market subsidizers generally subsidize because they want to learn the probability of something before it happens. Not to pay for a google alert that the event has already happened. I feel like past post bets are a waste of the liquidity subsidy except in rare cases where the news is premature

Alternatively we have volatility halts like the stock exchanges. We could gradually reinvent the market structure of successful stock exchanges from first principles or we could just copy it whole. If a market moves 10% suddenly, trading is halted for a bit and the creator is notified to give the creator time to resolve it if it needs to be resolved.

Stock exchanges don’t have N/A. They can bust some trades but only obvious errors. Fortunately you can rephrase all conditional markets as conjunction yes/no markets that never N/A, and get rid of N/A option on new markets. Volatility halts plus periodic limit orders mitigate the problem enough without introducing N/A

I really love this idea. Success on Manifold should be based on prediction, not being in the right place at the right time with a big balance.

Biggest downside from my point of view, though, is that some events don't have an obvious time at which they are known, especially if they are something where the first news might be unreliable. So presumably it becomes creator discretion. But this will inevitably upset some people. (Something like the death of Putin could arguably end up in his category.)

But then I suppose a good rule of thumb for the creator (and maybe more than a rule of thumb if you could think of a good way to codify it) would be "have there been any significant bets placed in the opposite direction since the news/rumour came out?". If so, maybe it was a rumour and the early news traders deserve their profits. If not, maybe actually everybody knows it was reliable news and it's fine to roll back.

arb /chrisjbillington/will-manifold-introduce-a-mechanism-a21e5ee4f7d5

One possible issue with this is that if you know your shares will retroactively be worth 0 in a few minutes or hours, you can sell them at any price, cash out, and then the N/A will put you in the red.

There'll be a delay before cash outs are cleared, sth like 24 hours

the best version of this should include partial n/a such that the trades are not fully reversed, leaving some incentive to still (deliver the) news trade.

you can N/A all limit order trades but not the AMM trades for example, and have the AMM have more of the liquidity near the market price than away from it, so that you get the near-market-price liquidity from news trade and the rest is minimal

Definitely some upsides to this! A downside is spoiling the beautiful feature of prediction markets where you can count on being able to watch the prediction market instead of the news. Like you hear about some crazy thing a politician did and you can just check the market and see if it actually matters or not.

Is there a way to get the best of both worlds? Like limit orders are retroactively canceled but not market orders? (Um, I don't think that makes sense, but, yeah.)

PS, @JonathanRay's proposal of volatility halts sounds sensible. But I think it needs to be combined with a broader revamp where markets never close indefinitely like in the status quo. Trading can just be temporarily paused for various reasons.

/chrisjbillington/will-manifold-introduce-a-mechanism-a21e5ee4f7d5

I was going to post this. The same market already exists and with 50 traders too!

ah, and here I did go post it before reading more than one comment. oops

I think it's a really good idea, but it seems like it has all the issues that resolving N/A normally does

With this change, we'd likely allow N/A for the first 24 hours of market creation. And set some rules on withdrawing prize points as cash where there's at least a 24 hour delay.