Will Manifold limit markets to 1-99% in 2023?
resolved Jan 1

To limit markets to 1-99%, the AMM would only allow bets up to these percents. Additional NO bets at 1% would be canceled or converted into a NO limit order at 1%.


  • At 1% most of the skill is at guessing resolution risk and RoI and this is not as fun as trying to predict the world.

  • Avoid problems with prediction market reliability at extremes.

  • Can display conspiracy theories as "<1%" instead of "0.5%", which I think is a better look (eg, "manifools say 0.6% chance moon landings faked").

  • Reduces profits available from rugpulls.

  • Reduces profits available from world manipulation to make previously unlikely events happen.

  • Slightly improves liquidity by spreading the same subsidy over a smaller range.

  • More resilient to tranche corrections.

Resolution criteria:

"Markets" in the title refers to binary (yes/no) markets, only. It does not include numeric, multiple response, free response, etc, which may have different needs.

It refers only to the AMM probability, not to limit orders.

It refers only to bets through the UI, not to the API.

The limit must be two-sided, eg a 10-100% limit does not count.

In the event of a partial implementation that effects only some markets, I may resolve to a probability between 1-99 that represents my assessment of how partial it is.

Get Ṁ600 play money

🏅 Top traders

#NameTotal profit
Sort by:

True now!

It's extremely difficult now!

I've updated resolution criteria, please comment if you feel this clarification is unfair. As written, the market will resolve NO if no code changes are made.

predicted YES

This would be a reasonable set of changes imo:

  • Put AMM liquidity from 1-99 instead of 0-100 (This is probably the weakest point here, the other points are more important imo.)

  • Allow limit orders at non-integer percentages when low/high (roughly, you want the same amount of precision in log-odds terms when trading at 50% or 2%). This also means people can set limit orders at 99.9%.

    • James has said he agrees with this: "Ok, I'm pretty convinced. It's a little annoying to implement b/c we have to support decimals in the input, but should be doable. Not sure when we'll get to it though"

  • Change the main display show >99% instead of 99.8% (this is a display change only, doesn't change the market mechanics). Since there will be no AMM at <1% or >99% you won't have an exact price, you'll have a bid-ask spread. (Not that the exact AMM price is very meaningful anyway because you typically get massive price slippage at extreme prices, which is effectively equivalent to wide spreads.) You could show the midprice, but that would be misleading usually.

predicted NO

@jack wow this seems good! So basically it just means the AMM is scraped dry when you hit 1% or 99% but you can still bet further? That's great from my POV; I don't want extreme cases for AMM value anyway.

So price would just say <1% or >99%, there is no AMM in that region, but you can still have limit order battles, and the next price is just your opposing limit? I assume you can still use the default yes/no buttons, but they'd just match against limits (if going down below 1%) and count as instant fills. And if there are no matching buys to your sub 1% short attempt, either nothing happens or you have the option to add a NO limit?

Wonderful. If trade history displays these fills meaningfully with buyer, seller, and price, it would be perfect - a big improvement for me. I gain limit improvements and much better precision and history and lose basically nothing. Thanks for putting this together.

EDIT: and the "actual" price with multiple sig digits beyond single percentages would be visible within the limit orders list, if any, or if the user tries to short when there are no YES limits so no matching bets

predicted YES

@Ernie Yes, exactly.

@jack this makes the scoring rule improper, right?

Put AMM liquidity from 1-99 instead of 0-100 (This is probably the weakest point here, the other points are more important imo.)

predicted YES

@EvanDaniel Not sure what you mean exactly (because scoring rule isn't a concept in these markets), but if your question is whether the payouts have the correct incentives, nothing changes there. Buying shares at x% still costs you x/100 per share.

The change proposed here is only to liquidity (i.e. who is willing to trade with you), not to the payout mechanism.

Uniswapv3 is a specific example of an AMM that allows you to put liquidity in specified ranges instead of 0-100.


Not sure what you mean exactly (because scoring rule isn't a concept in these markets)

Any AMM that guarantees to lose exactly a fixed amount for each resolution (path independence) is equivalent to a scoring rule. Right? Or am I misremembering Hanson's original paper?

predicted YES

@EvanDaniel Sure, I guess. In that case the storing rule is still proper.


Put AMM liquidity from 1-99 instead of 0-100 (This is probably the weakest point here, the other points are more important imo.)

Have you done the math to see what effect this has on market liquidity between 1-99? Judging from how much liquidity is available from 0-1%, it seems like it would be significant. This market has m940 subsidies and m523 is available by betting no. So it sure looks like most of the subsidy is wasted on 0-1% and 99-100%.

(never mind I'm bad at math right now)

predicted YES

@MartinRandall Yeah, I think it can definitely be significant, I think the math is complicated because it's going to vary depending on what p the subsidy was injected at. It should be less if the market spends most of its life near 50%.

predicted NO

Edit: the actual proposal is pretty great and seems to actually improve the range of things users can do so I no longer think this comment is relevant but leaving for history.

