If by “the current leading Bitcoin predictions market” you mean Bets of Bitcoin, that’s a single website, not a distributed application, and is susceptible to being shut down.
I thought the distributed part referred to Bitcoin or its epigones.
I don’t think there’s any ‘distributed application’ which seems fitted for a prediction market, except, well, a website. Really, if Bets of Bitcoin were to change its payout rules and reboot as a Tor hidden service, that’d basically be ideal (at least currently, while combinatorial prediction markets are being worked out).
I don’t think there’s any ‘distributed application’ which seems fitted for a prediction market
I don’t see any intrinsic reason why one couldn’t be written. Splitting the problem into pieces:
Posing questions A distributed database could hold these
Validating questions You’d want some form of reputation system or elected panel of moderators, to verify which newly posed questions had the required properties (eg not a duplicate, clear criteria on how the question is to be resolved, etc). Again, elections and webs of trust can both be handled in a distributed fashion with no centralised authority.
Resolving questions Marking which side of the ‘bet’ wins, or if it is still unresolved. Similar issues to validating, though more at stake. Need perhaps the possibility of submissions of evidence and counter-submissions.
Placing bets Easiest way would be for the person to place their sum of money in escrow, encrypted with the key spread out and requiring X out of Y holders of the key parts to release it upon resolution. Alternatively allow people to purchase insurance from ‘market makers’ who, in turn, have large sums in escrow backing the bets. No central authority or website required, though you could certainly use websites as convenience aggregator gateways to the underlying distributed data.
Metadata Keeping track of reputations of who is winning how much, which bets are busy or have a lot staked upon them, etc.
Can you think of an aspect of having a prediction market that REQUIRES there to be a single central website?
Can you think of an aspect of having a prediction market that REQUIRES there to be a single central website?
Merely a Chesterton’s fence: the observation that fully distributed websites are rarer than hen’s teeth, there have been recent failures (anyone using Diaspora?), and there’s almost no infrastructure anywhere which relies purely on distributed webs of trust and escrow despite PGP and GPG being around for decades. (Debian sorta works on a web of trust for developers; it’s also far from the largest Linux distribution these days.)
Why didn’t Silk Road spring up instantly after Bitcoin was released? Why is it a single website? Why does it use PGP keys for encrypting messages between buyers and sellers only? Why do all its competitors replicate this centralized single site model?
For a distributed service of this sort to spring up ‘naturally’ would require not just a threat model that requires it, but a critical mass of people who care sufficiently about the threat.
For a distributed service of this sort to spring up ‘naturally’ would require not just a threat model that requires it, but a critical mass of people who care sufficiently about the threat.
And apparently they don’t exist. That’s a pretty good reason.
I once wrote an open source cryptographic system that would let casinos prove to their customers that the casinos weren’t fixing the random generation behind the wheel spins, dice and card deals.
There was no interest in it. Not from the casinos OR from the customers who were betting their money.
There’s probably an interesting bias behind whether people properly take into account threats that are low frequency but high penalty.
I think there’s probably not much overlap between the sort of people who would understand the meaning of an open source cryptographic system enough to trust it, and people who frequent casinos, except probably for card counting.
If anything they seem to overemphasize low frequency, high-penalty threats. Or at least low-frequency, high-drama ones. Consider gun murder, terrorism, and child kidnapping; versus car accidents or family sexual abuse.
Maybe it is just a black-and-white thinking from most people. Either you trust online casinos, or you don’t. If you don’t, you don’t use them. If you do, you don’t use special cryptography to prove that they don’t cheat.
And the few people who understand cryptography and want to use online casinos and have a healthy degree of suspicion… are not enough to create a profitable market.
You could create a software which is easy enough to use, and then it would be used both by people with the healthy degree of suspicion and people who don’t care about safety. But 99% of them would also use a software which is easy to use and not provably safe… which is much cheaper to develop.
There was no interest in it. Not from the casinos OR from the customers who were betting their money.
What’s interesting here is that this was, at least initially, the motivation for SatoshiDice: that you could prove their lottery was fair. It’s exploded in popularity far more than that seems to warrant. I’ve never used it, so I don’t know how convenient it is, but if it’s as convenient as it sounds (no need to register accounts or log in, just send bitcoins to addresses, AFAIK), that’s more evidence that convenience is Really Important.
But there is SatoshiDICE, which is designed along similar principles and seems to get some use.
