5 ms·

Benford's Law is a pretty established method of detecting fraud in forensic accounting.

by Closi 5y ago

darawk 5y ago
Benford's law does not apply here. Benford's law is designed to separate human generated numbers from truly random ones. The numbers generated by exchange trades are generated by humans choosing trade sizes. This entire methodology is a complete joke.
srdev 5y ago
Benford's Law requires that the numbers being analyzed typically occur over a wide distribution spanning orders of magnitude, which typically doesn't happen in trading (except over very long time periods).
- Closi 5y ago
  Why would regulated and non-regulated exchanges differ in whether they follow Benford's law or not then? (As in the article)
dustingetz 5y ago
You don't even need to detect it. Name a single instance in human history where an unregulated and unaccountable industry didn't instantly degenerate into fraud. "2.5T" "dollars" (majority held by early insiders) is an incredible moral hazard
- solveit 5y ago
  "I'm sure there is a lot of fraud" is very different from "This is the type of fraud that is going on and here is the evidence".
  - bluecalm 5y ago
    The difference matters for legal proceedings but not for much else, surely not for personal investment decisions.
gillesjacobs 5y ago
Benford's law is extremely dubious in the field. [1] "Abstract. Is Benford's law a good instrument to detect fraud in reports of statistical and scientific data? For a valid test, the probability of ‘false positives’ and ‘false negatives’ has to be low. However, it is very doubtful whether the Benford distribution is an appropriate tool to discriminate between manipulated and non-manipulated estimates. Further research should focus more on the validity of the test and test results should be interpreted more carefully." 1. https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1468-0475.2010.00510.x https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1468-0475....
- ChrisLomont 5y ago
  That is not representative of "the field". Your paper, from 2010, has 21 papers citing it on the link you provided. None support the paper conclusions. The highest cited one of those, from 2019, ([1] with 172 citations) uses Benford's Law exactly as it's commonly stated and used. Here [2] is Google Scholar on Benford's Law. Of the 36,000 papers it pulled very, very few claim the law is not useful and valid. The vast majority (actually, every one of the first many pages) show how it's useful and demonstrate uses of it. If you want only recent papers, select from the left panel. Same result. In fact, the first several pages of results contain many papers showing empirical validation of the usefulness of the law. [1] https://www.degruyter.com/document/doi/10.1111/j.1468-0475.2011.00542.x/html https://www.degruyter.com/document/doi/10.1111/j.1468-0475.2... [2] https://scholar.google.com/scholar?hl=en&as_sdt=0%2C14&q=benford+law&btnG= https://scholar.google.com/scholar?hl=en&as_sdt=0%2C14&q=ben...
ur-whale 5y ago
> Benford's Law is a pretty established method of detecting fraud in forensic accounting Benford law never had to deal with exchanges where any customer can write a python bot and start trading at sub-second latencies via the exchange API's. I'd be very surprised if whatever statistical model they're relying on is in any way a match for what real (as in: legitimate orders from actual customers) trading goes on on crypto exchanges.
- MrMan 5y ago
  Python ... Subsecond latencies Ok
  - ur-whale 5y ago
    >Python ... Subsecond latencies Lol, fair enough. But I believe my point stands even if you go to python-level latencies.
  - wombatpm 5y ago
    Bitcoin transactions with sub second latency Double ok
    - tommek4077 5y ago
      You dont trade on the blockchain. We build software that does multi hundred millions orders per day. Millions of trades. But of course, not in python.
      - tata71 5y ago
        In Go, lol
  - FooBarBizBazz 5y ago
    "Subsecond" isn't much of a bar. A second is an eternity.
nostrademons 5y ago
Only if you assume "automated" == "fraud", which is the distinction that grandparent is getting at. Imagine you have a completely manual market which follows normal statistical distributions, including Benford's Law. Now you introduce one grid-trading bot with a large amount of capital. (A grid trading bot is basically a piece of software that automatically buys when the price falls below a certain level and then automatically sells when it goes above a certain range.) That one bot is going to make up the vast majority of transactions, because it's effectively "clamping" the price within its trading range. When random fluctuations take it below, it buys and sets a floor on the price. When random fluctuations take it above, it sells and sets a ceiling on the price. If you make the range small enough that most ordinary trades would end up occurring with the bot, it's going to take up the vast majority of volume. There's nothing illegal about grid trading. They work to dampen random price fluctuations in a market. In exchange, they take on the risk that fundamental supply & demand might shift enough that they're left holding the bag, eg. they run out of inventory to sell and then the price jumps sharply higher, or they collect all the inventory and then the price drops. But because they're non-human and take one side or another of most trades, they are going to account for a disproportionate amount of volume. This isn't fraud, it's that you don't understand the structure of the market.

