« Assorted links | Main | How an insurance mandate could leave many worse off »
How to flip a coin
Chris Blattman reports:
Using a high-speed camera that photographed people flipping coins, the three researchers determined that a coin is more likely to land facing the same side on which it started. If tails is facing up when the coin is perched on your thumb, it is more likely to land tails up.
How much more likely? At least 51 percent of the time, the researchers claim, and possibly as much as 55 percent to 60 percent — depending on the flipping motion of the individual.
The original research is here.
Posted by Tyler Cowen on October 24, 2009 at 11:03 PM in Games, Science | Permalink
Comments
this would make a difference if coin flipping were ever subject to the law of large numbers, but is it? probably never.
Posted by: rob at Oct 24, 2009 11:17:29 PM
I learned this from watching an episode of QI a while back, which led to a brief and unsuccessful career as a high-stakes coin-flipping hustler.
Posted by: Emperor Norton at Oct 24, 2009 11:33:32 PM
The real result from that work is that a massive advantage can be gained if you spin the coin (like a top). Some coins will land "tails" ~80% of the time, because the weight distribution matters.
Posted by: Joey at Oct 25, 2009 12:42:11 AM
Sorry, I just can't believe their conclusion, or rather I can't believe that it would apply in real-world situations.
Posted by: James at Oct 25, 2009 1:36:14 AM
A good candidate for the Ignobel prize.
Posted by: ranon at Oct 25, 2009 2:00:10 AM
Why did they need a high-speed camera for this?
Posted by: Daniel at Oct 25, 2009 3:17:40 AM
"Why did they need a high-speed camera for this?"
That was my initial question, too. I assume that they didn't just want to find out whether there was non-randomness, but also why. And -- if there were individual differences where some people produced more biased results -- what about the way they flipped the coin produced those biased results?
Posted by: Slocum at Oct 25, 2009 7:13:30 AM
nope.
Posted by: josh at Oct 25, 2009 7:42:59 AM
This seems like an application of Benford's Law in base 2.
Posted by: Cyrus at Oct 25, 2009 7:54:01 AM
You should have quoted from the conclusion on page 27 of the original paper:
"If we can find this much trouble analyzing a common coin toss, the reader
can imagine the difficulty we have with interpreting typical stochastic assumptions in an
econometric analysis."
Bias depends on technique. This can be shown with simple math. Assume the technique is to toss the coin so that it flips a very few times. Assume the distribution of flips is:
1 flip = 50%
2 flips = 25%
3 flips = 15%
4 flips = 10%
This gives a 65% chance that the coin will land on its opposite side. It's a rough model, but you get the drift. Learning a technique like that would be much easier than learning to, say, split a deck evenly and shuffle it so that the deck returns to the same order after 6 shuffles. Which lots of people know how to do.
Posted by: Bob Knaus at Oct 25, 2009 8:02:44 AM
what about that stats for people who flip real estate? how much of that real estate ended up landing on its tail? and why would anyone flip a coin? is it possible it could be worth more coming down than going up? is this because of deflation?
Posted by: babar at Oct 25, 2009 8:20:02 AM
Glad to see they nailed this down. I noticed it when I was in 2nd grade and have been waiting to be scientifically validated...
Posted by: df at Oct 25, 2009 3:45:08 PM
From a 'james franklin' science of conjecture perspective, this could be a discovery of great merit, amongst others bah, bah, bah
Posted by: -- at Oct 25, 2009 4:08:39 PM
sorry . . . emperor norton answered that in the second comment.
Posted by: -- at Oct 25, 2009 4:30:37 PM
This is actually quite believable result, as flipping a coin requires an impulse applied to it, which increases height of flight and spinning speed at the same time, if more powerful. And those two can cancel each other out to some extent when counting how many revolutions a coin made before landing.
In reality probability bias should depend on actual person doing flipping, as initial movement of hand matters a lot, as well as catching height :)
Posted by: Konstantin at Oct 26, 2009 5:52:18 AM
This is pretty old news (it was published in 2007). Here is the published paper: http://www-stat.stanford.edu/~cgates/PERSI/papers/dyn_coin_07.pdf
Posted by: Andy at Oct 26, 2009 10:25:27 AM
Of course, good etiquette for coin flipping has somebody other than the flipper call heads or tails. Even without this research, I would not have gambled quarters against somebody who insisted on flipping and calling the choice.
Posted by: Parke at Oct 26, 2009 4:25:45 PM
The obverse (principal side) of a coin typically features a symbol intended to be evocative of stately power, such as the head of a monarch or well-known state representative.
Posted by: estetik at Oct 27, 2009 4:56:35 AM
