4 ms·

One of the best pieces of advice I heard from our resident stats/machine learning researcher at Uni was this: If a paper is published which claims to show a tec

One of the best pieces of advice I heard from our resident stats/machine learning researcher at Uni was this: If a paper is published which claims to show a technique for profiting on the share market, then it doesn't work. If it did, they wouldn't have published it. Having said that, there is plenty of interesting papers in this area, and from the abstract, this paper seems to take an interesting approach too.

theintern 12y ago
My first thought too. This must already be a variable in high frequency trading software for a number of years.
jevgeni 12y ago
I think it was Michael Lewis' "Liars Poker": "Those that tell don't know and those that know don't tell."
chillingeffect 12y ago
Hmmmmmm if I apply that to that researcher's own words, then he is lying to confuse you! It's a Goedelian paradox! Actually, I think the bigger point is that researcher's rarely "tell all." Just like Moriarty in the final episode of Season 2 of Sherlock (on BBC 2013-2014), sometimes a work is created as an advertisement demonstrating the ability to draw larger returns. Those intimate with the data will always have more than than will or can possibly publish.
ganeumann 12y ago
Interesting academic work on this recently. It shows that, as expected, the publication of a working investment strategy decreases its efficacy. http://news.morningstar.com/articlenet/article.aspx?id=648576 http://news.morningstar.com/articlenet/article.aspx?id=64857... Of course, contra your ML researcher, this presupposes that there are effective strategies that are published.
mathattack 12y ago
Some of it is about the agenda of the person writing it. The Fama-French 3 factor model still holds. (http://en.wikipedia.org/wiki/Fama%E2%80%93French_three-factor_model http://en.wikipedia.org/wiki/Fama%E2%80%93French_three-facto...) Part of this is because Gene Fama doesn't have a hedge fund on the side. He advises DFA, but they want their customers to understand the model. (He also won this year's Economics Nobel)

theintern 12y ago

My first thought too. This must already be a variable in high frequency trading software for a number of years.

jevgeni 12y ago

I think it was Michael Lewis' "Liars Poker": "Those that tell don't know and those that know don't tell."

chillingeffect 12y ago

Hmmmmmm if I apply that to that researcher's own words, then he is lying to confuse you! It's a Goedelian paradox! Actually, I think the bigger point is that researcher's rarely "tell all." Just like Moriarty in the final episode of Season 2 of Sherlock (on BBC 2013-2014), sometimes a work is created as an advertisement demonstrating the ability to draw larger returns. Those intimate with the data will always have more than than will or can possibly publish.

ganeumann 12y ago

Interesting academic work on this recently. It shows that, as expected, the publication of a working investment strategy decreases its efficacy. http://news.morningstar.com/articlenet/article.aspx?id=648576 http://news.morningstar.com/articlenet/article.aspx?id=64857... Of course, contra your ML researcher, this presupposes that there are effective strategies that are published.

mathattack 12y ago

Some of it is about the agenda of the person writing it. The Fama-French 3 factor model still holds. (http://en.wikipedia.org/wiki/Fama%E2%80%93French_three-factor_model http://en.wikipedia.org/wiki/Fama%E2%80%93French_three-facto...) Part of this is because Gene Fama doesn't have a hedge fund on the side. He advises DFA, but they want their customers to understand the model. (He also won this year's Economics Nobel)