3 ms·

http://www.nanex.net/aqck2/4555.html http://www.nanex.net/aqck2/4555.html summarizes a study here: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1955965 h

http://www.nanex.net/aqck2/4555.html http://www.nanex.net/aqck2/4555.html summarizes a study here: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1955965 https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1955965 Key quote: "This manipulative practice [spoofing] is illegal under the U.S. securities law, yet it has been frequently discovered in both equity and futures markets." Here's an example of the penalty a middling HFT firm sees: http://www.reuters.com/article/us-sec-enforcement-spoofing-idUSBREA331DD20140404?feedType=RSS&feedName=businessNews http://www.reuters.com/article/us-sec-enforcement-spoofing-i... $1.9M in fines - no jail. By contrast Sarao is facing up 380 years in prison. Others faced fines of only several hundred thousand dollars: http://www.wsj.com/articles/how-spoofing-traders-dupe-markets-1424662202 http://www.wsj.com/articles/how-spoofing-traders-dupe-market... Note that the most connected HFT firms (Virtu, Citadel - oh hey, Ben "Revolving Door" Bernanke works there) don't even get fines, despite engaging in spoofing, and even being banned from China for spoofing: https://www.ft.com/content/d4b61458-398d-11e5-8613-07d16aad2152 https://www.ft.com/content/d4b61458-398d-11e5-8613-07d16aad2... Also note that there have been many "flash crashes" triggered by spoofing HFT's. There's nothing special about the Sarao one, other than the fact that he was an individual operating out of his home.

evanpw 10y ago
That paper uses simulated (i.e., fake) market data. The quote is just a bare assertion, not a conclusion. It's very difficult to identify spoofed orders in market data unless you're the exchange and know who sent all of the orders.
- ConfuciusSay02 10y ago
  The paper uses "fake" data to develop a baseline for their model, which they then compare to real data to test its accuracy, which they claim is 90%. Furthermore, the study is based on the premise that there is lots of spoofing going on. That's because EVERYONE knows that spoofing is rampant. Everyone has acknowledged it, including the SEC and the Fed, so to dismiss the study by simply saying "it used fake data" with no other context is not a forthright appraisal of the situation.
  - evanpw 10y ago
    > the study is based on the premise that there is lots of spoofing going on That's fine, but you can't then claim the study as evidence of spoofing, since it was injected into the simulations explicitly. > which they then compare to real data to test its accuracy, which they claim is 90% The validation with real data was just comparing things like trade volume, volatility, autocorrelation, etc. They didn't put a specific number on it (the 90% accuracy mentioned in the abstract is about detecting HFT in simulated data). And they added in the spoofing after the validation. There definitely is spoofing going on (some of my job involves trying to detect it), but I'm deeply skeptical of anyone who claims they can detect it accurately or quantify the amount. 92% accuracy for detecting spoofers in simulated data is insanely high, and makes me doubt the applicability to the real world.

evanpw 10y ago

That paper uses simulated (i.e., fake) market data. The quote is just a bare assertion, not a conclusion. It's very difficult to identify spoofed orders in market data unless you're the exchange and know who sent all of the orders.

ConfuciusSay02 10y ago
The paper uses "fake" data to develop a baseline for their model, which they then compare to real data to test its accuracy, which they claim is 90%. Furthermore, the study is based on the premise that there is lots of spoofing going on. That's because EVERYONE knows that spoofing is rampant. Everyone has acknowledged it, including the SEC and the Fed, so to dismiss the study by simply saying "it used fake data" with no other context is not a forthright appraisal of the situation.
- evanpw 10y ago
  > the study is based on the premise that there is lots of spoofing going on That's fine, but you can't then claim the study as evidence of spoofing, since it was injected into the simulations explicitly. > which they then compare to real data to test its accuracy, which they claim is 90% The validation with real data was just comparing things like trade volume, volatility, autocorrelation, etc. They didn't put a specific number on it (the 90% accuracy mentioned in the abstract is about detecting HFT in simulated data). And they added in the spoofing after the validation. There definitely is spoofing going on (some of my job involves trying to detect it), but I'm deeply skeptical of anyone who claims they can detect it accurately or quantify the amount. 92% accuracy for detecting spoofers in simulated data is insanely high, and makes me doubt the applicability to the real world.