August 19th, 2008
|03:55 pm - Long-shot Prediction Game: initial commentary|
Doesn't the world feel just the tiniest bit less... screwed than it did a month or two ago?
This is in response to the long-shot prediction game. Iain (the blogger formerly known as daweaver) has written a fascinating assessment of the twelve propositions and his reasons for accepting or rejecting them. As has happened before, he has saved me writing much of this article. (It's tremendously convenient when this happens!)
When I decided what the twelve propositions for the long-shot prediction game should be, my original intention was to provide a variety of propositions that I considered to be something like 20%-25% likely. Later on, I realised that it didn't really matter how probable I considered them to be and that the game would not necessarily work more or less well if there were some propositions which were more likely than others - and if it led to making some propositions simpler to adjudicate, so much the better.
Let's consider the twelve propositions individually.
USA to win most gold medals in Beijing: This one clearly wasn't a 20%-25% likely proposition, being much closer to even money. Indeed, shortly before the Olympics, you could get 5/6 against China and also 5/6 against the USA topping the medal table. (Not sure whether that was based on the "most golds" or the "most medals" metric, and dead-heat rules may apply when this proposition clearly calls for an outright win.) Analysts are split on the issue, with Luciano Barra's methodology predicting a win for the USA and Sports Illustrated's potentially pessimistic prognostications favouring China. Under current form one would have to back China, but I'm not completely sure that the current Chinese lead in the medal table isn't at least as much a result of the order in which the events are competed as it is a result of Chinese performances.
Incidentally, it's interesting to me to see the predictions for the team from Great Britain and Northern Ireland (and, sotto voce, associated Crown Dependencies). Considering the targets set by UK Sport, I suspect that British performances at this Olympics will be considered to have been about 60% awesome and 40% patchy; clearly the cycling team has massively, almost praeternaturally, overperformed compared to expectations, the rowing and sailing teams have done very well and Rebecca Adlington has carried the swimming team single-handedly.
It is difficult to conclude on the performances in other sports - and, as I type, we've just won a silver in the high jump, of all the things, which must be considered a glorious surprise - but archery, judo and shooting must be considered disappointing at least. (Incidentally, what are the comparative angles subtended at competition distances by the inner gold in archery, the 10-point circle in target shooting and the treble 20 in darts?) Just as well that the cyclists outperformed their target by six medals to make up for those three sports underperforming by six.
HP 6 to gross at least $285 million at the US box office: Poo, I was looking forward with great interest to seeing how this one would turn out, but the postponement of the release of the movie has put paid to that. There are compelling arguments both in favour and against this proposition.
* It's been a bad year at the US box office due to the credit crunch and the relative decrease of importance of cinema, with receipts below those of last year in real terms - and, had it not been for The Dark Knight, down in gross terms as well.
* Surely Harry Potter and the Order of the Phoenix gained from the fact that the
final seventh novel-length book in the series was released a week afterwards, and Pottermania presumably must decline now that the seventh novel has been released.
* There is a well-known major spoiler for the plot of Half-Blood Prince, which can't help the film's prospects.
* As bateleur points out, I was only considering data from the first eleven weeks or so of the film's receipts, not that that should make more than a very few percent difference.
* The US moviegoing public have been starved of a box office sensation for months, with The Dark Knight showing what can be achieved when the right film does come along.
* New screens are still being opened and ticket prices are going up.
* Half-Blood Prince has every prospect of being a much better movie than Order of the Phoenix, just by virtue of containing a story reasonably widely considered more entertaining and far more cinematic than its predecessor.
* If James Bond movies keep setting franchise records, why not Harry Potter ones?
After considering whether or not to void the category, I fear that I will have to be literal rather than liberal and declare that - unless the release date is changed again, and stranger things have happened yet, albeit infrequently - the movie is likely to have generated zero US dollars at the box office before the deadline. (The sporting thing to do would be to allow Twilight box office returns to count instead, but this would be illogical.) If I were to void the category and the release date were to revert to before the deadline, we'd all feel like right bananas.
Random weakly-related question. It is widely believed that the Deathly Hallows movie will be released in two parts. I wonder if the second movie will conclude with the epilogue from the book? If so, will I be the only one walking out of the theatre rather than watching the epilogue, then walking back in for the credits?
BBC Global 30 stock market index to drop below 4750:
Oil to trade at $200/barrel or higher:
USA and/or Israel to execute an overt air strike against Iran:
John McCain to win at least 336 Electoral College seats:
Iain correctly points out that there is a degree of correlation between the results of these propositions - indeed, arguably, the most likely numbers of long-shots to occur are, possibly, two followed by five and only then three or four.