Benford's Law is a pretty established method of detecting fraud in forensic accounting.

by Closi 5y ago

Benford's Law is a pretty established method of detecting fraud in forensic accounting.

Benford's law does not apply here. Benford's law is designed to separate human generated numbers from truly random ones. The numbers generated by exchange trades are generated by humans choosing trade sizes. This entire methodology is a complete joke.

srdev 5y ago

Benford's Law requires that the numbers being analyzed typically occur over a wide distribution spanning orders of magnitude, which typically doesn't happen in trading (except over very long time periods).

Closi 5y ago
Why would regulated and non-regulated exchanges differ in whether they follow Benford's law or not then? (As in the article)

dustingetz 5y ago

You don't even need to detect it. Name a single instance in human history where an unregulated and unaccountable industry didn't instantly degenerate into fraud. "2.5T" "dollars" (majority held by early insiders) is an incredible moral hazard

solveit 5y ago
"I'm sure there is a lot of fraud" is very different from "This is the type of fraud that is going on and here is the evidence".
- bluecalm 5y ago
  The difference matters for legal proceedings but not for much else, surely not for personal investment decisions.

gillesjacobs 5y ago

Benford's law is extremely dubious in the field. [1] "Abstract. Is Benford's law a good instrument to detect fraud in reports of statistical and scientific data? For a valid test, the probability of ‘false positives’ and ‘false negatives’ has to be low. However, it is very doubtful whether the Benford distribution is an appropriate tool to discriminate between manipulated and non-manipulated estimates. Further research should focus more on the validity of the test and test results should be interpreted more carefully." 1. https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1468-0475.2010.00510.x https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1468-0475....

ChrisLomont 5y ago
That is not representative of "the field". Your paper, from 2010, has 21 papers citing it on the link you provided. None support the paper conclusions. The highest cited one of those, from 2019, ([1] with 172 citations) uses Benford's Law exactly as it's commonly stated and used. Here [2] is Google Scholar on Benford's Law. Of the 36,000 papers it pulled very, very few claim the law is not useful and valid. The vast majority (actually, every one of the first many pages) show how it's useful and demonstrate uses of it. If you want only recent papers, select from the left panel. Same result. In fact, the first several pages of results contain many papers showing empirical validation of the usefulness of the law. [1] https://www.degruyter.com/document/doi/10.1111/j.1468-0475.2011.00542.x/html https://www.degruyter.com/document/doi/10.1111/j.1468-0475.2... [2] https://scholar.google.com/scholar?hl=en&as_sdt=0%2C14&q=benford+law&btnG= https://scholar.google.com/scholar?hl=en&as_sdt=0%2C14&q=ben...

ur-whale 5y ago

> Benford's Law is a pretty established method of detecting fraud in forensic accounting Benford law never had to deal with exchanges where any customer can write a python bot and start trading at sub-second latencies via the exchange API's. I'd be very surprised if whatever statistical model they're relying on is in any way a match for what real (as in: legitimate orders from actual customers) trading goes on on crypto exchanges.

MrMan 5y ago
Python ... Subsecond latencies Ok
- ur-whale 5y ago
  >Python ... Subsecond latencies Lol, fair enough. But I believe my point stands even if you go to python-level latencies.
- wombatpm 5y ago
  Bitcoin transactions with sub second latency Double ok
  - tommek4077 5y ago
    You dont trade on the blockchain. We build software that does multi hundred millions orders per day. Millions of trades. But of course, not in python.
    - tata71 5y ago
      In Go, lol
- FooBarBizBazz 5y ago
  "Subsecond" isn't much of a bar. A second is an eternity.

nostrademons 5y ago

Only if you assume "automated" == "fraud", which is the distinction that grandparent is getting at. Imagine you have a completely manual market which follows normal statistical distributions, including Benford's Law. Now you introduce one grid-trading bot with a large amount of capital. (A grid trading bot is basically a piece of software that automatically buys when the price falls below a certain level and then automatically sells when it goes above a certain range.) That one bot is going to make up the vast majority of transactions, because it's effectively "clamping" the price within its trading range. When random fluctuations take it below, it buys and sets a floor on the price. When random fluctuations take it above, it sells and sets a ceiling on the price. If you make the range small enough that most ordinary trades would end up occurring with the bot, it's going to take up the vast majority of volume. There's nothing illegal about grid trading. They work to dampen random price fluctuations in a market. In exchange, they take on the risk that fundamental supply & demand might shift enough that they're left holding the bag, eg. they run out of inventory to sell and then the price jumps sharply higher, or they collect all the inventory and then the price drops. But because they're non-human and take one side or another of most trades, they are going to account for a disproportionate amount of volume. This isn't fraud, it's that you don't understand the structure of the market.