Public Online Coin Toss Decider
August 22, 2017 6:52 AM   Subscribe

Is there a good way I can settle a 50/50 decision between two third parties at a distance, so that both will be sure that the decision was random?

I'm thinking something like a NewCoinTossEveryDay Twitter account, where I could get both parties to agree that we'd decide it based on tomorrow's coin.

Sadly that doesn't seem to exist.

Some other things in the right area might be a list of coin tosses at sporting events (as long as there was also as list of upcoming events that would definitely involve a coin toss so we could pick one in advance), or some other 50/50 chance sport event, or maybe whether the last digit of of some stock exchange thing is even or odd at a certain pre-determined point.

Ideally this would be something simple and transparent rather than technical and obscure, but I'm interested in any suggestions.
posted by PJMcPrettypants to Computers & Internet (19 answers total) 2 users marked this as a favorite
 


"PublicRNG.com allows individuals to settle agreements requiring a random element without any counterparty."

(Random.org's numbers aren't public or shared. They still require you to communicate a value, at which point you can lie.)
posted by whatnotever at 7:05 AM on August 22, 2017 [3 favorites]


And while you don't want "technical and obscure," just for completeness: How to perform a fair coin toss experiment over the phone [math.stackexchange].
posted by whatnotever at 7:07 AM on August 22, 2017 [3 favorites]


Stock prices can be good for this sort of thing. For example, check the last digit of the FTSE100 opening or closing price that day. Odd = heads, even or zero = tails.
posted by metaBugs at 7:13 AM on August 22, 2017 [10 favorites]


Video chat of you rolling a dice or flipping a coin?

Do you need to tell them the decision each day? Could you send them a pre-generated list of results like 10100001010111101010010001010 where each day's result is random but known in advance?
posted by EndsOfInvention at 7:15 AM on August 22, 2017


If you go Even vs. Odd: The high or low temperature for a given city or location according to a specific agency.
posted by soelo at 7:16 AM on August 22, 2017 [4 favorites]


There's this Twitter account - Daily Random Numbers (authenticity is unknown however).
posted by EndsOfInvention at 7:17 AM on August 22, 2017 [2 favorites]


Use the last digit of the Dow Jones close from tomorrow. You get odds your friend gets evens.
posted by paulcole at 7:17 AM on August 22, 2017 [2 favorites]


More immediate: Whether the timestamp of [insert verified, prolific Twitter account]'s next tweet will have an odd or even minute. CNN or NYTimes, e.g.
posted by supercres at 7:45 AM on August 22, 2017 [2 favorites]


"Stock prices can be good for this sort of thing. For example, check the last digit of the FTSE100 opening or closing price that day. Odd = heads, even or zero = tails."

Just checked this ticker for last year's data... 53.5% even, 46.5% odd, so this doesn't quite have the 50/50 odds you're looking for. Which, to be fair, neither does flipping a quarter.
posted by Grither at 8:44 AM on August 22, 2017 [1 favorite]


Perhaps to cancel out any bias in a particular stock index's hundredths digit, you could have each party choose a stock ticker symbol, then add the respective closing prices on the desired date and do odds/evens.
posted by letourneau at 8:57 AM on August 22, 2017


You can roll a virtual dice in Google Hangouts by typing /roll dN, where N is a number of sides of a dice. To flip a coin (a two-sided dice), type /roll d2.
posted by ringu0 at 9:41 AM on August 22, 2017 [2 favorites]


Perhaps to cancel out any bias in a particular stock index's hundredths digit, you could have each party choose a stock ticker symbol, then add the respective closing prices on the desired date and do odds/evens.

This is still not perfectly fair; even + even is even (as is odd + odd) and even is more likely than odd. To visualize it easily, think about the case where 99% of the results are even - in that case, adding the digits together would only make a small difference; instead of a 1% chance that one digit is odd, there would be a 2% chance digit that one of the two was odd and the chances of both even would be reduced, but only down to 98%. (A little higher because of the 1 in 10,000 chance of two odds coming up and producing an even number).

If 53.5% of digits are even (and even results are uncorrelated), then adding the two digits together produces a 50.245% chance that both are even. (0.535*0.535 + 0.465*0.465).

You can double up this procedure - pick two pairs of stocks and add - or compare - the digit. (I like to think of comparing the digits to make matching (both odd/both even) or not matching (one odd, one even) - this is mathematically equivalent). That can reduce bias to 50.0012% which is probably good enough.
posted by Homeboy Trouble at 10:07 AM on August 22, 2017 [1 favorite]


Decide who is odd and who is even (Trying it out with a coworker. I'm odd, she's even). Then both people say one word (me: pony, cw: paris) which we typed into the goog without quotes. We agreed that the second most significant digit in number of results will be our coin. Arbitrary? Sure!

So, trying this ten times results in:

pony paris: About 11,800,000 results (0.76 seconds) - odd, I win.
solzhenitsyn puppy: About 27,600 results (0.43 seconds) - odd, I win.
guacamole tincap: About 10,400 results (0.65 seconds) - even, she wins.
kate boy: About 23,300,000 results (0.74 seconds) - odd, I win.
motherfucker season: About 3,160,000 results (0.63 seconds) - odd, I win.
coffee poker: About 6,830,000 results (0.59 seconds) - even, she wins.
slackbot hungry: About 142,000 results (0.67 seconds) - even, she wins.
cigarette taco: About 2,900,000 results (0.59 seconds) - odd, I win.
platypus trainwreck: About 46,000 results (0.69 seconds) - even, she wins.
what carnivorous: About 2,830,000 results (0.48 seconds) - even, she wins.


...gets an even 50/50 split. And some funny results. And Google's impression of me is going to be a little bit more skewed, which is a nice bonus.

Yes, there are flaws to this system. But I don't think it is gameable. Anyway, shut up. It was fun.

Also, I hereby christen this the Pony Paris coinflip.
posted by dirtdirt at 10:47 AM on August 22, 2017 [6 favorites]


What about tomorrow's daily number lottery in *insert city*. One gets even one gets odd.
posted by bowmaniac at 10:48 AM on August 22, 2017 [5 favorites]


dirtdirt, I like the way you think, and your way is the most fun, but I get 25,300,000 for pony paris using your link, so maybe location has an effect on it? I think both parties would have to be in the room for that one.
posted by PJMcPrettypants at 2:42 PM on August 22, 2017


A twitter bot that does the coin toss you want would be trivial to set up. If you don't find a better option and can't set it up yourself, drop me a line and I'll make it.
posted by mishafletch at 6:10 PM on August 22, 2017 [1 favorite]


If you type "flip a coin" into Google search, Google will flip a coin for you and tell you the outcome in the search results page. Works on mobile and desktop in my experience. Also on Android: "Ok Google, flip a coin".

http://google.com/search?q=flip+a+coin
posted by czytm at 5:55 AM on August 23, 2017


I was kindof hoping I could prove the randomness beyond doubt, to make it "trustless", but in a way that seemed easy and casual like I wasn't going out of my way to do so.

I don't want to cause offense by implying that we distrust each other, but I want to avoid the possibility of anyone having creeping doubts later on that I was in cahoots with the other party.

The lottery number thing would have been good except that there isn't a daily lottery where I live, using a foreign lottery seems like going out of my way, and waiting until the weekend would have created cruel and unnecessary suspense.

It's also worth noting that some lotteries have more odd than even numbers.

In the end I went with the Google Hangouts solution.
posted by PJMcPrettypants at 7:38 AM on August 27, 2017


« Older Help me get out of temping and into data analysis?   |   To report or not to report? Newer »
This thread is closed to new comments.