Home Creators Posts Import Register Favorites Logout
haven't archived this post yet. have a subscription? use the importer!

Content

In case the 40-minute video wasn't enough for you today, I also did a video with Numberphile about a fun conundrum from probability known as Bertrand's paradox. It was originally something I wanted to mention as a side comment in the cube shadows video, but it shaped up quite nicely to just make it a full separate video and have it posted as a sort of soft collaboration.

Linked above is the first part, but if you want more commentary going deeper into the problem, there is a second part: https://youtu.be/pJyKM-7IgAU 

As a side note, one of the things I really appreciate about the patreon model is how I don't really have to wonder whether the time spent making animations for something like this is "worth it".  Lessons are lessons, it doesn't really matter what channel they live on.

Files

Previews only

Bertrand's Paradox (with 3blue1brown) - Numberphile

Featuring Grant Sanderson, creator of 3blue1brown. Extra footage from this interview: https://youtu.be/pJyKM-7IgAU 3blue1brown video on the shadow a cube: https://youtu.be/ltLUadnCyi0 More links & stuff in full description below ↓↓↓ 3blue1brown: https://www.youtube.com/channel/UCYO_jab_esuFRV4b17AJtAw Grant on The Numberphile Podcast: https://youtu.be/A0RH93XvSyU Numberphile is supported by the Mathematical Sciences Research Institute (MSRI): http://bit.ly/MSRINumberphile We are also supported by Science Sandbox, a Simons Foundation initiative dedicated to engaging everyone with the process of science. https://www.simonsfoundation.org/outreach/science-sandbox/ And support from The Akamai Foundation - dedicated to encouraging the next generation of technology innovators and equitable access to STEM education - https://www.akamai.com/company/corporate-responsibility/akamai-foundation NUMBERPHILE Website: http://www.numberphile.com/ Numberphile on Facebook: http://www.facebook.com/numberphile Numberphile tweets: https://twitter.com/numberphile Subscribe: http://bit.ly/Numberphile_Sub Videos by Brady Haran Patreon: http://www.patreon.com/numberphile Numberphile T-Shirts and Merch: https://teespring.com/stores/numberphile Brady's videos subreddit: http://www.reddit.com/r/BradyHaran/ Brady's latest videos across all channels: http://www.bradyharanblog.com/ Sign up for (occasional) emails: http://eepurl.com/YdjL9

Comments

Kevin Iga

What's the wooden pi circle maker? Where can I get one?

3blue1brown

That was actually a custom-made piece from my dad, who enjoys woodworking on wanted to practice a gold leaf inlay on something.

Kevin Iga

Ah. I will just free hand them on the whiteboard like I have always been doing, then. But if you do find a way to mass produce them and sell them on your store, I'd be interested in buying one.

Anonymous

I want to post my solution, but I can't paste an image... Sigh, Patreon....

Poker Chen

Within the context of Bayesian reasoning, your assumptions are free and you declare them together with the proposed answer. The usefulness is more about the process of deduction than the certitude of the final values. I.e., one would declare their beliefs on the possible definitions of a chord, resulting in a probability over the three answers 1/4, 1/3, and 1/2. It makes intuitive sense in the dice rolling analogy, where the more I know Grant (e.g. what table top games he plays), the better I can tell what kind of dice he has actual access to. He may have D100s, but he's unlikely to have stockpiled more than a few of them, whereas if he plays 40K Orcs he's might actually have 100 D6s, etc. So I can choose to believe that a proposed answer 1/100 is going to be a lot less probable than an alternative 1/6.

Poker Chen

Other response. Here is a Desmos visualiser on the probabilities when you define the chord according the video's three ways: [ EDIT: https://www.desmos.com/calculator/hhebl403x8 ] Notice how a uniform probability over your starting choice is distorted by each definition. Here lies the secret. - Option 1 "choose randomly two points on the perimeter" essentially boils down to "choose randomly the subtended angle". One dice-throw equivalent over this angle exerts a cos^-1 distortion. - Option 2 "choose randomly a point in the circle" boils down to "choose some random radius but biased for larger r", whose dice-throw equivalent is tranformed by a Pythagorean relationship. - Option 3 is related to option 2, but no longer biased for larger r. What you've done is to shift the 2D dice-throw from (x,y) to (r,theta).

Anonymous

Hi Grant, great video as always. My take on the paradox is that the naive assumption that the number of point in the larger annulus (in method 2) is greater than the number of points in the smaller central circle. There are the same number of points in each and so the probability of choosing from one set is the same as from the other set. Likewise with method 1, although the length of the circle where the chords would be shorter is twice as long, the number of points in the longer and shorter sections are equal and so the probabilities are equal. The approximation methods, being inherently discrete, create distortions which throw off the estimates.

Kevin McCurdy

I thought of a variation on the 1/3 way, where the distribution is biased by the length of the lines (so corresponding with area). That one gives the area of the 60 degree sector (probability of around .609).

James Matheson

This feels like there should be a connection to that other numberphile video you did about the dart game with the area of the dart board shrinking based on the size of chord generated by a dart hit. Wouldn't the third method correspond to the dart throwing method where you are taking a random point drawing a line from the centre and then taking the chord? In that way the third method does feel somehow less biased than the other methods, in that it can be generated by uniformly blasting the circle with points but the first method still somehow feels right too.... I wonder what happens if instead of two random points you have one rando point on the circumference and then a distance traveled.

John Nichols

This was very fun to watch and listen to. I also found the 1/2 answer to seem the least appropriate when originally looking at the problem. In a house-fire scenario, I would pick the 1/3 probability, as I define a chord as the line which connects two points on the perimeter of a circle. I'm now thinking about the very first question which is what is the mean length of that chord... How would you set that up for the third (1/2) definition of a chord? Could you just integrate the height of the circle over x and divide by x?

Bernard

nice... rather than thinking about translation in-variance, could it also be correct to think in terms of infinite sets of cords? Like should the solution not be the union of all infinite sets of possible cords and thus maximum of the set of probabilities of possible sets of cords? thus 1/2?