I guess part of the problem is that people who play roulette are not very sophisticated about it, since it’s an expected loss anyway. I could imagine poker players being interested, but this protocol doesn’t really fit: it reveals what the deck was at the end of the round (players who folded do not want other players to know what their hand was). I think that the attack that online poker places worry the most about is colluding players, and P2P cryptography doesn’t help there.
You propose to create a system that has the potential to be highly dangerous and illegal. In such a system I don’t think it’s easy to set up a web of trusts and trustworthy elections.
In a straightforward manner, the system would allow anoymous trading on stock options. We have laws against insider trading that you can’t effectively enforce when you have an anoymous prediction market.
If I can anonymously place a bet of $1,000,000 that a specific person dies, than I also have an assassination market.
I note that perfectly legal bookmakers in Ireland already accept bets that President Obama will be assassinated while in office. link
It isn’t precisely an assassination market, because people other than the assassin can join in the bet, diluting the odds.
But, yes, I take your point. With a single central authority, such bets could be rejected as invalid. With multiple authorities, it would require all of them to reject it. On the other hand, authorities that don’t reject such bets as invalid might end up with a poor reputation for reliability.
One way to deal with Validating Questions would be to automatically allow questions that match an approved list of formats. To add to or amend the list of formats you’d circulate a proposed definition and allow other proposals to be added, then let people meeting some criteria (eg users with more than X bitcoins in escrow on currently outstanding bets) to vote the different definitions up or down over a 7 day period.
So, for example, one format might be:
{DOW_JONES} on {FUTURE_DAY} will match {PRICE_RANGE}
Where, for example, {DOW_JONES} for a particular {FUTURE_DAY} is defined as the closing price of the Dow Jones Industrial Average, as retrieved at noon on the day after {FUTURE_DAY} by looking at a list of 5 URLs such as:
parsing each one via a supplied perl regexp or similar to get just a single number, and then taking the median value of that number.
In effect, the bet would be actually about values on a number of different web pages, rather than about the Dow Jones index itself, and the people placing the bets would be assuming the risk that the web pages on the resolution day didn’t reflect the actual stockmarket price they expected it to.
Another way would be to have different ‘authorities’. Each authority can pick which proposed questions are valid for that authority. If someone wishes to place a bet on a question that’s been validated by at least one authority, then they get a list of authorities available for that question, and choose which one to use (which may be influenced by whether there are others betting via that authority in the other direction).
Different authorities could use different systems. So one might required the pre-agreed format style, another might use voting to resolve and allow any question, a third might use a panel of experts and allow free text questions by only in one domain (eg sports).
A reputation system could keep track of how popular different authorities are for different domains of question.
You’d have to pay a percentage of proceeds to the authorities in order to monetize staying an authority over selling your reputation for a guaranteed win in bitcoins. Especially in a single-authority system.
The basketball score is 94-72 late in the fourth quarter. “Second team wins” contracts go down to pennies on the dollar. But, the arbitrating authority has the fix in for “Second team wins”, and buys out massive numbers of the same contract, then takes vast quantities of money out of the system.
For sports, however, there’s an easy solution. Just have Google be your authority. So, for figuring out who won the Oregon-UCLA game last night, the system would pay out on the results of the Google search “Oregon UCLA basketball score” as of a certain time. Reading your earlier posts this seems like what you already described, though.
You’d have to pay a percentage of proceeds to the authorities in order to monetize staying an authority over selling your reputation for a guaranteed win in bitcoins.
Depends on the authority type. The pre-agreed format type of authority shouldn’t need paying; nor one that relies upon mass voting to make decisions. Because in neither case is there a small group of controllers with the power to ‘fix’ the result.
One that uses a panel of experts will need some way to prevent a cabal of experts colluding together, and some form of reward for their time. The reward might be the reputation they gain from being seen as an expert (like with WikiPedia). The anti-collusion measure might be by having quite a large panel and selecting 7 members from it at random for each decision (preferably, selected only after the bets have been laid). However, yes, the system ought to be set up in a sufficiently flexible manner to allow each authority to be able to set in advance a rate of ‘cut’ for bets using them as the decider (with competition between authorities keeping such rates low).
So, for figuring out who won the Oregon-UCLA game last night, the system would pay out on the results of the Google search “Oregon UCLA basketball score” as of a certain time.
You need to have someone that you trust to perform that Google search and read it and then decide who won the money. Otherwise the two people who bet might disagree on what the Google search says.
A quick search shows that there are lots of websites providing a ticker of sports results in XML form. Rather than doing a google search which, as you say, would need interpretation, I think it would be more reliable to pick in advance a basket of such XML feeds, where you know in advance the precise format they’ll use to announce the result.
