The story of Augur, an Ethereum prediction market

Some gambling (e.g poker, blackjack, roulette) doesn't give predictions about the real world.

Prediction markets are a specific form of gambling that yields predictions about the real world. So all prediction markets are gambling, but not all gambling is a prediction market. One noteable difference between 'prediction markets' and the normal operation of bookies is who sets the prices. Usually, bookies offers odds on a bet, and for every new bet, they can set new odds taking the previous bets into account. This is called 'fixed-odds' betting.

In 'prediction markets' people on all sides of a bet put money into a pool. Suppose the total pool has a value of $Z, you put in $X on outcome A, and in total people put $Y on outcome A, then if outcome A is the true outcome, you get $ Z * (X / Y) for your bet. This is called parimutuel betting.

Beyond parimutuel betting, the idea is that once the pool is fixed, you can sell your bet. So in the above scenario, I could sell my X/Y share in outcome A to someone else. Later, the price of such shares allow us to track the markets opinion on changing odds.

In reality, I think they tend to skip the parimutuel part, and instead just auction of an equal amount of shares for each outcome. The proceeds of this initial auction then become the shared pool.

> I sat through some early presentations of Augur at the big monthly Bitcoin meeting that's been running for years in Amsterdam, and my first thought upon seeing it was, "who says what the factual outcome is?".

Is that a genuine question, or are you confused about the consensus mechanism? The answer to 'who says' is the core of what Augur is: people vote on what the outcome was, in a way that incentivizes them to vote for the answer they believe to be correct.

> So it's back to the original problem with smart contracts - you still have to trust an oracle, and I would bet ;) that if the stakes are high enough, an oracle will be manipulated in some way.

No, you do not have to trust an oracle. The whole mechanism is actually quite cool, you should check it out:

https://www.augur.net/whitepaper.pdf

> Let's take an election for example. We've had enough madness just in the US in the last 20 years to be skeptical of initially reported outcomes. In fact, some elections get reversed after study and investigation. Augur isn't going to reverse a funds exchange. And if someone has enough riding on an outcome, they may find it worth spending extra money to at least temporarily cause the outcome to match their bet. They just need to win briefly, even if the results get overturned later.

That's easy to solve. You simply define the question in a way that obviates those kinds of issues. In Augur, people vote on the outcome they believe is correct. The outcome of this process is defined as the truth, within the system. Since this is transparent, you place your bets with this in mind. You aren't betting on who will win the 2020 election, you're betting on which answer people will converge on for the winner of the 2020 election. A subtle, but important difference.

This is true, so it comes down to how well the consensus process reflect reality. It's a pretty interesting process: there are multiple layers of dispute escalation, and the ultimate backstop is that Augur splits in two. You keep REP only on the side of the split you voted for.

The theory is that the version of REP that reports the truth will be more valuable on the open market, and that the split is a credible threat which keeps everyone honest, so it will seldom actually happen.

Numerai is a much superior data prediction tool imo... don't know what all the fuzz is abt with augur.

If you are curious about the Augur usage metrics you can check http://dappboard.com/app/application/5a58431c5d7064dc7fb37cc...