Context: Random twitter person (who is followed by ChatGPT official account) claims something big will happen in an hour. https://x.com/iruletheworldmo/status/1819161723250921976
I will resolve according to my subjective assessment of "bigness". I will count both major releases OR announcements of major releases (e.g. GPT-5, real-time voice, etc).
I will not be that strict with timing. If there's a release any time before 7ish pm PT, that will count.
1,000
3.00🏅 Top traders
| # | Name | Total profit |
|---|---|---|
| 1 | Ṁ287 | |
| 2 | Ṁ166 | |
| 3 | Ṁ156 | |
| 4 | Ṁ112 | |
| 5 | Ṁ22 |
I'm very 😍 about this method of calling someone on their BS. Possibly in this case this person deserved to be ignored altogether, idk. In any case, I'm now wondering about how, in general, to ask a question like this where timing is critical. Saying "before 7pm ish still counts" probably mostly suffices but it feels like setting oneself up for a Sorites paradox. Like if this person hadn't been trolling and were just a little off on the timing, how should it have resolved? Is 9pm still "7pm ish"? I think in principle there's a fraction correct that that person would be that starts at 100% at 6:30pm (the literal hour they claimed) and gradually goes to 0% in, idk, a day?
My tentative contention is that we should pick a standard protocol along those lines where you name an exact time when the correctness starts dropping and an exact time the correctness hits zero and say that it decays linearly in between and resolve-to-PROB accordingly.
(I don't know if that fully captures the spirit of it. Like if GPT-5 is launched this weekend maybe that probable-troll seems prescient after all? But maybe it's a start.)
I think it's simpler to just set a specific criteria (generally I would go with "what the person claimed, plus a little bit of wiggle room), and then have it be a binary market. Otherwise I think it's less intuitive to price correctly. The thing we're trying to predict is pretty binary (do they know something or not) and so IMO it's most intuitive to predict it with a binary market.
To clarify, I do mean a binary market and I think traders can still reason about it in binary terms: do I think the thing will happen in that time frame? Just that the answer might end up being "well... kind of" and it makes sense to me for the resolution to reflect that.
Picking an exact cutoff is cleaner in some sense but I noticed Stephen was reluctant to do that in this market and it makes a lot of sense why he was. If you pick a precise deadline and miss it by 3 seconds then it's going to feel really unsatisfying to have to resolve NO when it feels like such a YES in spirit.
I'm actually getting curious to try my protocol now. (But I think a much bigger issue is that most "will XYZ happen by time T" markets really want to be numeric "when will XYZ happen" markets.)