Wall St. Computers Read and Trade on the News | Hacker News Reader

Wall St. Computers Read and Trade on the News

nytimes.com

29 pointsJ3L240418 comments

I appreciate that the Times calls this what it is: an arms race, and it got me wondering, has there ever been an arms race that has benefited humanity?

Competition over tangible consumer technology, I understand. An arms race (what this is, because most would call finance markets a zero-sum game) I do not understand.

This sort of algorithmic black-box HFT with uncontrolled news feed input is like mutual assured destruction, all over again. What is it going to take to give the financial industry a sense of responsibility?

Well, there are nuances here. HFT may not be a net win for humanity, but it's a long leap from there to "no arms race benefits humanity".

The last large "arms race" was a decades-long contest between the US and the USSR. It resulted in millions of deaths (via proxy wars and famine), the near-bankrupting of several contestants, and several cases where we almost turned our civilization into glowing dust.

It also got us, as a species, to the Moon. It encouraged hundreds of thousands more people to study engineering and hard sciences, created the interstate highway system, modern communications, telemetry, trauma medicine, epidemiology, etc.

I won't get into a discussion of whether all that was worth the cost, but it's no coincidence that technology advances quickly during times of war and near-war.

It probably makes the market efficient even at a resolution of sub-second intervals. But what societal benefits accrue from that is not clear to me.

because most would call finance markets a zero-sum game

Most what would call what finance markets zero-sum games? Futures are, the stock market isn't, regardless of what most call them.

This really isn't news, Reuters et al have been selling tools like ClearForest for years and lots of quant funds have their own tools.

This reminds me of the recent New Yorker article debating the social value of Wall St.:

http://www.newyorker.com/reporting/2010/11/29/101129fa_fact_...

Isn't this what caused a tizzy in the markets a few months ago?

Some news article came out that trigged some of the algos to sell...then the other algos started seeing this and selling...then the news reports started flowing in about something weird going on...the algos thought the sky was falling and started freaking out until somebody hit the scram button?

Ultimately what caused the tizzy in the markets was the mid-office of financial institutions not being rigorous enough. Not enough backtesting. Inadequate modeling. Weak correlations.

Analyzing news, per se, isn't an issue. There are already plenty of data brokers that provide this information (i.e. Thomson Reuters' Guidance, Significant Developments, and many more). Decreasing the latency also isn't an issue. Models that enter runaway states are the issue. Quantitative finance is about rigor -- eliminate the rigor and it's advantages over qualitative strategies are all but gone.

Can we get a link without the sign-in?

you can google the headline

Googling the headline gives me a link with the same sign-in barrier.

http://viewtext.org/article?url=http%3A%2F%2Fwww.nytimes.com...

This has been done for a LONG time...

Yeah, you can find mentions of this sort of thing as a standard motivating application for information-extraction/sentiment-analysis/etc. NLP systems dating back to the 1980s. The 80s academic papers I've read do mostly propose it as a semi-automated thing: extract a bunch of information and present it to a human trader, i.e. function as a decision-support system. Not sure when it was first plugged into a closed loop that directly traded on the analysis without human intervention (it's possible it's not publicly known who first did that, since hedge funds tend not to announce such things).

Generally that's true of most algo/AI stuff, initially it goes in as human advisory and then gets switched over to automatic once people trust it.

^H^H^H^H^H^once the volume compensate the bad decisions.

there. fixed it for you.

Many of the people who wrote those papers went on to work at rentec. Rented, by dollars, is the most successful hedge fund ever.

this is exploitable both ways. humans will always be able to outsmart themselves.

Reply on news.ycombinator.com