Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Using Benford’s Law to Detect Bitcoin Manipulation

statmodeling.stat.columbia.edu

Post reply on HN
Print view View original

51–60 of 96 posts

Re: Using Benford’s Law to Detect Bitcoin Manipulation

#51

Post by x4e »

fluidcruft wrote: post #21
qPM9l3XJrF wrote: post #10

I don't buy it. My intuition here is that Benford's law is a thing because for pretty much any statistical distribution, large numbers are less likely than small numbers, and small numbers are more likely to start with 1. But I'd only expect this effect to show up when aggregating across many different statistical distributions (e.g. looking at all stocks in the S&P 500 at once) rather than looking at individual dist…

I don't really know anything about this but every time I read these analyses using Benford's Law I don't understand why anyone would expect Bitcoin's exchange rate to begin with the digit "1" ~30% of the time. Once you're not talking about human-guestimated numbers, it seems more like a question of scale factors. If bitcoin's value fluctuates between 30k and 70k it's never going to start with a "1". Or if its value f…

I believe it’s about which numbers it ends with, not starts with

Re: Using Benford’s Law to Detect Bitcoin Manipulation

#52

Post by iratewizard »

mrfredward wrote: post #32

Benfords law is used to find evidence that the numbers came from a person, not a measurement or mathematical process, right? So anyone who knows what a limit order is should not be surprised to find evidence that humans are involved in picking the prices, right? It should be obvious that violating Benfords law isn't evidence of fraud or manipulation or even fomo, just evidence that the price is impacted by the people…

This is exactly it. If you look at price changes as bitcoin approaches round numbers, you can see that a significant number of people have their limit price set to something like $10,000. When it would approach those round numbers, it would be stuck just under that number for a while. If the price cracked the round number, meaning all those limits got sold, the price would then slingshot much higher.

Re: Using Benford’s Law to Detect Bitcoin Manipulation

#53

Post by contravariant »

dannyw wrote: post #7

What is the statistical likelihood of this appearing by chance? Note that 2014 to today isn't actually that long of a timeframe, as prices in a time-series are highly linked to the previous data point. Berkshire Hathaway has been trading for several decades; so that's 40 years of data; as compared to 7 years of data. yet the author, by using the same charts, seems to falsely imply that these are remotely comparable.…

Benford's law doesn't give much of a guarantee except in highly specific scenarios. You can however show it's pretty accurate if you've got a 'smooth' probability distribution spanning multiple orders of magnitude (how much of a guarantee you can give depends on how specific you define what it means to be 'smooth').

The rigged elections example you gave was, I believe, mostly explained by the fact that the number of votes they were looking at were all roughly the same order of magnitude [1].

[1]: https://www.youtube.com/watch?v=etx0k1nLn78

Re: Using Benford’s Law to Detect Bitcoin Manipulation

#54

Post by SamPatt »

louwrentius wrote: post #2

Frankly, I can’t judge the merits of this article because I lack the knowledge required. But I think at this point we know that all cryptocurrencies are ‘greater fool’ “investments”. [1] They are in every way totally irrelevant and detrimental to society. Governments are unfortunately slow to crack down on the exchanges, although progress is being made. And that effort is essential to battle the cancer that is the ra…

"at this point we all know..."

Sorry that's an appeal to majority.

You're wrong. Bitcoin is useful.

Re: Using Benford’s Law to Detect Bitcoin Manipulation

#56

Post by y04nn »

fluidcruft wrote: post #21
qPM9l3XJrF wrote: post #10

I don't buy it. My intuition here is that Benford's law is a thing because for pretty much any statistical distribution, large numbers are less likely than small numbers, and small numbers are more likely to start with 1. But I'd only expect this effect to show up when aggregating across many different statistical distributions (e.g. looking at all stocks in the S&P 500 at once) rather than looking at individual dist…

I don't really know anything about this but every time I read these analyses using Benford's Law I don't understand why anyone would expect Bitcoin's exchange rate to begin with the digit "1" ~30% of the time. Once you're not talking about human-guestimated numbers, it seems more like a question of scale factors. If bitcoin's value fluctuates between 30k and 70k it's never going to start with a "1". Or if its value f…

Yes, I don't know if the prices should have a better Benford's Law fit when the price is around and 10k or 100k but then a lesser fit when the price is between 5/50k and 9k/99k if you have not enough data. But if there is enough data, maybe it should better fit the Benford's Law, and this may be a proof that arround 10k/20k the price was rigged and should have stay around it longer but manipulation put it around 30k/50k.

