🏅 Top traders

This can resolve NO.

@MartinRandall Resolving NO on the grounds that the existing stock markets don't seem as good/useful to me as the any other market types, as of Apr 21.

I would like a "freeform" market that allows the market creator(perhaps a bot) control over every primitive market mechanism. It'd be a general "abstract stock", useful as public bookkeeping for payout calculations.

Most people should not create freeform markets or bet in them without reading the details or trusting the creator.

Primitives include:

• Adding mana to a pool

• Creating shares into a pool

• Giving shares to a user

• Giving mana to a specific shareholder from the pool

• Dividends: Giving mana to every shareholder from the pool

• Resolving a market distributes the liquidity to every shareholder.

• Canceling a market distributes liquidity to subsidizers.

• Closing: Enabling/disabling limit orders from being entered.

• Shares bought/sold with limit orders only

• The official Manifold account does not provide automatic liquidity to these markets, since trader/bot limit orders would handle everything and the semantics given too much freedom would complicate it.

• (Optional) Multiple shareclasses, liquidity is split according to a distribution

• (Optional) Loans can be given to the market itself, or the markets can be ineligible for loans.

• (Optional) Negative shares could be allowed, but they might also cause moderation hassle because a dividend could take money out of someone's account at any time.

Examples include:

• Investing in someone's bot trading strategy. "Dividends" would be the bot reinvesting subsidizing the market and then triggering a redistribution.

• Automated derivative markets: Options trading on a numerical market("I don't think Mira's bot will be successful, but to hedge against it taking over the entire market I'd like call options on shares at some very high price. I expect to lose money on the options, but they're valuable anyways."

  • In this example, shares would've been created in the pool earlier. When the option is exercised, the bot would assign those shares to the user exercising the option. In this way, two markets can have their semantics "linked", where shares in one directly convert to shares in the other.

• Large or complex markets: For something like the US 2024 election, maybe it's annoying having 300 variations on who wins what state or county. People might want a different and external betting interface, that keeps 50 share classes in the same market for every politician and pays out based on a scoring rule or something. Consider that multiple-choice or free response markets are kind of like this and people want to use them, but the payout structure is too rigid so they settle for binary markets instead.

  • For example, someone could buy "Trump stock", and the scoring rule could say: "He won 20 out of 50 states, but didn't get elected, but the state representatives all voted for him; and given a scoring rule, your payout is X" which sounds complicated, but is less complicated than trying to hedge against every possibility and keep them up-to-date using many binary markets.

• Experimenting with new market types. It's a lot easier for someone to write their own bot using the base mechanisms, than to get an official feature added to the site. And then if something's popular, maybe you add special support for that type of market later.

  • For example, "quadratic funding" markets were implemented and seem to be gone now? If you had a generic freeform market, it's much cheaper to roll out a bot or market-management script than to change the backend/frontend/etc. and work out all the details upfront for every experimental market type you want to add.

• Oracle markets: If somebody has an observable like "share price of $SPY on close of 2023-04-12", we could have one place on the entire site that defines that instead of 100 duplicate markets and creators. If derivative markets were possible, then many markets could get automatically paid out based on shares created/trigger by a single permanent oracle market that defines what the numbers mean. So there'd be less moderation hassle overall resolving random people's markets that forgot about the site.

This sounds similar to the one linked in your description, with the distinguishing feature being "This market type exposes the market primitives in the list above to the market creator", rather than trying to guess everything someone might want to do with it.

@Mira Kudos for the detailed writeup! We currently have something implemented that's about 50% of what you've asked for (supports general stocks, dividends, with a CPMM for providing liquidity) -- but it doesn't have limit order or API support at the moment, along with a whole range of other issues.

I do think it'd be really great if we could simplify down our API to a point where individual bots can easily create "smart contracts" and surface them to users; though it's not a priority at the moment, unfortunately.

Well it's already better than the quadratic funding markets so

One rough idea I have is to utilise a ranking mechanism:

• People can buy or sell a stock, moving its price

• The price and no of holder determine the its ranking among all stocks. Ranking calculate daily.

• Every time the stock gain a rank, its Yes holder gain and No holder lose some MANA. The opposite for losing a rank. Overall should be zero sum.

• Some details to be fine tuned like exact distribution of rewards, how to segment ranking etc

This could incentive people holding stock that they think are undervalued and go against stock that they found overhyped, relative to other things.

For duplication issue, rules can be specified that market with lower than certain number of trader is exclude from the ranking (so just behave like a binary market now). Or only the top 500 stocks will have ranking. This should reduce visible clutter and encourage people to focus on existing market.

All are rough ideas. Others please scrutnize and try to think of exploit.

@ian has been working on this at https://github.com/manifoldmarkets/manifold/pull/1667 - be excited!