Isn't it wonderful that the price of oil is only - "only" - in the low $110s per barrel! Certainly the $200 figure was one of the hardest to set, and from memory it was set when prices were in the high $130s and rising. The principle behind it was that oil prices had risen by about 50% in the last six months, so I was asking you to speculate whether it might rise by another 50% in the next six months. Given that the oil price has rewound so far, a $200 barrel looks somewhat further off. Nevertheless, it could be argued that speculators managed to double the price of a barrel of oil, so perhaps they might manage to make it double again. (Or halve it from its recent peak, which would suit me down to the ground as a consumer.)
Iain points out that the recent fall in oil prices has correlated almost exactly with the strengthening of the US dollar, as investors favoured oil and gold in response to the weak US dollar and retreated from commodities as the US dollar started to strengthen once again. There are some other fundamentals (i.e. solid reasons for prices being what they are) at work, too: demand in the US actually falling as a result of price increases, supply problems caused by Atlantic hurricanes and African instability being less bad than feared and political pressure to keep supplies flowing.
The selection of the 4750 target level for the BBC Global 30 was based on similar logic - a fall of some number of points over a period of time could conceivably be repeated over the next six months. (An arithmetic application of global recession, in this instance, rather than a geometric one.) We hear discussion of possible impending bank failure in the US which might set the rest of the stock markets tumbling in sympathy.
The possibility of an air strike on Iran came directly from this one on InTrade. I saw it, I liked the sort of probability estimates it's coming up with and the fact that I'm settling up until the end of January 2009 probably puts the event fairly squarely in my desired probability range. (Part of the whole thesis of this enterprise is to see whether there's any sense trusting in the opinions of prediction markets, after all.)
The market about McCain was inspired by the fascinating (and distinctly plausible) FiveThirtyEight.com. They have a model which simulates the US presidential election and they run it 10,000 times daily. I picked the target level of 336 because the Electoral Vote Distribution seemed to suggest the chance of the election turning out with 202 or fewer Obama Electoral Votes looked like about a fifth. I'd like to see the simulated probability distribution function integrated into a cumulative probability distribution, to see if my guess of 202 is about right.
My gut suspicion is that if there is an increased prospect of conflict in Iran then US voters are likely to prefer a continuation of the current administration, even if they don't have a "war president" as such. I also suspect that such conflict would have economic knock-on consequences likely to trigger the stock market decline and oil price increase propositions.
Microsoft to takeover Yahoo: doesn't look so likely any more, but these things have a habit of rumbling on and later bids might come at any point.
Gordon Brown to stop being Prime Minister, Labour leader or both: this is a classic betting market. Paddy Power now offer only evens against Brown going by the end of 2008, so would presumably quote slight odds in favour of Brown going in the timescale we're considering. On the other hand, they're quoting a slightly silly 120% book, so they probably think it's really about a 40%-45% likely proposition. Remember, you never see a bankrupt bookie. (Except John McCrirrick, and possibly Angus Loughran might well have been doing his share of laying on the exchanges as well as backing...)
Firefox usage to reach 33% or higher: this is the one where I don't agree with Iain's assessment; I chose 33% as a level which was about 20%-25% likely, not almost impossible. Remember, I specified three different counters, and the W3 counter global stats for July suggest a combined share for Firefoxes 1.0, 1.5, 2.0 and 3.0 of 29.68%. That represents a gain of 0.66% over the previous month, but this comes after a period of relative stability for the browser - only an increase of 1.29% over the six months, suggesting a figure possibly around 31% by the end of January 2009. There's always the possibility of retreat, too; Firefox 3.0 isn't going to be released every month, after all.
Named British Prince or Princess to marry publicly: yeah, I was struggling with this one. Looks pretty unlikely to me. Sorry, huskyteer. On the other hand, if you turn out to be correct here, you have the job of Royal Correspondent for life.
No more than 2 major hurricanes in 2008 Atlantic Hurricane season: Hurricane Bertha briefly crept in to Category 3 to count as one of the two. Of course, intensity of a storm doesn't necessarily correlate terribly closely with the damage it does. Needless to say, we all hope for a minimum of damage and loss, even if some of us are hoping for a couple more Category Threes. Always interesting to see if the predictors break down along national lines - whether US inhabitants know something the rest of us don't, or vice versa - but I don't immediately spot a pattern.
New England Patriots to win Super Bowl XLIII: as long as Cheatin' Bill Belichick is in charge, and there isn't a mass personnel exodus, the Pats are going to be there or thereabouts. Backing one team against 31 is always taking a punt, but this prediction market is offering you odds of 79/21 for your money. (That's about 3.7/1.) Given that the markets are quoting it as a 3/1 to 3.5/1 shot, I reckon this is the juiciest opportunity of the twelve.
If I had been playing along, I think I would have gone for the two longest shots - I choose to earn 79 points if "New England Patriots to win Super Bowl XLIII" happens, instead of earning 21 points if it doesn't and I choose to earn 82 points if "John McCain to win at least 336 Electoral College seats" happens, instead of earning 18 points if it doesn't - and left the other ten alone. Some of the others are very closely priced decisions, though, which is what I was hoping for.