If by “the current leading Bitcoin predictions market” you mean Bets of Bitcoin, that’s a single website, not a distributed application, and is susceptible to being shut down.
I thought the distributed part referred to Bitcoin or its epigones.
I don’t think there’s any ‘distributed application’ which seems fitted for a prediction market, except, well, a website. Really, if Bets of Bitcoin were to change its payout rules and reboot as a Tor hidden service, that’d basically be ideal (at least currently, while combinatorial prediction markets are being worked out).
I don’t see any intrinsic reason why one couldn’t be written. Splitting the problem into pieces:
Posing questions A distributed database could hold these
Validating questions You’d want some form of reputation system or elected panel of moderators, to verify which newly posed questions had the required properties (eg not a duplicate, clear criteria on how the question is to be resolved, etc). Again, elections and webs of trust can both be handled in a distributed fashion with no centralised authority.
Resolving questions Marking which side of the ‘bet’ wins, or if it is still unresolved. Similar issues to validating, though more at stake. Need perhaps the possibility of submissions of evidence and counter-submissions.
Placing bets Easiest way would be for the person to place their sum of money in escrow, encrypted with the key spread out and requiring X out of Y holders of the key parts to release it upon resolution. Alternatively allow people to purchase insurance from ‘market makers’ who, in turn, have large sums in escrow backing the bets. No central authority or website required, though you could certainly use websites as convenience aggregator gateways to the underlying distributed data.
Metadata Keeping track of reputations of who is winning how much, which bets are busy or have a lot staked upon them, etc.
Can you think of an aspect of having a prediction market that REQUIRES there to be a single central website?
Merely a Chesterton’s fence: the observation that fully distributed websites are rarer than hen’s teeth, there have been recent failures (anyone using Diaspora?), and there’s almost no infrastructure anywhere which relies purely on distributed webs of trust and escrow despite PGP and GPG being around for decades. (Debian sorta works on a web of trust for developers; it’s also far from the largest Linux distribution these days.)
Why didn’t Silk Road spring up instantly after Bitcoin was released? Why is it a single website? Why does it use PGP keys for encrypting messages between buyers and sellers only? Why do all its competitors replicate this centralized single site model?
My immediate thought is: because it is easier
For a distributed service of this sort to spring up ‘naturally’ would require not just a threat model that requires it, but a critical mass of people who care sufficiently about the threat.
Crypt21 Cryptographic Library—“This ain’t your daddy’s crypto!”
And apparently they don’t exist. That’s a pretty good reason.
sighs Yes, you’re right.
I once wrote an open source cryptographic system that would let casinos prove to their customers that the casinos weren’t fixing the random generation behind the wheel spins, dice and card deals.
There was no interest in it. Not from the casinos OR from the customers who were betting their money.
There’s probably an interesting bias behind whether people properly take into account threats that are low frequency but high penalty.
I think there’s probably not much overlap between the sort of people who would understand the meaning of an open source cryptographic system enough to trust it, and people who frequent casinos, except probably for card counting.
If anything they seem to overemphasize low frequency, high-penalty threats. Or at least low-frequency, high-drama ones. Consider gun murder, terrorism, and child kidnapping; versus car accidents or family sexual abuse.
Maybe it is just a black-and-white thinking from most people. Either you trust online casinos, or you don’t. If you don’t, you don’t use them. If you do, you don’t use special cryptography to prove that they don’t cheat.
And the few people who understand cryptography and want to use online casinos and have a healthy degree of suspicion… are not enough to create a profitable market.
You could create a software which is easy enough to use, and then it would be used both by people with the healthy degree of suspicion and people who don’t care about safety. But 99% of them would also use a software which is easy to use and not provably safe… which is much cheaper to develop.
What’s interesting here is that this was, at least initially, the motivation for SatoshiDice: that you could prove their lottery was fair. It’s exploded in popularity far more than that seems to warrant. I’ve never used it, so I don’t know how convenient it is, but if it’s as convenient as it sounds (no need to register accounts or log in, just send bitcoins to addresses, AFAIK), that’s more evidence that convenience is Really Important.
Maybe poker players. Also how practical is your system, e.g., how long would it take to shuffle a deck of cards with it?
But there is SatoshiDICE, which is designed along similar principles and seems to get some use.
