Skip to main content
MANIFOLD
(GPT-4 Sudoku challenge) How many puzzles in the January pool will be solved by any solution?
20
Ṁ865Ṁ3.2k
resolved Jan 17
82%71%
20
18%22%
10
0.3%
0
0.5%
1
7%
31

Main Market: /Mira/will-a-prompt-that-enables-gpt4-to

The January fixed pool should have exactly 31 puzzles. If fewer than 31 puzzles are published, I will make up the difference using easy puzzles from a different source. If more than 31 are published, I will use only the first 31.

The Solve Count is the number of these puzzles solved by any candidate in the GPT-4 Sudoku challenge eligible for extended testing. If one candidate solves 70% of puzzles, and another candidate solves only the other 30%, then this market resolves to 100% since each puzzle was solved by at least one candidate. If testing in the main market is stopped early because resolution is guaranteed, the solve percentage will be extrapolated to the remainder.

Resolves to the linear interpolation between nearest bounding answers of the Solve Count.

  • If no puzzles are solved, resolves to 100% 0

  • If 10 puzzles are solved, resolves to 100% 10

  • If 17 puzzles are solved and (10, 20) are the nearest bounds, resolves to 30% 10 and 70% 20.

Market context
Get
Ṁ1,000
to start trading!

🏅 Top traders

#TraderTotal profit
1Ṁ102
2Ṁ70
3Ṁ53
4Ṁ20
5Ṁ19
Sort by:

The final solve percentage was 10/17 or 59%. 10/17 * 31 = 18.2 puzzles extrapolated over the remainder. So it resolves 82% 20 and 18% 10.

@Mira I was trading on this under the assumption that all 31 days would be tested.
I know with respect to the main market you said you would resolve the other markets as soon as the resolution criteria were guaranteed but this one stated that it resolves to how many out of 31, no more than 31, are solved, and makes no mention of this market's resolution being based upon a partial set of the 31.

never mind I found it:

If testing in the main market is stopped early because resolution is guaranteed, the solve percentage will be extrapolated to the remainder.

I need to do a better job reading. lol 🤣

@ShitakiIntaki That was edited in on January 1, when the possibility of partial resolution was brought up. I didn't want to harm bettors that bet based on a 31-puzzle pool by changing terms, so I added the extrapolation clause to give them a similar EV for their shares. Your bets were placed on Jan 6 and later.

@Mira is there only one candidate testing for the main market?

So the resolution will be something along the lines of ?

@ShitakiIntaki Emily's December 1st submission is the only one being tested.

@Mira Would your prompt count for this market?

@EmilyThomas I forgot about that. I think I shouldn't accept it here, because it wouldn't be a "candidate" for the main contest. This market was also intended for "live betting" which is impossible if there's a hidden solution that could pop out at any time.