are less than 5% of the standard deviation in monthly UAH residuals about a naïve regression of temperatures over the last 20 plus years and probably less than 10% of the standard error in the best professional climate models, i.e. Bullseyes. However, Tamara won zero Quatloos while Rotondo got 86!. And Rotondo got 29 times more Quatloos than 3rd ranked Anteros.
    –
    For Tamara to make up for that June loss to Rotondo she would have to win 29 similar 3rd place rankings and Rotondo would have to lose 29. Clearly Rotondo’s estimate demonstrates no greater skill than any of the others in the top 4, yet Rotondo gets a disproportionate share of the Quatloos. That’s pure luck.

    In my scheme, nominally: all 4 would have grossed 4 times their bets. The awards would drop off after that down to 15th ranked Pavel (with a 0.063C error) with 1.5 times his bet.

  31. Lucia,

    ————————————————————–
    “Your central objection seems to be that you think your payout scheme somehow reduced the amount of luck in some way or another. Am I mistaken? ”
    ————————————————————————
    –
    Too be more clear, my objection (as detailed above) is to the excessive amount of luck in the current scheme. I am not particularly enamored with my scheme; most any simple and transparent payout scheme that improves on the skill to luck ratio would be more fun IMO. Adult games require significant skill to keep peoples interest – unless their is a cash reward. And success – or the lack thereof – in the Quatloo game can be fodder for discussion about climate models or GW in general if it’s not just a guessing game – again more fun.

  32. DavidE.
    The betting has nothing to do with climate models. It doesn’t even have much to do with the size of residuals about the linear fit to UAH.

    Anyone who uses the linear fit to UAH to make their guesses about next month’s temperatures lacks understanding and has no skill and doesn’t understand anything about what factors you need to forecast imminent weather. Not everything is about climate models.

    Adult games require significant skill to keep peoples interest – unless their is a cash reward.

    I’ve been running these bets for … how long? Three years?

  33. DavidE. (Comment #103616),

    The only reason I keep arguing the point is not because I am still trying to convince you to recode the scheme, but rather I don’t think you understood my central objection. And I like to argue. LOL.

    David, stop being such a PITA. This is Lucia’s game… if you don’t like the way the game is being run, then don’t play.

  34. Lucia,

    I don’t know how I could offer a suggestion more contritely or helpfully. I wasn’t trying to make a challenge to your domain or get into a pissing contest. Judging by your many posts you are obviously intelligent, educated and fully capable of rational discourse and recognizing the shortcoming of the scheme which I attempted to point out, but now you are just being defensive and trying to score points with tangential proclaimations of the obvious and a disingenuous representation of my argument.
    Tedious. Pointless.
    –

    Disappointed and Gone.
    –
    –

Comments are closed.