I guess part of the problem is that people who play roulette are not very sophisticated about it, since it’s an expected loss anyway. I could imagine poker players being interested, but this protocol doesn’t really fit: it reveals what the deck was at the end of the round (players who folded do not want other players to know what their hand was). I think that the attack that online poker places worry the most about is colluding players, and P2P cryptography doesn’t help there.
You propose to create a system that has the potential to be highly dangerous and illegal. In such a system I don’t think it’s easy to set up a web of trusts and trustworthy elections.
In a straightforward manner, the system would allow anoymous trading on stock options. We have laws against insider trading that you can’t effectively enforce when you have an anoymous prediction market.
If I can anonymously place a bet of $1,000,000 that a specific person dies, than I also have an assassination market.
I note that perfectly legal bookmakers in Ireland already accept bets that President Obama will be assassinated while in office. link
It isn’t precisely an assassination market, because people other than the assassin can join in the bet, diluting the odds.
But, yes, I take your point. With a single central authority, such bets could be rejected as invalid. With multiple authorities, it would require all of them to reject it. On the other hand, authorities that don’t reject such bets as invalid might end up with a poor reputation for reliability.
How would you deal with Validating and especially Resolving Questions?
One way to deal with Validating Questions would be to automatically allow questions that match an approved list of formats. To add to or amend the list of formats you’d circulate a proposed definition and allow other proposals to be added, then let people meeting some criteria (eg users with more than X bitcoins in escrow on currently outstanding bets) to vote the different definitions up or down over a 7 day period.
So, for example, one format might be: {DOW_JONES} on {FUTURE_DAY} will match {PRICE_RANGE}
Where, for example, {DOW_JONES} for a particular {FUTURE_DAY} is defined as the closing price of the Dow Jones Industrial Average, as retrieved at noon on the day after {FUTURE_DAY} by looking at a list of 5 URLs such as:
https://www.google.com/finance/historical?cid=983582&startdate=Dec+25%2C+2012&enddate=Dec+26%2C+2012
parsing each one via a supplied perl regexp or similar to get just a single number, and then taking the median value of that number.
In effect, the bet would be actually about values on a number of different web pages, rather than about the Dow Jones index itself, and the people placing the bets would be assuming the risk that the web pages on the resolution day didn’t reflect the actual stockmarket price they expected it to.
This greatly limits the kind of questions that can be asked.
Another way would be to have different ‘authorities’. Each authority can pick which proposed questions are valid for that authority. If someone wishes to place a bet on a question that’s been validated by at least one authority, then they get a list of authorities available for that question, and choose which one to use (which may be influenced by whether there are others betting via that authority in the other direction).
Different authorities could use different systems. So one might required the pre-agreed format style, another might use voting to resolve and allow any question, a third might use a panel of experts and allow free text questions by only in one domain (eg sports).
A reputation system could keep track of how popular different authorities are for different domains of question.
You’d have to pay a percentage of proceeds to the authorities in order to monetize staying an authority over selling your reputation for a guaranteed win in bitcoins. Especially in a single-authority system.
The basketball score is 94-72 late in the fourth quarter. “Second team wins” contracts go down to pennies on the dollar. But, the arbitrating authority has the fix in for “Second team wins”, and buys out massive numbers of the same contract, then takes vast quantities of money out of the system.
For sports, however, there’s an easy solution. Just have Google be your authority. So, for figuring out who won the Oregon-UCLA game last night, the system would pay out on the results of the Google search “Oregon UCLA basketball score” as of a certain time. Reading your earlier posts this seems like what you already described, though.
Depends on the authority type. The pre-agreed format type of authority shouldn’t need paying; nor one that relies upon mass voting to make decisions. Because in neither case is there a small group of controllers with the power to ‘fix’ the result.
One that uses a panel of experts will need some way to prevent a cabal of experts colluding together, and some form of reward for their time. The reward might be the reputation they gain from being seen as an expert (like with WikiPedia). The anti-collusion measure might be by having quite a large panel and selecting 7 members from it at random for each decision (preferably, selected only after the bets have been laid). However, yes, the system ought to be set up in a sufficiently flexible manner to allow each authority to be able to set in advance a rate of ‘cut’ for bets using them as the decider (with competition between authorities keeping such rates low).
You need to have someone that you trust to perform that Google search and read it and then decide who won the money. Otherwise the two people who bet might disagree on what the Google search says.
A quick search shows that there are lots of websites providing a ticker of sports results in XML form. Rather than doing a google search which, as you say, would need interpretation, I think it would be more reliable to pick in advance a basket of such XML feeds, where you know in advance the precise format they’ll use to announce the result.