Smart money is moving the markets! @jim and @JonathanRay and @JonasVollmer yes you are doing it right way. @Joshua and the @jack and the @jacksonpolack maybe not so much here? I am always searching for the edge of it here, and everyday your bets are bringing me closer to it. I agree!

We've made a new official YES/NO presidential market! Simpler format, same 100k subsidy!

"I have the concepts of a plan."

If you expect the debate to have a significant impact, but you have mean 0 on the direction of the move, you can buy some variance on the outcome of this very market, in the market below:

(Or if you think little will change between now and the election, you can buy a more sharply peaked distribution there!)

Perfectly balanced

It's over for communistcels

@skibidist now what happened? our country is doomed. 😞

@WesReisen Seems pretty 50/50 now.

I collected all of @ManifoldPolitics markets for each states and ran the election 100_000 times simulating the outcome.

https://johannorberg.github.io/mm_us24el/

It is updated once every hour. Right now it has D: 68% - R: 32% with median outcome D: 282 - R: 256

How did you account for correlations between states?

They should use /EvanDaniel/pairwise-state-results-which-pairs of course! 🤪

I added a simulation using the pairs from the market you linked.

It goes through all pairs https://manifold.markets/EvanDaniel/pairwise-state-results-which-pairs simulates outcome.

I don't think the correlations affect the final result much, As it stands now I have

Normal: D: 68% - R: 32%
Adjusted (taking everything below 25% to 0% or above 75% to 100%) D: 60% - R: 40%
Simulated using correlation pairs: D: 68% - R: 32%

I think the reason is that the pairwise correlation markets significantly underestimate the correlations

How often does the simulation update? I just bet a bunch of the pairwise correlation options up, so that should make the simulation accounting for correlation more accurate.

Every hour

@JohanNorberg im curious about why the result of this simulation is so radically different from what the market says

@metachirality It's possible that, even accounting for the outcomes in the pairwise market, the sampling still doesn't reflect higher order correlations somehow. Based on the code it looks like it currently samples pairs of states in a random order, so even now I think the correlations aren't being fully considered.

What I think is maybe happening is that for the solid states, there aren't any state correlation markets, so those states are still being treated as completely independent of all the other states. But simultaneously, their probabilities are way underestimated.

In most simulations, at least one, and probably more, of the solid states flip, so they have an outsized impact on the result.

@PlasmaBallin I updated so all odds for adjusted simulations so all below 25% to 0% and 75% to 100%.

I also added what states were assigned.

Manifold now have more EV for Republican. Mainly Arizona and Pennsylvania has flipped red. Running Monte Carlo simulations still favors Democrat but I think I need come up with a smarter way to adjust the odds.

EV doesn't directly translate into who's favored to win, since that's the median, but Harris's EV should be higher than her median because of the chance of winning Florida or Texas.