This feels a little weird. Because group A is convinced something is not possible they're trying to prevent group B from being allowed to do it, meanwhile people in group B are saying hey wait what?

I do appreciate the moderation and resolution risk points, though. If this really reduced those costs I can see the value. I've been assuming long term mm will need scalable community solutions for those regardless

Trying to manipulate PR by hiding variant opinions doesn't seem right. (The bit about being able to say "<1%".)

Feels off to me that the absolute freedom in market creation, description and rule adding freedom, would be paired with an arbitrary price restriction. Since when does mm think it knows better than users?

predicted YES

@Ernie I disagree with this framing - imo it's not about "preventing people from doing X". It's more like "let's modify the AMM to distribute liquidity from x-y instead of 0-100%." Manifold is providing an exchange, and many exchanges have tick sizes that only allow trading at integer percentages, so this wouldn't be particularly novel.

predicted NO

@jack if most agree with your premise that markets don't work at extreme values then the case seems clear.

For me, given that we're hobbled from testing the claim by info hiding in price+limit order power, based on my exp an manifold I personally am not convinced yet. But if most others do believe that, seems like the proposal will be accepted.

To get my vote, though, I'd like to see a real test and demo of extreme markets not working even with working tools, a reason for why 1% and not another number, a long term plan for how trading is supposed to work here with hobbled limit orders (I believe the distortion in the cost and info environment that happens at extreme values is damaging to the market because our tick size is too large. IE extreme markets are dominated by limit order walls where more central value markets aren't, because the relative price change by a percent tick becomes very large, and info sharing via prices in trade history becomes weaker), and product wise how this restriction of freedom fits with the otherwise very free set of site rules. So, a fair estimate of cost vs benefits including complexity of adding this restriction. I acknowledge some benefits to the idea in some ways, but don't feel the costs from my pov are being weighted or accepted as legit.

And regardless I will work on my extension which shows full prices since it's clearly valuable to me at all price levels.

predicted NO

Why do you propose 1%? That's decimal bias. If we had 16 fingers you'd be proposing 1/256th.

The natural line is the maximum: 0% and 100% asymptote

Run a virtual first past the post vote across all evolutionary histories of intelligence, each species voting their preference and zero wins. 1% is not special

@Ernie I already addressed this in my initial proposal, the general proposal is not fixated on a specific percentage threshold. @MartinRandall just chose a specific threshold for this question:

@Ernie When I create games I create them for ten fingered species. What do you do?

predicted NO

@MartinRandall yes numerically 1% makes sense for human understanding. But I haven't seen evidence that 1% is better than 1/64th or 1/256s as the place markets lose usefulness

Manifold Love markets are limited to 10%+. Please discuss if this should resolve the market YES or to some non-zero probability. The market is closed pending discussion to reduce drama.

I wrote in the comments that

A 2-98% limit resolves this yes, because they prevent betting beyond 2% and thus beyond 1%

So the 10% limit is sufficient at the low end.

@MartinRandall I had recently looked at this and interpreted your criteria as requiring a limit to all markets, even if it could be a limit besides 1-99

predicted YES

@MartinRandall I think the common interpretation would probably be requiring a limit on most markets? I don't think all is required, e.g. if the limit was applied to all yes/no markets but not to multi choice, that seems like it would count. But having the limit on a small subset of markets seems insufficient for YES.

I suppose another possibility is resolving to the % of markets that have the limitation.

@MartinRandall If Manifold had added a new market type with the limit, leaving all other markets alone, would that have been sufficient to resolve this YES? That's effectively what's happened here. That doesn't feel like a limit placed on markets, that feels like a new option we didn't have before.

I suppose another possibility is resolving to the % of markets that have the limitation.

I don't think this makes sense; when are you doing the check? The most logical point would be right after the limit is implemented in the code or right after the first market with the limit is created, but that means this just resolves to 0 anyway.

predicted YES

@IsaacKing Fair point, but I think the most logical point would be to resolve to the maximum at any point in the year. Anyway, I think just resolving yes/no based on whether most markets have a limit is the most sensible approach.

@IsaacKing @jack I think when I made this market we only had binary markets and perhaps the numerical markets that are reskinned binary markets, and the question justification is clearly focused on binary markets. But the Manifold Love markets are "unlinked multiple response" which are basically four binary markets in a trench coat. So I think they still count in spirit and in letter. Also the justifications given for the Love 10% limit overlap with the justifications in this question.

Points against:

  • Only a small subset of markets, not all or most

  • Only a one-sided limit (10-100%), not two-sided.

There's also this in the comments:

Q: How will this resolve if Manifold only disables quick bet arrows for extreme probabilities, but allows regular bets and limit orders?

A: That's a great feature but doesn't change market resolution. The market is not limited.

But I would accept a UI limit that can be bypassed via the API. Grudgingly.

I'm currently leaning towards a 1% resolution (assuming no change over the rest of 2023).

If Manifold had added a new market type with the limit, leaving all other markets alone, would that have been sufficient to resolve this YES?

I think that should at least count for a partial resolution, given that the market criteria doesn't specify all markets vs most markets vs some markets vs any markets.