Holy cow, and we're already almost there! $34 may not be high enough!

Earlier in this thread, I jokingly said that if you want to speculate on prices higher than $99 by July 1st, just buy bitcoins. Now I have to decide if I can provide a mechanism to do so. Any ideas? Should I just expand the range to include values up to 1.99 BTC? 2.99 BTC? (Obviously that would only apply to bets posted after the change is announced).

The 5% thing was a "maybe for next time" idea. The rules are set in post #1 and will not change, except as needed for clarification.

Regarding inclusive bets, that is a good question which is not addressed in post #1. I doubt that situation will arise, but if it does, I think I HAVE to make them inclusive. It's much better to have an extra winner than to have someone be exactly $0.50 away and be excluded by a technicality.

Post #1 has been updated accordingly. It now says:

Thanks! There have been some websites which have sprung up to fill this need since I started this thread, but there are still 6 unclaimed bitcoins in the pot, so I updated the thread title to hopefully get some attention that this contract will be ending soon.

Ah, I was looking for this thread last night. Maybe my prediction will turn out to be true after all.

You are correct that a late entry can lose money even if they guess correctly, but that is only true if they are late enough and there aren't many losing bets. That is intentional.

I'm pretty sure it is not possible for a late bet to cause an earlier winning bet to lose money. Here is my example:

A bets 2 BTC 10 days in advance, loses
B bets 2 BTC 10 days in advance, wins
C bets 10000 BTC 1 day in advance, wins

The pot is 10004 BTC.
I would get my 2% fee = ~200 BTC, winners get 98% = ~9804 BTC

A gets nothing (loser!)
B's weight is 10 days * 2 BTC = 20
C's weight is 1 day * 10000 BTC = 10000
Total of weights = 10020

B's share is 20/10020 = 0.2% (19.57 BTC, a 10x gain)
C's share is 10000/10020 = 99.8% (9784 BTC, oops!)

Of course, in reality there will be a lot of losers, and only a couple winners, so I expect all winners will do well as long as they didn't bet an absurdly high number of BTCs. Doing that only helps me

It's possible that there IS some way to game the system, but I haven't thought of one. This is an experiment. If anybody can see a way to cheat, definitely post your idea (or just go ahead and cheat, and I'll try to figure out what to do about it).

I stand corrected.

However, your system might penalize winners by discouraging new losers (although also discouraging competition, which you'll probably correctly argue is a bigger fear of a early better) because there is little incentive to bet on a position if it is already taken (whether eventually winning or loosing).

Depending on a bet, there are likely more loosing positions than winning positions, so everyone should want an inflated market as long as at no moment is there a position of greater than 50% assurance. If you close the bet well in advance of the moment-of-reckoning, I think your winners would have appreciated more volume. Obviously if you allow betting until the moment-of-reckoning (with no winners loosing) then all bets will accelerate toward the winning position, thus deflating the early winning bids.

I write this forty hours before the close. I've got coins burning in my wallet, but can not rationally choose a position that comes close to the risk (though, I admit the market has shown abnormally low volatility the last three days). On second thought...

An optimistic moron if you ask me.

Looking at the blockexplorer log, it appears that the timestamps are UTC. Pacific Time is currently -7 hours from that (we are in pacific daylight time right now). I noticed that at one point in my original post I said the time would be "PST" which is not technically correct since right now my clock is running in "PDT". I have updated post #1 to just say "pacific time" now, because I did not mean to imply that I would be ignoring the daylight saving adjustment.

On blockexplorer timestamps, this contest ends: 2011-07-01 19:00:00 (12+7=19)

The 1.165 BTC bet came in at 2011-07-01 11:26:44 UTC which was 7 hours, 33 minutes, 16 seconds early, which is 0.31476851851851... days early. If that bet is a winner, its weight will be 1.165*0.31476851851851... = 0.36670532407407407...

The 0.16 bet came in at 2011-06-28 09:03:50 UTC which was 3 days, 9 hours, 56 minutes, 10 seconds early, which is 3.414004629629629... days early. If that bet is a winner, its weight will be 0.16*3.414004629629629... = 0.5462407407407...

I calculate that any more winning bets placed at this time would definitely be unprofitable (unless the price falls below $15.5). At this point, the pot is 9.743 BTC. My 2% is 0.19486 BTC (which does not cover the 0.5 BTC I gave away to inspire early betting), leaving 9.54814 BTC to be split among the winners according to their weights.

Edit: Fixed several mistakes in the calculations above. I will look for at least one person to double-check my math before this is all over.

