Augur Turbo's contracts are structured such that they are mostly independent. This makes it easy to upgrade smart contracts without special authority.
The major components are: MarketFactory, Fetcher, AMMFactory, MasterChef, and FeePot. The contracts can be found at https://github.com/AugurProject/turbo/tree/dev/packages/smart/contracts.
Every type of market has its own MarketFactory. The market factory accepts USDC and returns a "complete set", which is one outcome token for each outcome. The user will trade away the outcome tokens they don't want, for those they do, giving them a "position" that pays out a profit if they guess correctly.
Every market factory has the same common code for defining markets, minting outcome tokens, and claiming winnings. These are defined in the contracts "AbstractMarketFactoryV*", with "V3" being the latest.
Each market factory additionally defines rules for market creation and market resolution. Sports market factories encode some of the game rules, like defining spread or specifying how to handle ties. The crypto market factory pulls from the price feeds.
The market factories are mostly built out of other contracts, acting as mixins. The design goal has been to move any common code into mixins. The market factory contracts themselves hold a lot of boilerplate code used to invoke code defined in mixins.
The market factories also define structs for additional information about markets. For sports markets this is the "Event" struct, which ties together several markets that are resolved by the same game scoring information. For crypto this is the "Coin" struct, storing information like price feed address.
The sports supported right now are: NBA, NFL, MLB, and MMA/UFC. Other sports can be added fairly easily using mixins, if they're similar to other sports.
The crypto market factory deployed right now only supports price feeds. The next version, which has been written, supports market cap as well. In actuality the new version supports any Chainlink feeds.
There are some other market factories that aren't supported for one reason or another. Some are older versions that aren't used anymore, like SportsLinkMarketFactory. Others were made in expectation of being used but have been indeterminately delayed or cancelled, like GroupedMarketFactory.
A market factory can contain virtually infinite markets. Iterating over all of them to get the current state of markets isn't always possible because many nodes limit query runtime or response size. To work around this, we have the Fetcher contracts.
Fetcher contracts return paginated lists of "interesting" markets. These are markets that are either unresolved (open) or have unclaimed winning outcome shares. There are two Fetcher contracts that matter right now: SportsFetcher and CryptoFetcher.
This concludes the overview of the market factories. They're all you need to run Augur Turbo. However, in reality you want an easy way to trade, which can be promoted through liquidity provision rewards.
Trading is enabled via BalancerV1 AMMs interacted with via the AMMFactory. There need only be one AMMFactory for all market factories. It handles creating AMMs, buying and selling outcome shares, and interacts with the MasterChef contract to enable rewards.
The ability to set weights is necessary for adding liquidity to Augur Turbo AMMs because of the way outcome tokens are created. Thus, we need to use Balancer AMMs. The ability to use several outcomes is also helpful since most markets have three outcomes due to the inclusion of the invalid outcome.
Why are weights needed? A naive AMM sets the price of tokens by their relative balances in the AMM. This works reasonably well for normal tokens because cheaper tokens are actually easier to acquire in larger quantities. This is not true of outcome tokens because they are minted in equal quantities.
If you have a 50:50 market then a complete set of outcome tokens can be added to the AMM's liquidity pool without any tokens being kept by the LP. But if the price isn't exactly 50:50 then some tokens will be kept by the user lest they move the price towards 50:50.
Normal outcome prices don't differ enough to matter much up until the outcome is known, which causes the price to rapidly approach 100:0. But the invalid outcome (usually labeled "No Contest") is uncommon so it's set to 2%, the smallest default ratio using BalancerV1. To provide the same amount of liquidity for a 98:2 AMM is many times more than needed for a 50:50 market. That inefficiency is extremely undesirable since it locks up so much capital as well as giving the LP a larger position than they may desire.
The initial weights the AMMs use are set at market creation using the lines from the sportsbook. They do not update as the sportsbook lines change.
Note that crypto markets do not have an invalid outcome and always start at a price ratio of 50:50, but still use BalancerV1 AMMs. There's no deep reason: it's easy to use the existing infrastructure needed for the sports markets.
In an attempt to offset impermanent loss, a rewards contract is provided whose purpose is to give incentive liquidity providers. The contract is a highly modified version of the MasterChef contract provided by SushiSwap. It can be used to distribute any ERC20 token as a reward. This is distributed based on percentage of stake in the pool a given account has. Rewards are paid out over a predetermined period of time which is set on a per event/pool basis. There is also an incentive available for early liquidity providers who join within the first 10% of the reward window. Both of these amounts must be set per market factory or will default to 0.
This concludes the overview of trading and rewards. Finally, there's the FeePot contract.
We don't use it right now. It's designed to pay REPv2 holders for providing a secure backstop in case of bad market resolution. That functionality hasn't been implemented yet so the fee that would go to the FeePot is set to zero.
So long as AugurV2 (on Ethereum mainnet) never forks, there will only ever need to be one FeePot contract. If it works then a new FeePot must be deployed.