🏅 Top traders

I have an alternative but half-baked suggestion that will redistribute gains from HFT bots without damaging instant bets.

Currently the AMM is a single percentage. Suppose that a market should be at 50%. Adam bets it to 75%. A bot bets it back down to 51%. The AMM was "smart" to sell those YES shares to Adam for a good price, but "dumb" to sell the NO shares to the bot at a bad price. This transfers mana from Adam to the bot, via the AMM, and the fastest bot wins.

Instead, I propose that after Adam's bet, the AMM will sell additional YES shares at 75%, but it will still sell NO shares at the original 50%. Instead of displaying a market probability of 75%, it displays as 50-75%. Over the course of some minutes with no further bets, the 50-75% raises to 75-75%, as the AMM gains confidence in Adam's bet.

Now, if a bot wants to profit from the market after Adam's bet, it can set a limit NO order at (eg) 51%. A short time later the AMM fills the limit order and displays a probability of 51%. Overall this mostly transfers mana from Adam to the AMM, but the bot takes a cut. If there's only one bot on a market it can take a large cut by waiting for the AMM to update to 75% before correcting it. But if there are multiple bots then they compete to take the smallest cut.

(if there are multiple limit orders at the same percent, maybe split the order between them)

If a user wants to bet in a 50-75% market they can buy NO at 50%, YES at 75%, or they can set limit orders as usual. Maybe there would be a yellow warning for instant bets on an evolving market like this.

@MartinRandall That's an interesting idea. But that seems like it's reducing price accuracy of the market - traders (whether human or bot) who think Adam was wrong have less incentive to correct the price quickly. If they think nobody else will notice they would want to set a high limit order to take more spread, and move it down over time - that sounds like a lot of hassle. If there is a lot of competition then it ends up being kind of like they bid in an auction for how small of a spread they are willing to take - except it's not a formal auction, so it's much more annoying - maybe using an actual auction would be better :)

FWIW what I care about most is I just want more liquidity - how much fast trading is rewarded is a secondary concern to me, and seems like it has pros and cons.

@jack Compare it to call auctions. With a call auction every minute the price updates are delayed by up to a minute. If the AMM takes a minute to go from "50-75%" to "75%" then the maximum delay is the same. In both cases we can tune the delay according to the desired outcome.

If they think nobody else will notice they would want to set a high limit order to take more spread, and move it down over time - that sounds like a lot of hassle.

If you think nobody else will notice and don't want any hassle you can wait for the price to get to 75% and then bet it back down to 50%.

maybe using an actual auction would be better :)

Maybe. How about this:

1. On an inactive market, start an auction when someone makes an instant bet that moves the probability by 1% or more.

2. While the auction is ongoing, display the probability as a range based on instant bets.

3. Instant bets always get at least the shares they were promised, but might get more.

4. Once the auction completes, there may be a surplus. Distribute the surplus between the AMM and the instant bettors according to your values.

Compared to regular call auctions, my goal is to support instant bets, and reduce the losses such bets currently face from arbitrage and other bots.

Here's a paper on effects of (ending) batch auctions at the Taiwan Stock Exchange

Abstract: We exploit the switch from frequent batch auctions to continuous trading at the Taiwan Stock Exchange to show that liquidity did not change significantly for stocks, while efficiency improved substantially for mid-cap and small-cap after trading became continuous. We also find positive and significant abnormal returns for mid-cap and small-cap, whilst for large-cap abnormal returns are positive but not significant. Continuous trading increased profits for fast investors and losses for individual investors in mid-cap and small-cap. The results inform the debate on optimal market design and support previous empirical studies such as Amihud, Mendelson, and Lauterbach (1997) suggesting that investors highly value the opportunity to trade continuously.

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3733682

Frequent Batch Auctions vs. Continuous Trading: Evidence from Taiwan

We exploit the switch from frequent batch auctions to continuous trading at the Taiwan Stock Exchange to show that liquidity did not change significantly for st

@SG Incidentally, my view is the reason TSE batch auctions failed was that the duration (5 seconds) was way too long. The optimal time is probably more like 100-300 milliseconds, slow enough to obliterate pure HFT speed plays, but fast enough for efficiency gains from bot trading.

@SG In the context of Manifold, I think a batch duration long enough to make a difference would likely be a worse user experience: having to wait to a minute or two to get your bet filled is terrible.

@SG What are your thoughts on having a longer auction right after a market is created?

@SG interesting info! What about using auctions in addition to continuous trading, not instead of it? Like the opening and closing auctions. Looks like on the NYSE they account for over 20% of total trading volume.

@jack Placing limit orders while a market is closed seems similar enough to that, and would be useful in other contexts too. ("misleading news just dropped, so I'm closing the market for a day to reduce impulse bets and so everyone can adjust their orders")

@Mira I think the main use I'm thinking of is that if such auctions are run periodically (say every day, for example) they provide a way for people to trade high volumes without either party taking on the risk of the market price moving on news between when you place the limit order and when it fills.

@Mira That’d be nice! Or for closing the market to clarify resolution criteria. I dislike having markets open when resolution criteria are in major dispute, but then if I wanna reopen it the first person snags the “spoils”. An auction like you’re describing might solve that!

@SG Hmm, for me 99.9% of the time I don't care whether my order fills now or 30 minutes from now - what I do care about is whether or not the order will fill, and at what price. Usually, what I want is to trade as much as possible up to a certain amount of price slippage, and while minimizing the risk that the market moves on news in the meantime. So I generally don't like market orders and an order that took longer to allow better price execution would almost always be a big win for my trading.

I recognize that this is probably not the typical trading pattern for most users, but I imagine it's reasonably common for power users.

@SG > Continuous trading increased profits for fast investors and losses for individual investors in mid-cap and small-cap.

And that's bad, right?

@SG I haven't had a chance to investigate deeper but that paper seems to contradict other research I know of. Or you may be right that it all comes down to the frequency. The following paper (by my PhD advisor though I'm not otherwise very up the research) seems to suggest that 200ms is optimal: http://strategicreasoning.org/wp-content/uploads/2013/02/ec38-wah.pdf

Two ways continuous markets have deadweight loss:

1. Obscene expenditure on shaving milliseconds off of trading algorithms

2. Poorer matching of buyers to sellers compared to waiting a second

I like bare AMM for most trading, but I'd definitely use a feature where the market opens at some point after creation and trades before that are an auction. Just to remove the initial settling period, and maybe give people time to clarify rules

@citrinitas Definitely! I think AMM can work with the call auctions too. The way I'm imagining it, people submit bids (effectively, limit orders) and the market resolves all pending bids periodically against each other and also against the AMM.

Amen. See also the research of Michael Wellman and Elaine Wah.