I realized I can actually download the trades for the time period above from the list of recent trades here: http://mtgox.com/code/data/getTrades.php

Here's the most interesting/relevant chunk of data:

{"date":1309546534,"price":"16.06869","amount":"0.1","price_int":"1606869","amount_int":"10000000","tid":"1309546534570843"}{"date":1309546714,"price":"16.06869","amount":"0.70603902","price_int":"1606869","amount_int":"70603902","tid":"1309546714395447"}{"date":1309546714,"price":"16.065","amount":"19.79800996","price_int":"1606500","amount_int":"1979800996","tid":"1309546714419664"}{"date":1309546714,"price":"16.05011","amount":"0.15","price_int":"1605011","amount_int":"15000000","tid":"1309546714475023"}{"date":1309546714,"price":"16.05","amount":"1.18","price_int":"1605000","amount_int":"118000000","tid":"1309546714499844"}{"date":1309546714,"price":"16.0234","amount":"1","price_int":"1602340","amount_int":"100000000","tid":"1309546714528446"},
{"date":1309546714,"price":"16.02001","amount":"1","price_int":"1602001","amount_int":"100000000","tid":"1309546714548907"},
{"date":1309546714,"price":"16.011","amount":"2.23962072","price_int":"1601100","amount_int":"223962072","tid":"1309546714566519"},
{"date":1309546714,"price":"16.01","amount":"50","price_int":"1601000","amount_int":"5000000000","tid":"1309546714601679"},
{"date":1309546714,"price":"16.0007","amount":"13.91720362","price_int":"1600070","amount_int":"1391720362","tid":"1309546714633848"},
{"date":1309546714,"price":"16.0007","amount":"10.94612668","price_int":"1600070","amount_int":"1094612668","tid":"1309546714691336"},
{"date":1309546857,"price":"16.05101","amount":"1","price_int":"1605101","amount_int":"100000000","tid":"1309546857305532"},

Using a handy unix timestamp converter (http://www.onlineconversion.com/unix_time.htm), I come up with the following sequence of events:

Fri, 01 Jul 2011 18:55:34 GMT: Somebody buys 0.1 BTC at 16.06869
Fri, 01 Jul 2011 18:58:34 GMT: Somebody sells 100.937 BTC all at once, driving the price down to 16.0007
Fri, 01 Jul 2011 19:00:57 GMT: Somebody buys 1 BTC at 16.05101

It's probably a coincidence, but we can't really know for sure. If someone was trying to manipulate the outcome with the big sell, they failed. It could also be that one or more of the buy orders in place right above $16 which stopped the price from going under were also trying to manipulate the outcome, in which case they succeeded.

As the lucky, late, and optimistic moron at $16.5 I'd like to thank Dacoinminster for an entertaining run and in particular, I thank all the losers who took the time and cost to pay the boss. Since I would have been the only one to have benefited from the price above $16 and Mr. 16eg1Q to have benefited from a price below, I can only conclude that we have a market manipulator among us! I can assure you, I have done everything I could since I bought coins months ago, to keep the price as high as possible. I am therefore happy the market agrees with (albeit much lower than) my greatest expectations.

Thank you, thank you, thank you. With my winnings I shall take my girlfriend and two other ladies out for a luxurious coffee (Mr. 16eg1Q if you'd like to join us up where the sun don't set, we'd love to share a cup under the midnight sun). I owe my warm caffeinated happiness to all of you. Thank you.

Dacoinminster, I'd like to propose, should you administer a similar bet again, that you set a closing time on wagers followed by perhaps an hour during which a volume weighted average is calculated, rather than a discreet closing price. I expect a VWAP would be impractical or too costly to manipulate. I think (as long as you are dividing by the total pool rather than the loosing bids) there will always be a variable period after which betting will be unprofitable. So, setting a close time well before the VWAP period would have no negative consequence on the game.

I continue to recommend a compromise to the parimutuel. Perhaps the winnings pool is composed of 100% of the loosing bids and some percentage (50% ?) of the winning bids. And/or weighting by the square (or ^1.5) of the time. I expect these changes would secure early bidders with a great advantage (encouraging early bids) while minimally discouraging later bids, thus increasing the volume throughout the betting period. And more profits for graceless jerks like me and Mr. Market Manipulator 16eg1Q.

EDIT: Because there will be a period of lost winning bets, a closing time before determination and payout will not be necessary. I think it is safe to start the VWAP period before the final close of the market. For example, all bets must be in by 24:00 UTC, where the VWAP is calculated 23:00-24:00 UTC. Betters will likely loose if they wager before 23:00 and I suspect will loose with wagers much earlier still.