Re: Using Benford’s Law to Detect Bitcoin Manipulation

#57

Post by JumpCrisscross »

mrfredward wrote: post #47
JumpCrisscross wrote: post #41

Earlier quoted context omitted.

> If the author had spent 5 seconds thinking about how markets work The author has spent a career thinking about this, and has written a good fraction of the textbooks on statistics in market contexts.

I guess we need to make a distinction between the blog post and the Gary Smith post it links to here. Gary smith (the person I think you're referring to having spent a career in this) says this: >The market manipulation, the irrational price gyrations, and the enthusiasm of so many investors for investing in bitcoin (and other cryptocurrencies) is ample evidence that market prices are not invariably equal to intrinsi…

> the world is full of people who feel very different paying $100.00 vs $99.99

Agree, though that effect is not constrained to Bitcoin. Retail orders, for instance, follow Benford's law. This is despite well-documented psychological biases towards e.g. certain digits, whole numbers, round numbers, et cetera [1]. Benford's law [2] derives from deeper mechanics.

As you point out, however, a better control would have been not all prices in public stocks, but retail orders.

[1] https://mro.massey.ac.nz/bitstream/handle/10179/2695/02_whol...

[2] https://en.wikipedia.org/wiki/Benford's_law#Krieger–Kafri_en...

Re: Using Benford’s Law to Detect Bitcoin Manipulation

#58

Post by axiosgunnar »

SV_BubbleTime wrote: post #49
nootropicat wrote: post #28

Earlier quoted context omitted.

LN isn't actually used - the way it actually works is: Strike holds dollars for the user. The user wants to send dollars to someone. Strike buys btc using dollars (from itself), sends btc via ln - to itself - and then sells btc for the same amount of dollars, again to itself, and credits the receiver. What's actually happening is that Strike is a normal payment provider like Paypal, Venmo, Revolut that fakes the btc…

I read that twice and now it makes less sense. What is the purpose of the fake transfer steps?

> because pretending to be only a btc wallet means less regulation than a full payment provider.

Re: Using Benford’s Law to Detect Bitcoin Manipulation

#59

Post by tedunangst »

dannyw wrote: post #7

What is the statistical likelihood of this appearing by chance? Note that 2014 to today isn't actually that long of a timeframe, as prices in a time-series are highly linked to the previous data point. Berkshire Hathaway has been trading for several decades; so that's 40 years of data; as compared to 7 years of data. yet the author, by using the same charts, seems to falsely imply that these are remotely comparable.…

Like what does the graph look like for TSLA? Or GOOG?

Re: Using Benford’s Law to Detect Bitcoin Manipulation

#60

Post by TomSwirly »

dannyw wrote: post #7

What is the statistical likelihood of this appearing by chance? Note that 2014 to today isn't actually that long of a timeframe, as prices in a time-series are highly linked to the previous data point. Berkshire Hathaway has been trading for several decades; so that's 40 years of data; as compared to 7 years of data. yet the author, by using the same charts, seems to falsely imply that these are remotely comparable.…

> prices in a time-series are highly linked to the previous data point.

And how many data points are there in seven years of BTC?

> seems to falsely imply that these are remotely comparable.

Why is it "false" that they are "remotely comparable" if we are looking at time series of security prices?

You are long on the insults and short on the reasoning. Given the author is a well-known statistician and you are some anonymous individual, I'm going with them.

Post reply on HN