Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Tau is clearer in some domains and pi is better in others. The mental overhead from switching between the two is too much, so just pick one already. And let the rest of us know, so we can go back to having a clear standard and not requiring the mental overhead of switching while studying different authors.

Oh yeah, we went through that 2000 years ago and the winner was pi. I haven't seen any compelling reason to switch yet. (Says the guy who uses "j" for the imaginary constant. :P)



In which domains is pi better? And was there a discussion 2000 years ago that led to pi "winning"? (If so, it'd be really fun to read its reasoning.) [Both of these are honest questions; I'm not a mathematician.]


http://www.thepimanifesto.com/

The Pi Manifesto has a few examples of pi beating tau (statistics, polygons, and complex numbers mainly), while also pointing out how silly and biased some of the tau examples are. However, the argument isn't very convincing from either side.

And as to your second question, I have no idea. I just know the idea has been settled for a long time now and I think this whole debate is needlessly distracting. Not that it isn't fun to watch or think about, but it confuses people who are just trying to learn and use math.


Their argument about the normal distribution is completely off though. There is completely no reason to group the two with the standard deviation. Their biggest argument seems to be that the area of unit circle is exactly pi. Sadly, there is absolutely no need to work with fractions of areas of the unit circle, while if you're working with angles the unit circle is a most natural standardisation. The normal distribution has been a lot clearer to me since I've understood the two should be grouped with the pi.

Their argument about trigonometric functions is completely wrong and obviously so. Trigonometric functions work with angles and it's already shown (and pimanifesto readily admits) that tau shines there.

Their argument about Euler's identity is as inane as the tauists' is.

They don't understand quadratic forms.

I'm guessing this site is sarcastic.


We can start with any domain where torque is a thing. Torque is a particularly bad thing to conflict with.

I've also seen arguments that it's better when dealing with triangles (all angles in a triangle add up to pi radians). I don't think I buy that one, since everyone uses degrees for that anyway.


Try http://www.thepimanifesto.com/ for a πle of cases where pi works out more cleanly, and other good arguments for pi.


> a πle of cases

τché!


> Tau is clearer in some domains and pi is better in others.

I always thought Tau was better because it was easier to understand when teaching it. For using it, it doesn't much matter.

But when you are a kid learning about it, having a clear picture of why specifically that number is very important, and Tau is much more obvious.


After reading both the tau and pi manifestos I get the impression that it's quite an arbitrary choice: either the circumference is fundamental and angles are natural (tau), or the radius is fundamental and areas are natural (pi). It boils down to a trade-off and you can cherry-pick examples where either is "more natural", but in a mathematical sense it doesn't matter a single bit and that's why we should not waste time on this, it doesn't bring anything new to the table.


Tau is clearer in some domains and pi is better in others.

This argument is addressed explicitly in the section "What is really going on here?":

http://tauday.com/tau-manifesto#sec:what_is_really_going_on


Re: 2000 years. pi as standard notation goes back to Euler, though of course Archimedes and others calculated it much earlier.

I haven't heard of tau being considered pre-2001, though it must have happened here and there (I had the same idea independently, for example).


At uni, decades ago, I was told that whenever a "mathematician" appeared up on the telly, it was to present their new idea of definining a new constant equal to 2pi...

Not the same now, of course, with shows like Numb3rs and Big Bang Theory, and a greater public awareness of the importance of numeracy.


The mental overhead from switching between the two is too much...

Really? It's just two constants that are both easy to remember and trivial to convert between.

I don't have a horse in this race. I don't mind having two distinct constants to deal with.


Next time you write some code, try this:

#define ONE 1 #define TWO 2 #define THREE 3 ... etc.

Then use the words in place of integers wherever you need them. Hell, mix and match. And tell me if you are just as productive as you would be sticking with the regular integers you know and love.


Sure. And in return, you can't use constants, but must instead type out 3.141592653... to whatever decimal place you deem important. Nor typedefs. Nor named flags (`FileRead | FileWrite` is just long-hand for `3`). At the end of the day we'll see who's pulling their hair more.

Then we'll go back in a couple weeks, and look at the same code again. Mine's wordy. Yours is cryptic, and requires looking up the use of everything.


That seems like a bad comparison, writing out multiple digits in place of a single digit should ALWAYS be less efficient.


Agree....wouldn't having more language for these things make it MORE efficient? I'm down for a symbol for pi/2 (tau/4)!


I think that wouldn't be a meaningful bottleneck on my productivity. But I could be wrong.


x = ONETHREEZEROEIGHT;


I don't ever remember ever reading anything about how 2000 years ago we had a debate on if pi would be better than tau (if 3.14 is better than 6.28).

goes online to see if he can find answer to question before asking... Nope, sorry.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: