Which betting system will Manifold be using on April 1st?

Plus

Ṁ14k

resolved Apr 5

70%60%

CFMM (constant function) using @Pepe's k = y^(1-p) * n^p

17%15%

CPMM (constant product market maker, a la Uniswap)

11%12%

CPMM for binary markets, something else for free-response markets

2%9%

DPM (dynamic parimutuel; our current system)

0.0%Other

0.0%

CLOB (central limit order book)

0.5%

LS-LMSR (Liquidity sensitive log market scoring rule, variant of Hanson's market maker)

Multiple options will be supported, with market creators choosing between them.

Help us decide whether to keep our current DPM system or switch to something different. You can influence the outcome of this market by suggesting an alternative or arguing in favor of one of these options. Close date updated to 2022-03-15 5:50 pm Close date updated to 2022-03-15 5:47 pm Apr 4, 9:40pm: The system in place on April 1st was CFMM for binary markets and DPM for free-response markets. (Since our CFMM is also a CPMM, this answer got some weight in the final resolution.)

This question is managed and resolved by Manifold.

Get

1,000

and

3.00

14 Comments

38 Holders

178 Trades

Sort by:

You can sell your current position and buy in using the Manifold account.

Is there anywhere summary how this different systems work?

@Robert McIntyre (gold) "One of the things I want to do is make "knowlege hunt" markets, which almost work like quora questions but with actual betting" Something like https://manifold.markets/M/looking-for-compilationoverview-of ? Multianswer with well fed ante seems like a good solution.

Thankfully I'm not the one working on our market maker~ I realize now that the right thing to do was probably to buy from the Manifold account regardless; ah well! I hereby commit to not being the one to resolve this market.

Uh oh, if too many people buy into this answer Austin is no longer incentivized enough to actually implement it...

Small bet because coding is hard and getting this done in that time frame, such a central feature and the possibility to kill interest in the platform if it is wrong. I think this decision will take time. I don’t think there is a major problem with the system. They could tune the cost function in favour of earlier bettors. In other news, i have too many side project ideas but a cool idea is a simulation of all the options, along with common trading scenarios. Getting the intuitive feel for the mechanism is as important or more so than understanding the maths. At least until this is real money with HFT bots and what have you.

Would like to see a comparison writeup of all these options in terms of how they look to the lay user. What situations do markets get funky in? How accurate are the given estimates? What's the winning strategy?

I'm pretty surprised to realize that the implied odds are not how you should be choosing where to place bets in the current system. That strikes me as missing the basic point of a prediction market: people should be able to make profitable bets when they think the market is wrong. It seems obvious to me that whatever system is used needs to have that property. Have the manifold developers explained anywhere why they don't like the alternative approaches?

Not an opinion, just a bet 😊

When I prototyped my own prediction market thingy I went with LMSR over DPM because I think it is important that early predictions are rewarded more than later ones. I wasn't aware of CPMM until now so I can't comment on that.

@johnnbeshir One of the things I want to do is make "knowlege hunt" markets, which almost work like quora questions but with actual betting. (like this one: https://manifold.markets/RobertMcIntyre/scipapers-generational-inheritance) Do you have thoughts on what betting system works well for those?

I'd like "Payout if chosen" to indicate how much profit I get if I'm right. That is, buying should lock-in that profit number.

DPM's liquidity is really nice. A case against it, I think "buying YES when the market is below your estimated probability is positive in expectation" should be guaranteed without needing to assume that you're the last trader, and without that I feel quite lost in knowing when and how buying/selling is actually in my benefit, or in interpreting implied probabilities given that they could be at equilibria in places other than the probabilities people assign. Right now my bigger profits have been by going to already resolved or almost resolved markets that are behind on price updates, and buying them to their actual resolution, taking return from the people who estimated the correct outcome earlier so they get a lot less than they would have had projected to them, making their prediction probably negative EV from their perspective. This doesn't feel like a stable equilibrium. There's a lot of interesting markets (like the Ukraine ones!) subject to sudden resolution and avoiding them outright is probably not great, so I'm bearish on "learn to make and use only markets for which the difference between final return and initial projections of return are small" as a sufficiently encompassing solution.

I don't have a confident opinion on the right answer here, but I'm bullish on both CPMM and LS-LMSR, and bearish on DPM (per https://manifold.markets/kjz/will-manifolds-developers-agree-wit)

Related questions

Related questions