4 ms·

I'm not sure I follow. Isn't one supposed to select a scale which allows the compiled data to be clearly and informatively expressed? And doesn't the data compi

I'm not sure I follow. Isn't one supposed to select a scale which allows the compiled data to be clearly and informatively expressed? And doesn't the data compiled so far fit comfortably on a linear scale? Would you recommend using a logarithmic graph because you expect Bitcoin use to increase exponentially? In that case, wouldn't the use of a logarithmic scale at this point imply a direct belief in the future of Bitcoin use?

mpyne 12y ago
> Would you recommend using a logarithmic graph because you expect Bitcoin use to increase exponentially? In that case, wouldn't the use of a logarithmic scale at this point imply a direct belief in the future of Bitcoin use? This is pretty much it exactly, and why the whole "OMG log graphs!!!" thing pops up in finance only when you're talking about Bitcoin and its alts. Though even then I can't figure out why, you only would want log graphs for exponential growth or decay, which isn't something we're seeing out of Bitcoin except for a brief period of tulipmania.
- ars 12y ago
  No, that's not correct in the slightest. > pops up in finance only when you're talking about Bitcoin If you are looking at linear stock value graphs you are being mislead. If you want to make correct decisions based on graphs they must be logarithmic (except percent graphs). Any stock trader that uses linear graphs will lose his shirt. (And I am well aware that the default stock graphs are all linear.) > you only would want log graphs for exponential growth or decay No, that's not correct (for the common meaning of exponential meaning "huge"). You want a logarithmic graph anytime the future value of something depends on the present value (i.e. the mathematical meaning). Which is most of the time. Most things in the world, the amount you have today influences how much you will have tomorrow. It's actually rare to want a linear graph. You only want that when how much you have tomorrow has nothing to do with today. So population should be logarithmic (the more people you have the more will be born), but infant mortality per capita should be linear (future levels of infant mortality do not depend on how many died in the past).
  - logicallee 12y ago
    ars, I think the position you're taking is a bit strong. The author expressed the relationship between transaction volume and price, to reach a certain daily transaction amount: this implies he's perfectly aware that the relationship isn't linear (a move to twice the transactions at half the price is the same movement, irrespective of the price you're asking at - even though it looks like a larger swing if the basis is larger). But the fact that the shape doesn't show this as nicely as the shape you'd like does, doesn't mean that we can't follow the author's reasoning. Why not drop him a line by email explaining that his post would benefit from logarithmic graphs, since htey show the relationships he's discussing better.
    - ars 12y ago
      > Why not drop him a line by email I wanted to, but the site has no place for comments and no info on how to reach him. His conclusion at the end is that the total dollar volume is approximately flat, and he bases this on the graph, but it's not completely flat - it's bumpy, so you can't actually tell on a linear graph if it's flat or not.
  - kszx 12y ago
    > Any stock trader that uses linear graphs will lose his shirt. Can you show me a stock trading platform that uses log values by default? Yes, log graphs are very useful when you're looking at developments of stock values and many other financial variables - predominantly over longer terms when the changes are high enough that the difference matters.
ars 12y ago
> Isn't one supposed to select a scale which allows the compiled data to be clearly and informatively expressed? And doesn't the data compiled so far fit comfortably on a linear scale? It has nothing to do with scale. It has to do with change. The change in value for bitcoin is what we care about, we want to see how it changes over time. > Would you recommend using a logarithmic graph because you expect Bitcoin use to increase exponentially No, that's not why at all. It's because if the value goes from $10 to $20 - that's exactly the same as from $20 to $40. A logarithmic graph shows that correctly, a linear graph makes it looks like 20 to 40 is twice as large as 10 to 20, but it's not.

mpyne 12y ago

> Would you recommend using a logarithmic graph because you expect Bitcoin use to increase exponentially? In that case, wouldn't the use of a logarithmic scale at this point imply a direct belief in the future of Bitcoin use? This is pretty much it exactly, and why the whole "OMG log graphs!!!" thing pops up in finance only when you're talking about Bitcoin and its alts. Though even then I can't figure out why, you only would want log graphs for exponential growth or decay, which isn't something we're seeing out of Bitcoin except for a brief period of tulipmania.

ars 12y ago
No, that's not correct in the slightest. > pops up in finance only when you're talking about Bitcoin If you are looking at linear stock value graphs you are being mislead. If you want to make correct decisions based on graphs they must be logarithmic (except percent graphs). Any stock trader that uses linear graphs will lose his shirt. (And I am well aware that the default stock graphs are all linear.) > you only would want log graphs for exponential growth or decay No, that's not correct (for the common meaning of exponential meaning "huge"). You want a logarithmic graph anytime the future value of something depends on the present value (i.e. the mathematical meaning). Which is most of the time. Most things in the world, the amount you have today influences how much you will have tomorrow. It's actually rare to want a linear graph. You only want that when how much you have tomorrow has nothing to do with today. So population should be logarithmic (the more people you have the more will be born), but infant mortality per capita should be linear (future levels of infant mortality do not depend on how many died in the past).
- logicallee 12y ago
  ars, I think the position you're taking is a bit strong. The author expressed the relationship between transaction volume and price, to reach a certain daily transaction amount: this implies he's perfectly aware that the relationship isn't linear (a move to twice the transactions at half the price is the same movement, irrespective of the price you're asking at - even though it looks like a larger swing if the basis is larger). But the fact that the shape doesn't show this as nicely as the shape you'd like does, doesn't mean that we can't follow the author's reasoning. Why not drop him a line by email explaining that his post would benefit from logarithmic graphs, since htey show the relationships he's discussing better.
  - ars 12y ago
    > Why not drop him a line by email I wanted to, but the site has no place for comments and no info on how to reach him. His conclusion at the end is that the total dollar volume is approximately flat, and he bases this on the graph, but it's not completely flat - it's bumpy, so you can't actually tell on a linear graph if it's flat or not.
- kszx 12y ago
  > Any stock trader that uses linear graphs will lose his shirt. Can you show me a stock trading platform that uses log values by default? Yes, log graphs are very useful when you're looking at developments of stock values and many other financial variables - predominantly over longer terms when the changes are high enough that the difference matters.

ars 12y ago

> Isn't one supposed to select a scale which allows the compiled data to be clearly and informatively expressed? And doesn't the data compiled so far fit comfortably on a linear scale? It has nothing to do with scale. It has to do with change. The change in value for bitcoin is what we care about, we want to see how it changes over time. > Would you recommend using a logarithmic graph because you expect Bitcoin use to increase exponentially No, that's not why at all. It's because if the value goes from $10 to $20 - that's exactly the same as from $20 to $40. A logarithmic graph shows that correctly, a linear graph makes it looks like 20 to 40 is twice as large as 10 to 20, but it's not.