Hacker News new | past | comments | ask | show | jobs | submit login
Ask HN: Automated trading on April 1
83 points by dman on April 1, 2011 | hide | past | favorite | 29 comments
Would love to hear from folks who build automated trading systems based on Semantic analysis about how they handle the goofy stories on April 1.


Serious HFT players are either relying solely on news feeds that won't include April Fools stories or are factor in the date for sources that might report them (like CNN). All of the April Fools stories from legitimate news providers have also clearly explained these were jokes and didn't focus a piece primarily on one company.

A shame that there aren't any responses (yet) from people with relevant projects.

Is it possible that those programmers merely take the easy (although not completely fool proof) route and exclude all data from April 1, and all data that refers to April Fools Day?

Somewhat Related:

Does Anne Hathaway News Drive Berkshire Hathaway's Stock?


Valid. If everyone else is using the same news, doesn't matter if it makes sense or not.

Related: are there any good sites/communities/blogs to discuss HFT that anyone can point to?

Am really interested in this field, anything like a crash course in what it involves would be nice by anyone familiar.

How much cash do you generally start with, and how liquid are you?

Thanks, good links in the blogroll as well, appreciated.

Try NuclearPhynance.com

check out quant.ly

hmm: http://www.downforeveryoneorjustme.com/quant.ly :(

Is it just quant talk though? Isn't there enough of that on NP, Wilmott, and QuantNet?

Or am I silly in assuming quants on those sites work mostly for i-banks, hedge funds, and more established players?

Do partners in HFT firms function on the same level, as say the guy structuring CDOs at Bear Stearns? I thought there was a difference, but this is another reason for the wide questions; the nature of what HFTs do is a little vague and not clear from Google searches.

Do partners in HFT firms function on the same level, as say the guy structuring CDOs at Bear Stearns?

Quant can mean a lot of things, just like "programmer" can. Your question is like asking, "do programmers work at Google, or do they write HR reporting systems for no-name companies?" They can do either and they'll still be called programmers.

Ah, I see. I guess I'm just trying to flesh out the difference between all the terms I hear thrown around all the time that seem to overlap; day-trader, quant, HFT partner, etc. I see it's not as clear-cut as I thought. Thanks.

Yes, those all overlap. I work for a department in an investment bank called "Quantitative Analytics and Research". We basically write software to move data from one system to another. The "quant" in there is just to sound important, not to actually mean anything.

In theory, "quants" develop models and "traders" use those models (combined with market data) to make trading decisions. But plenty of people called traders make their own models and plenty of people called quants press the button to make the trade.

(HFT, as I've seen it here, is basically people looking at graphs. But we are not known for electronic trading or anything.)

Awesome, thanks for the great response. Curious if you're familiar with the common trend in the last half-decade~decade or so of Masters in Financial Engineering degree programs popping up.

Would you say spending the 150K+ and years needed to do well on the 7City exam to work as an "official quant" at famous investment banks is reasonable? There seems to be little hiring/demand for quants coming out of these programs, despite the as-of-now still high admission and graduation rates.

If the job description is as variable as you say it is, would it not make more sense to work on becoming a trader at a number of different banks, firms, and prop shops and not waste all the time and money on getting an MFE? It was popular pre-housing bubble and apparently job placement for graduates was high, yet I'm not sure this is the case is now. Could you comment on this, seeing as you're working there already?

No idea. I'm a programmer who happens to work at an investment bank, but finance does not interest me any more or any less than any other programming problem. The words we use to describe things are finance-related, but the problems we solve aren't.

I would personally never pay $150k for a degree, but that's because the programming world doesn't care about degrees at all. I have never tried to be a quant so I don't really know. My guess is that if you come in as a programmer on a HFT team and interest the right people, you can be a quant without any particular credentials.

I'm told by my boss that many of the people in our application support team are "blown up" traders. Which means they were traders, and now they help people reset passwords, leading me to believe that there aren't many expensive degrees required to become a trader. That's all I know, though.

The 7City quant course fee is around U$20k, not 150k!

I think before the 2008 crisis (I was going to say "Lehman shock" but I'm not sure if this is common outside Japan) it was considered a good deal: you could easily increase your value by more than $20k/year, especially if you were already working in finance. Now, I don't know. Although the surviving banks/funds started re-hiring, I don't think there's as much demand for quants now as there was in (say) 2007.

Ah yes, that is true. The 150K figure is of the usual (ballpark) total cost of a Masters in Financial Engineering program.

quant.ly is down temporarily because of the .ly domain issues with LibyanSpider but it'll be back up as soon as that resolves itself.

The thing about building automated trading systems (related or not to Semantic analysis) is that the things you tend to track are compared to other things you are tracking. After all, if there is an issue with a data stream you wouldn't want to bet the farm on it. You check with other streams. So goofy stories on april first wouldn't affect the algorithms too much unless they were reported the same way in disparate news/feed sources.

How many goofy stories have you seen today that affect publicly traded companies? I don't think I have seen any yet.

Google launching a new product. Atlassian entering into gaming. TSA mandating self pat down system. Oracles Ellison accused of running Executive fighting ring. Facebook and Zynga team up to merge Romance and gaming. Toshiba develops 3-D monocle. Nvidia to unlock SLI on AMD 990 series motherboards. NOTE: the Nvidia and Ellison stories might not be April fools jokes, theres no way to tell until the silliness subsides. Edit: Google hiring auto-completers. Could affect employment related numbers.

  date = today();
  if (date.month() == APRIL && date.day() == 1) {

Read as "Automated Trolling."

"Trading" is discussed on Hacker News at about a 1.5-to-1 ratio with "trolling". FYI, Google Fight accepts the same "site:" specifier that Google Search does. For example:


If you have reading problems, you could use that technique to determine which of 2 words you're more likely to be reading here.

Cool to see that the second time you submitted the question you got some answers.

It wasnt the second time. Since it was an Ask HN, people continued voting for it from what I guess is the ask tab on the main page. At some point it picked up ~15 votes and broke on through to the main page. Also, if you look at the posts so far, none of them answers the question.

  > It wasn't the second time. Since it was an Ask HN,
  > people continued voting for it from what I guess is
  > the ask tab on the main page. 
Ah - nice to know that happens sometimes. I guess you must have deleted your rather wistful comment lamenting that no one had replied.

  > Also, if you look at the posts so far, none of them
  > answers the question.
I wonder if there's no one here who uses semantic analysis in automated trading systems. Maybe the semantic analysis people who are here felt excluded by your question.

Who knows. Perhaps there's a better question to ask, or a better way of hooking the people you want into answering.

But I'm pleased you got traction, even if you haven't actually got answers.

Indeed I did delete my comment :). I have been interested in the white spy / black spy aspect of semantic trading for sometime and the trust related issues. However I have no expertise in the area and no one in my network works in the area. Folks doing technology in finance usually document their work much lesser than other fields, perhaps because of the financial incentives to keep secrets. So I thought tapping the wisdom of HN was my best bet to scratch my curiosity scratch.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact