3 ms·

Discussing the merits of HFT is similar to discussing the merits of different approaches to cryptography. Because there's such a vast amount of prerequisite kn

Discussing the merits of HFT is similar to discussing the merits of different approaches to cryptography. Because there's such a vast amount of prerequisite knowledge required for analysis, one either has to slog through the learning curve or look for expert opinions. So who are these experts in HFT that should be referenced? Certainly not me or Michael Lewis. One source I like to follow is Eric Hunsader. One of his claims to fame is catching a leak in FOMC data that led to a policy change by the Federal Reserve [1]. My own take on the HFT debate: HFT is here to stay; the method of trading is not the problem, the danger is that no regulator can monitor the massive stream of quotes and trades and accurately sift out bad actors. What's the solution -- I have no clue, but arbitrarily slowing down trades is not useful. [1]http://www.nanex.net/aqck2/4473.html http://www.nanex.net/aqck2/4473.html Edit: Apparently no true Scotsman would follow Nanex as a source. Duly noted. :) I'd be interested to hear of other sources that at least attempt in depth analysis like Nanex.

gd1 11y ago
>So who are these experts in HFT that should be referenced? Certainly not me or Michael Lewis. >One source I like to follow is Eric Hunsader. Well done. I nearly spat my coffee. Are you trying to be funny? I can't tell.
- Nrsolis 11y ago
  Yeah....most people in the know think Hunsader is full of crap. He certainly sounds smart to people outside of the business of HFT, but he's not quite the droid everyone seems to think he is. His opinions of HFT and the various market microstructures that evolve are quite hilarious to anyone with a trading background.
evanpw 11y ago
The second order problem is how to select which "experts" to believe. In my opinion, Hunsader sees conspiracies where none exist and interprets all unusual or hard-to-explain behavior as evidence for his existing beliefs. (Not to say he's never right, of course). But you have no reason to believe me, some random guy on the internet, and in general most experts on the subject are going to have some kind of conflict of interest.
- tptacek 11y ago
  I've gotten the same read about Hunsader and Nanex from other experts: good data, and interesting but questionable analysis.
patio11 11y ago
Because there's such a vast amount of prerequisite knowledge required for analysis, one either has to slog through the learning curve or look for expert opinions. Thomas, Erin, and I possess voluminous experimental evidence that suggests generalist engineers can know more about market making than Michael Lewis (and, if one assumes Lewis is not committing malpractice with regards to framing quotes, Brad Katsuyama) in under two hours. This is not a commandingly high bar. An order book is just a data structure. It tends to have a lot of money associated with it, but that is also true of at least some hash tables, and you understand hash tables. I am frustrated with HN discussion of HFT occasionally for the same reason I get frustrated with HN discussions of some privacy technologies: there exists a political valence to the discussion and some participants treat that political valence as carte blanche for just ignoring that we are talking about actual codebases which have actual rules which can be described in words, including words which definitively falsify some claims about them which feel right to people due to political valence.
- drewblaisdell 11y ago
  > Thomas, Erin, and I possess voluminous experimental evidence that suggests generalist engineers can know more about market making than Michael Lewis (and, if one assumes Lewis is not committing malpractice with regards to framing quotes, Brad Katsuyama) in under two hours. After which level in Stockfighter is it likely that I know more about HFT than Lewis?
  - patio11 11y ago
    Depending on how exactly you decide to solve it and how good of a mental model you have about what your code is doing, level three. Again, a high bar Lewis' level of understanding is not. I know it feels like this is an abrasively combatative statement from me, and it's quite uncharacteristic about any topic other than Bitcoin, but if you've coded a trading system and then read Flash Boys it's like several hundred pages of "I used a GUI to query optimize the traceroute with MongoDB and a red/black tree." Those words, individually they have meaning, but in that configuration it is resoundingly unclear that the speaker understands what is going on. It is resoundingly unclear that Lewis understands what an order book is, what order types are (other than (paraphrase) a mechanism by which sophisticated operators cheat mom-and-pop hedge funds out of their hard-earned 2 and 20), etc. If you'd like this critique at literally book length, read Flash Boys: Not So Fast. It's brutal. It's a point-by-point and page-by-page refutation which brings in lots of crunchy detail (which Lewis scrupulously avoids), quotes extensively from experts (including ones who would be incentivized to say the opposite thing except for a respect for the truth), comports with the understanding of our informal advisors, and does not require me to suspend belief in core principles of math, physics, or computer science, which Flash Boys does multiple times.

gd1 11y ago

>So who are these experts in HFT that should be referenced? Certainly not me or Michael Lewis. >One source I like to follow is Eric Hunsader. Well done. I nearly spat my coffee. Are you trying to be funny? I can't tell.

Nrsolis 11y ago
Yeah....most people in the know think Hunsader is full of crap. He certainly sounds smart to people outside of the business of HFT, but he's not quite the droid everyone seems to think he is. His opinions of HFT and the various market microstructures that evolve are quite hilarious to anyone with a trading background.

evanpw 11y ago

The second order problem is how to select which "experts" to believe. In my opinion, Hunsader sees conspiracies where none exist and interprets all unusual or hard-to-explain behavior as evidence for his existing beliefs. (Not to say he's never right, of course). But you have no reason to believe me, some random guy on the internet, and in general most experts on the subject are going to have some kind of conflict of interest.

tptacek 11y ago
I've gotten the same read about Hunsader and Nanex from other experts: good data, and interesting but questionable analysis.

patio11 11y ago

Because there's such a vast amount of prerequisite knowledge required for analysis, one either has to slog through the learning curve or look for expert opinions. Thomas, Erin, and I possess voluminous experimental evidence that suggests generalist engineers can know more about market making than Michael Lewis (and, if one assumes Lewis is not committing malpractice with regards to framing quotes, Brad Katsuyama) in under two hours. This is not a commandingly high bar. An order book is just a data structure. It tends to have a lot of money associated with it, but that is also true of at least some hash tables, and you understand hash tables. I am frustrated with HN discussion of HFT occasionally for the same reason I get frustrated with HN discussions of some privacy technologies: there exists a political valence to the discussion and some participants treat that political valence as carte blanche for just ignoring that we are talking about actual codebases which have actual rules which can be described in words, including words which definitively falsify some claims about them which feel right to people due to political valence.

drewblaisdell 11y ago
> Thomas, Erin, and I possess voluminous experimental evidence that suggests generalist engineers can know more about market making than Michael Lewis (and, if one assumes Lewis is not committing malpractice with regards to framing quotes, Brad Katsuyama) in under two hours. After which level in Stockfighter is it likely that I know more about HFT than Lewis?
- patio11 11y ago
  Depending on how exactly you decide to solve it and how good of a mental model you have about what your code is doing, level three. Again, a high bar Lewis' level of understanding is not. I know it feels like this is an abrasively combatative statement from me, and it's quite uncharacteristic about any topic other than Bitcoin, but if you've coded a trading system and then read Flash Boys it's like several hundred pages of "I used a GUI to query optimize the traceroute with MongoDB and a red/black tree." Those words, individually they have meaning, but in that configuration it is resoundingly unclear that the speaker understands what is going on. It is resoundingly unclear that Lewis understands what an order book is, what order types are (other than (paraphrase) a mechanism by which sophisticated operators cheat mom-and-pop hedge funds out of their hard-earned 2 and 20), etc. If you'd like this critique at literally book length, read Flash Boys: Not So Fast. It's brutal. It's a point-by-point and page-by-page refutation which brings in lots of crunchy detail (which Lewis scrupulously avoids), quotes extensively from experts (including ones who would be incentivized to say the opposite thing except for a respect for the truth), comports with the understanding of our informal advisors, and does not require me to suspend belief in core principles of math, physics, or computer science, which Flash Boys does multiple times.