While three stages to the voting procedure are quite enough, if we wanted to add a fourth, we could open a market on which of the 21 players is most likely to win the prediction game. :-) My money would be on canadianpuzzler, who has avoided what has turned out to be the HP6 trap and who (like me) favours the Patriot and McCain long-shots. On the other hand, he has chosen to follow six of the long-shots, which is rather more than most, and I reckon most of them aren't too generously priced. (Conversely, the players who have only selected one or two long-shots went for USA gold medals and/or HP 6, neither of which is looking too sharp.) I don't have a strong feeling about this one at all.
Heck, we could open a virtual stock exchange in which stock exchange players trade funny-money contracts concerning the performance of prediction game players! Then we'd start to speculate over who's going to win the virtual stock exchange game... :-)
I do think there's scope to play lots of interesting games on LiveJournal, even if I can't really work out cunning ways to take advantage of the fundamental social nature of the service. (I'm not yet doing anything more sophisticated than the old play-by-mail publish-the-results-to-all-in-a-'zine days, except with a more convenient user interface.) This may be an attempt to harness the wisdom of crowds, something very weakly related to the Delphi method. There are more games to be played in the future.
For instance, I really enjoyed the points race in the Olympic cycling and have a suspicion that it ought to be possible to do a cracking cycling game on LJ. (I'm thinking one lap per day for 160 days...) It ought to be able to come up with some sort of ruleset which will permit people to submit orders with fewer than 30 seconds' worth of thought, default neatly when people forget to submit orders (which they will, probably more than half of the time) and still offer interesting decisions that simulate the quandaries of when to time your charges in order to win the sprints or attempt to gain a lap on the pack.
Much more simply, anyone for a game of Breaking Away? :-)
Current Mood: delighted to be British
Deathly Hallows: Why would you skip the epilogue? Are you one of those entitlement fandomers who disagree with what an author should be allowed to write? Have you read the Twilight series? (Should make this a poll LOL) Did you flip out over BD or did you see it as just a continuation of the lunacy of the first 3 books? What if I prefer to be a left banana???
Might want to check Poblano's site again...think McCain's in real trouble there. With a very little bit of luck, it will be a miracle if he gets 236 electoral votes, not 336. Obama's actually trailing the rest of the Democratic party (and yes I see the angst over this elsewhere but it really is superficial. If the media actually does what the McCain campaign wants and focuses more on their candidate we could be talking about Reagan blowout numbers). Still, the reason that Obama does not have those same numbers (>336) now is the same reason McCain will not either, even with a Iran strike. We're divided, despite the President's approval rating, and it takes time for people to move from one party to another, either out of ignorance or or stubbornance. So, we're a 51-49 nation. Make it 53-47 and that's a landslide...and that's where we'll be.
For instance: http://www.dailykos.com/storyonly/2008/8/18/184025/398
"My gut suspicion"
I have never understood this concept except that we have been shown to be cowardly since 9/11. I hope you're wrong, but nothing I've seen (re: the Bin Laden tape before the 2004 election) has shown that to be verifiably false, unfortunately. Home of the brave...
Pats....Favre...Jets! Jets! Jets! :-)
"favours the Patriot and McCain long-shots"
Money...mouth...is? I'm looking to add a wing onto the house and those are better odds than I would get in Vegas...
Nevertheless, it could be argued that speculators managed to double the price of a barrel of oil, so perhaps they might manage to make it double again.
This is a popular myth, so I can understand why you've picked it up, but in fact it's wrong.
In fact not only is it wrong, it's backwards! Speculators don't drive up high prices. They buy when the price is low (raising it) and sell when the price is high (lowering it). That's how they make a profit. For speculators to cause a price peak would mean they were making a massive loss!
Hat tip to Tim Harford (the Undercover Economist) for the explanation.
I think you are ascribing far more rational utility functions to speculators than may be the case in practice. Arguably if speculators were incentivised to act more rationally - and, more to the point, more in terms of long-term benefits rather than a term as short as their next bonus round - then the world may well have been in less trouble right now than it is.
I am not sure how much misincentivised priorities really have contributed to the global financial problems; probably more than you may think and less than I may fear.
You'll have noticed that my votes for what is going to happen are rather different from my votes for the ones that were most likely to happen. All right, I admit that I changed my mind about the Olympics, but the rest of it is tactical. I haven't quite followed your tactic because although I think McCain has a decent chance of winning the election, there's no way he's going to get 336 seats. I don't know enough about hurricanes or, these days, American football to judge the likelihood of those events, but the odds did look very tempting in points terms.
I've actually really enjoyed the multi-stage nature of this game, as it allows for complicated tactics like the above. In Parish's similar game, I several times got the most correct predictions, but never won. I hope that I have finally learnt my lesson.