A handful of questions about Hivemind

[bitedge]

Hi guys, Until recently I was a doubter of prediction markets as it seemed they were just betting operators trying to disguise themselves as financial products. Hearing Paul Sztorc, Robin Hanson and the Augur guys I have realized prediction markets can be very powerful and I want to learn more and share it with others.

I am James Canning, Webmaster of http://www.bitedge.co/ a leading source of information about bitcoin gambling and I am going to write one or two blog posts about bitcoin prediction markets. I hope you can answer a few questions to clarify things for me.

1. What is the relationship between Truthcoin and Hivemind? I thought Hivemind was the new name for Truthcoin meaning Truthcoin was no longer a thing but then I see activity on the Truthcoin website and docs after Hivemind already exists so is it more like Truthcoin is the concept/invention/protocol and Hivemind is one implementation of that concept/invention/protocol?

2. About predictions with scalar outcomes like how many Electoral College votes will Hillary win is it the case that there would/could be a prediction that pays 1/538 of $1 per vote won? So if she wins 0 votes its worth $0.00, if she wins 269 votes it pays out $0.50c if she wins 538 votes it pays out $1.00c and so on for every amount in between?

If that is how it would work and the market price for that prediction is $0.50c how would a new trader come along and bet that Hillary will win less than 269 votes? This trader does not hold the prediction to be able to sell it so is there a shorting mechanism or could/would there be an equal and opposite prediction available for sorters that pays out $0.00 if Hillary wins 538 votes, pays out $0.50c if Hillary wins 269 and pays out $1.00c if Hillary wins 0 votes?

3. Do you imagine predictions being tradable outside the Hivemind system/sidechain? I guess by the keys for those predictions being traded privately potentially at a different price than the price on the Hivemind system/sidechain? 

If predictions could be traded in the dark, off market does that degrade the price signaling that is required for wisdom of the crowd accuracy?

Although someone who trades a prediction at a different to market price has an incentive to arbitrage that trade on the market he might have real life knowledge that the prediction will soon go to $1.00c or $0.00c and real life incentive to not let the market know that. In such a case he might trade the predictions off market in a way that would change the price up if it were happening on market. In this case the market and the prediction does not gain from his knowledge right?

4. When a market creator makes a new market how do the predictions get their initial price? Does the creator set the initial price by setting a loss limit per outcome? Like if he sets a $6.00c loss limit if Hilary wins and a $4.00c loss if Hilary losses does that mean the prediction that Hillary will win's initial price will be $0.60c?

5. Are betting exchanges prediction markets? I do understand the difference in trading mechanism, the concept of a prediction being trading between parties as different to parties matching each other's bets and that the logarithmic market scoring rule will give Hivemind the affect of liquidity. 

Bur isn't the valuable part of a prediction market, the fact that the price implies a wisdom of the crowd probability, exactly replicated by a betting exchange?

If I go to Betfair and check the odds on Man U Vs Arsenal those odds are just as much a wisdom of the crowd probability with all the same incentives at play that would be for a prediction market right (except Betfair is centralized and non-anonymous)?

I get that Hivemind is decentralized and anonymous but that is just as much a case in favor of a decentralized anonymous betting exchange is it is in favored of Hivemind.

So what would be the advantages or disadvantages of a decentralized anonymous betting exchange compared to Hivemind? Particularly if there are any advantages or disadvantages to the accuracy of its price signal as a forecast.

6.  I assume a lot of markets will be created automatically by bots and I assumed they would be voted on by bots as well. But then I gather what each voter has to vote on is random which would make bot voting more difficult so do you think people will do voting manually by looking stuff up on google?

Thanks for any info you can give, I have had a few ahha moments with this and I think I will get a lot more involved.

[psztorc]

Hi James,

Always happy to help.


> 1. What is the relationship between Truthcoin and Hivemind? I thought Hivemind was the new name for Truthcoin meaning Truthcoin was no longer a thing but then I see activity on the Truthcoin website and docs after Hivemind already exists so is it more like Truthcoin is the concept/invention/protocol and Hivemind is one implementation of that concept/invention/protocol?

Names don't really matter, do they? Especially for a protocol, where one can use it without knowing the name. A protocol is just a set of rules, and the rule-set doesn't need to include it's own name, or any names.

Nonetheless, Truthcoin was the original name for the concept/invention/protocol, and it was also the name of one implementation of that concept, and thirdly it was the website where I hosted the project as well as my blog. However, people were a little uncomfortable with something that matched the Altcoin naming-schema, so I split my one website into two...one for both my blog and the project (emphasizing the blog), and a new website 'hivemind' for only the project and project-related content.


