Why this is important

High stakes makes Manifold fun. There would be no Tumbles Financial Complex without markets where @Tumbles can bet at size (e.g. the US and Canadian presidential markets). Manifold can't provide enough liquidity to make every market fun, so we need to rely on users to bet against each other.

The problem

Leaving open limit orders on prediction markets is dangerous: news could break at any moment, letting traders fill your limit order placed with old, now out-of-date information. This is true in stock markets as well, but stocks are much less likely to go to 0 (or 100% if you’re holding no).

Possible solutions

Timed limit orders (capped fill amount per hour)

• How it works: You set a max % (or mana amount) fill per hour (or other minute interval). If the price moves beyond your limit price, your limit order fills at the minimum of your fill threshold and order amount every hour until your limit order is cancelled or fills completely.

• Pros: This lets you set a maximum amount of the order you lose due to news breaking. You might get a better price for shares if traders move the AMM above your price. User experience is relatively easy to get right and implementation is relatively easy.

• Cons: Your order is less likely to fill completely as opposing traders trade more against AMM liquidity, which is capped at 1 and 99%. You can still lose all of your order if you don't see the limit fill notifications in time.

Limit offers (require creator agreement before filling)

• How it works: You can create a new type of limit order that requires your consent after placement and before it fills. A trader will see a table of limit offers and click a button to request to fill yours. You get a notification and have an hour (or other minute threshold) to agree, otherwise your offer is cancelled.

• Pros: This completely protects your limit order, you can check if there are any relevant news before agreeing. This allows trading to scale to any size.

• Cons: We'd need a new way to create and expose these limit offers, which sounds like a challenging user experience problem. Many limit offers may go unfilled (even without relevant news bc a user is away from their device), disappointing traders. Implementation is non-trivial.

• Credit: @Ernie's suggestion doc

Conditional limit orders (order fills only if AI with web search thinks there is no recent, relevant news)

• How it works: You can toggle AI protection on your limit order, which will use a fast AI model (likely gemini flash 2.0 with web search) to check if there are any news articles related to the market since you made the order that might influence your decision, and only fill if no news appears relevant. Maybe you could write a custom prompt as well. The AI might have to be queried every minute or so to make sure trading isn't slowed, or maybe we would just slow trading down on that market if a trade would get a significantly better price at >1000mana and query the ai at the time of the trade.

• Pros: Off-loads your news-searching work onto an AI and protects your orders from news if none has been published yet. Should be pretty smooth user experience.

• Cons: The AI might make a mistake when evaluating if news is relevant, or news might be so recent that the AI can't find it (though you might not be able to find it, either). If we slow down trading to let the AI search for news (although we could cache the result and run the search at a max of every minute), this would be a worse experience for traders. Perhaps traders could choose to ignore these orders?

Have any other ideas or comments?

Related Poll: /ian/why-dont-you-place-more-limit-order

Related market: /ian/percent-limit-offers-limit-orders-i