Show Posts
61
General Discussion / Re: A Call to the Truthcoin Prediction Market
« on: April 02, 2014, 01:18:08 pm »
These are good questions, much better than I expected. (I expected people to not even read the paper, and make all kinds of wrong claims about the accuracy or desirability of PMs).
I don't know a lot of crypto, there probably is a better way of doing this. Of course, by the time you expose that key, it is useless, but I agree that automated (or even manual) exposure of private keys is inconvenient and a risk. There needs to be a way to encrypt the votes in a way that they can't be decrypted by anyone else, if there were nothing at stake individuals could publish the decryption key and allow other people to read their votes, which would weaken (and complicate) voting strategy in a number of ways.
Yes I have. In the ancient past, one can observe these lingering notes in line 237 the original R code:
But I later decided it wasn't worth it. I don't remember all of the reasons for that decsion. Clearly, it would take more time. I was thinking an average across outcome-results, with later results counting more (as manipulators are short term). However, I reasoned that, if the first outcome were 'wrong' the total smoothed outcome would now be more likely to be wrong than right. In a way, it unseals some of the votes in advance, disrupting the coordination game.
If the dispute completely replaced the old outcome, then I think I was also worried that attackers would abuse it by endlessly lodging disputes, but you might be on to something there by making it expensive. I'm not sure the benefit of this outweighs the cost, however.
Although the entire incentive scheme is designed on off-path reasoning (if people lie, if people refuse to vote), the Nash Equilibrium has everyone voting on all the outcomes [market depth & market price are irrelevant] on their branch (this is a major reason for branches...too much work for voters). I really do expect participation to be nearly 100% all the time, particularly in the medium to long run.
Yes you've missed this part:
Check the Appendices for how I punish people who refused to vote. This will have to be **clearly** explained to shareholders (as if a large proportion of them do nothing, their shares will actually start to disappear). One could design a class of 'owner-only' shares, but this separates ownership from control, introduces agency cost, and otherwise threatens the model. My research into the financial history of this planet indicates to me that separation of ownership from control produces something permanently unworkable. Bitcoin miners control Bitcoin, and are paid in Bitcoin. Bitcoin owners can only destroy their own Bitcoin. Perfectly logical.
1. It seems you have to expose your private keys when you want to vote, I think there may need to be some alternative to this. (I know NXT has very low 'forging' participation rate, a) because the incentives are low but b) because users don't want to frequently expose their private keys to potential keyloggers etc.)
I don't know a lot of crypto, there probably is a better way of doing this. Of course, by the time you expose that key, it is useless, but I agree that automated (or even manual) exposure of private keys is inconvenient and a risk. There needs to be a way to encrypt the votes in a way that they can't be decrypted by anyone else, if there were nothing at stake individuals could publish the decryption key and allow other people to read their votes, which would weaken (and complicate) voting strategy in a number of ways.
2. Have you considered a fail-safe dispute resolving mechanism?
Yes I have. In the ancient past, one can observe these lingering notes in line 237 the original R code:
Quote from: https://github.com/psztorc/Truthcoin/blob/40b859cc3d396af3e2ab068be3f23c8312db6e4b/lib/consensus/ConsensusMechanism.r
#Voting Across Time
#Later Votes should count more
#! ...simple change = ConoutFinal becomes exponentially smoothed result of previous chains.
#! require X number of chains (blocks) before the outcome is officially determined (two weeks?)
# Will need:
# 1] Percent Voted
# 2] Time Dimension of blocks.
But I later decided it wasn't worth it. I don't remember all of the reasons for that decsion. Clearly, it would take more time. I was thinking an average across outcome-results, with later results counting more (as manipulators are short term). However, I reasoned that, if the first outcome were 'wrong' the total smoothed outcome would now be more likely to be wrong than right. In a way, it unseals some of the votes in advance, disrupting the coordination game.
If the dispute completely replaced the old outcome, then I think I was also worried that attackers would abuse it by endlessly lodging disputes, but you might be on to something there by making it expensive. I'm not sure the benefit of this outweighs the cost, however.
Although the entire incentive scheme is designed on off-path reasoning (if people lie, if people refuse to vote), the Nash Equilibrium has everyone voting on all the outcomes [market depth & market price are irrelevant] on their branch (this is a major reason for branches...too much work for voters). I really do expect participation to be nearly 100% all the time, particularly in the medium to long run.
In Truthcoin's current form, do people bet using Bitcoin and pay a trading fee that gets paid to Truthcoin holders or do they buy Truthcoin and use that to bet?
Yes you've missed this part:
For c, in this blockchain the 'Coins' *are* reputation. There are basically two coins at once (although the network-currency "Bitcoin" are just inputs in a signed message to the market maker). Then they are 'withdrawn'. This withdraw thing is a little complicated, but some people on Bitcointalk helped me with several possible solutions, which are in the Implementation Details section.
Check the Appendices for how I punish people who refused to vote. This will have to be **clearly** explained to shareholders (as if a large proportion of them do nothing, their shares will actually start to disappear). One could design a class of 'owner-only' shares, but this separates ownership from control, introduces agency cost, and otherwise threatens the model. My research into the financial history of this planet indicates to me that separation of ownership from control produces something permanently unworkable. Bitcoin miners control Bitcoin, and are paid in Bitcoin. Bitcoin owners can only destroy their own Bitcoin. Perfectly logical.