> 2. About predictions with scalar outcomes like how many Electoral College votes will Hillary win is it the case that there would/could be a prediction that pays 1/538 of $1 per vote won? So if she wins 0 votes its worth $0.00, if she wins 269 votes it pays out $0.50c if she wins 538 votes it pays out $1.00c and so on for every amount in between?

Scalar outcomes have 2 states: one for 'min' and one for 'max'. If, say, the prices are $0.50 and $0.50, and the max and min chosen were "0" and "538" (although these would probably indeed be the choices, this is not an absolute requirement -- someone might consider those outcomes to be excessively extreme, and might pick "40" and "500"), then this would correspond to an expectation of 269 (which is halfway between the two extremes. So it is very similar to what you've laid out, but not quite.


> 3. Do you imagine predictions being tradable outside the Hivemind system/sidechain? I guess by the keys for those predictions being traded privately potentially at a different price than the price on the Hivemind system/sidechain? 

It is certainly possible. I designed a way for this to be done which is conceptually similar to Bitcoin's lightning network. In addition to that, it is possible to move the 'shares' to the a trusted third party website, at which point users would have all of the benefits and drawbacks of a traditional database.


> If predictions could be traded in the dark, off market does that degrade the price signaling that is required for wisdom of the crowd accuracy?

No. In the very worse case, the signal is degraded to the exact extent of "someone who was going to make a trade, does not make one at all", but in all other cases the info quickly diffuses through the rest of the market.


> Although someone who trades a prediction at a different to market price has an incentive to arbitrage that trade on the market he might have real life knowledge that the prediction will soon go to $1.00c or $0.00c and real life incentive to not let the market know that. In such a case he might trade the predictions off market in a way that would change the price up if it were happening on market. In this case the market and the prediction does not gain from his knowledge right?

Perhaps not at first, but buyers (of anything) will always seek the best deal (and 'buyers' includes this "insider"), and arbitrageurs will always seek risk-free profits, so I don't think this will be a problem. Markets do spread information much much faster than, say, official reports or scientific papers or expert testimony or conversations, but they do not work magically, they do need a little bit of time to work.


> 4. When a market creator makes a new market how do the predictions get their initial price? Does the creator set the initial price by setting a loss limit per outcome? Like if he sets a $6.00c loss limit if Hilary wins and a $4.00c loss if Hilary losses does that mean the prediction that Hillary will win's initial price will be $0.60c?

When the market is created, it has uniform prices. However, the creator of the market is the first to know that the market has been created at all, so he is in an opportune position to make an the very first trade on this market. If his initial odds are well-selected, he will profit as a result.


> 5. Are betting exchanges prediction markets? I do understand the difference in trading mechanism, the concept of a prediction being trading between parties as different to parties matching each other's bets and that the logarithmic market scoring rule will give Hivemind the affect of liquidity. 

> Bur isn't the valuable part of a prediction market, the fact that the price implies a wisdom of the crowd probability, exactly replicated by a betting exchange?

A "bet" is almost a prediction market...each bet is missing some of the 'market' part: that the betting odds are dynamically repriced as (many) individuals trade (in the same place).

> If I go to Betfair and check the odds on Man U Vs Arsenal those odds are just as much a wisdom of the crowd probability with all the same incentives at play that would be for a prediction market right (except Betfair is centralized and non-anonymous)?

Betfair is very close to a prediction market, yes. Dynamic parimutuel betting is mathematically similar (and, sometimes, equivalent) to market scoring rule betting. The key thing is that the odds are dynamically repriced.

I get that Hivemind is decentralized and anonymous but that is just as much a case in favor of a decentralized anonymous betting exchange is it is in favored of Hivemind.

> So what would be the advantages or disadvantages of a decentralized anonymous betting exchange compared to Hivemind? Particularly if there are any advantages or disadvantages to the accuracy of its price signal as a forecast.

They would be very similar. Markets tend to do well with scale, as do blockchains, and technology in general, in my opinion.


> 6.  I assume a lot of markets will be created automatically by bots and I assumed they would be voted on by bots as well. But then I gather what each voter has to vote on is random which would make bot voting more difficult so do you think people will do voting manually by looking stuff up on google?

They will probably have software to help them, and manually check things over to make sure the software hasn't screwed up.


> Thanks for any info you can give, I have had a few ahha moments with this and I think I will get a lot more involved.

My pleasure.