Unexpected Rationalussy Paradox: If EY first tweets rationalussy in the last 5 days of 2023, will the day be a surprise?
25
470
470
resolved Jan 2
Resolved
NO

The Rationalussy Wiki describes the Unexpected Rationalussy Paradox as follows:

Eliezer Yudkowsky tells you that he is going to tweet (or xeet) the word "rationalussy" in 2023, but that the day will be a surprise. More specifically, you will not be able to know what day it is going to occur on until he actually tweets it.

You conclude from this that he will not tweet it for the first time on Dec. 31, since then the day would not be a surprise: You would know as soon as 12/31 midnight comes around that he would be tweeting it on Dec. 31, since he had not tweeted it on any other day, and it is the last day of the year to do it. You therefore buy lots of NO shares in the corresponding market (though not too many - after all, he could tweet "rationalussy" multiple times, with only one of them being on the last day).

However, you then realize that it can't happen on Dec. 30 either. After all, if it hasn't happened by the end of Dec. 29, then you would know on midnight, Dec. 30 that it was going to happen that day, since you've already eliminated Dec. 31 as a possibility. And the same reasoning can be applied to Dec. 29, Dec. 28, and so on, all the way down to the current date. You conclude that what Yudkowsky said is impossible. There is no way he can tweet the word "rationalussy" in 2023 while keeping the day a surprise. So much for being a rationalist! You go about the rest of your year feeling smug about having schooled Eliezer Yudkowski in basic logic.

Then, on Dec. 29, 2023, Eliezer Yudkowsky tweets the word "rationalussy", in a total shock to you, and to the market on the subject. Despite your seemingly valid argument, every part of Yudkowski's original announcement was true. He did tweet the word "rationalussy" in 2023, and you had no idea what day it was going to be until he actually made the tweet, despite being told in advance that he would do it on a surprise day.

We now have markets asking whether Eliezer Yudkowsky will tweet "rationalussy" on each of the last five days of 2023:

/PlasmaBallin/will-eliezer-yudkowsky-tweet-the-wo

/nanob0nus/will-eliezer-yudkowsky-tweet-the-wo-58fbde5fbba4

/ButtocksCocktoasten/will-eliezer-yudkowsky-tweet-the-wo-d79032c3984b

/ArmandodiMatteo/will-eliezer-yudkowsky-tweet-the-wo-d7c0423fe232

/ArmandodiMatteo/will-eliezer-yudkowsky-tweet-the-wo-4ceb54c7287c

And of course, there is the main market on whether he will tweet it in 2023 at all:

/SG/will-eliezer-yudkowsky-write-a-twee

This market resolves N/A if Eliezer Yudkowsky doesn't tweet the word in 2023, or if he tweets it before Dec. 27. Otherwise, it resolves YES if the average conditional probability for the first day that he tweets it on is <50% on the day that he tweets it, and NO otherwise. The conditional probability for each day is just the probability for the market about that day divided by the probability of the main market.

I am using the Central Time Zone to delineate days, so the average will be taken from midnight CT on the day Yudkowski tweets the word to the time that the tweet is actually posted.

Get Ṁ200 play money

🏅 Top traders

#NameTotal profit
1Ṁ143
2Ṁ105
3Ṁ100
4Ṁ38
5Ṁ25
Sort by:
predicted NO

I didn't calculate the exact averages, but the main market was never at a probability 2x greater than the Dec. 31 market or more at any point on Dec. 31 before the tweet was posted. So it is impossible for the average conditional probability to have been <50%. It was more like 100%, as expected, since the two markets became equivalent on midnight CT.

bought Ṁ60 of NO

12 hours for this to possibly resolve YES.

predicted NO

@dph121 Technically speaking, it could still resolve YES if someone keeps the average probabilities of the Dec. 31 market and the main market separated by a factor of 2 tomorrow, despite them being equivalent.

God picked the wrong month for me and E.Y. to be in the same league

This is a great market but sadly, it will most likely just resolve N/A.

bought Ṁ10 of YES

10/10