What is quantitative trading?

[electrical]

I am writing this only after searching heavily, both here and on Google. Please excuse the cluelessness - at worst the thread can be thrown in the trashcan.



What do people mean when they say "quantitative trading?" It is something I have heard while looking for my first job, and I have not been able to fully understand what it would mean to work in that line.



My understanding as of now - designing semi-to-all automated (probably should say un-supervised) programs to trade. Skills required are good understanding of DSP, pattern recognition etc.



Is quantitative trading "Algorithmic Trading" with quantitative strategies (as opposed to the ones listed on the Wikipedia article linked above)?

[FDAXHunter]

Finance is full of ill-conceived and ill-defined nomenclature. Different people mean different things when they say something.

My definition of algorithmic trading is different than Wikipedia's for example (that article seems to contain a bit of inaccuracy anyway).

Sometimes this is due to the nature of the beast (for example: finding a definition for the term "derivative" in finance that satisfies all parties is impossible).



Quantitative Trading, in my book is trading based heavily or entirely on hard information (quantified attributes, whatever shape or form they may take).

This need not necessarily be automated, but it excludes any trading of the type "Oh I like this stock, I'll buy that." It need not be unsupervised. Quantitative trading has different subfields like algorithmic trading, systematic trading, assisted trading, etc.



"Hunting" (a way of trading that is my one claim to fame) is quantitative trading. It is not systematic trading however. It's assisted trading



From my other posts, you have likely seen that I put a lot of emphasis on using the correct terminology. I hope this is useful, but do be aware that there are people who will use any of the above terms in a slightly (or completely!) different manner.

[electrical]

FDAXHunter,



Thanks - I have spent many hours reading your other posts on this website.



In the context that I heard the words, I think it had a narrower meaning - a meaning in which "boxes" of various shades (black/grey/white) came in.



Also - any thoughts about the wiseness of starting a career along the above lines? For an electrical engineer



a) whose research has a strong "random-processes" (as opposed to signal processing) bent and



b) who hasn't coded heavily for the duration of his PhD (but has coded quite a bit before that)?



Thanks again.

[FDAXHunter]

The different colors of boxes also have different definitions depending on who you talk to.



To some people, "Black Box" means a systematic trading strategy whose workings are totally unknown. One day the magic box will say: "Buy 1,350 EDU8 at market" and you have no idea why. This is the original use of the term.



Some (fewer) people will use "Black Box" to simply refer to a systematic trading strategy in the sense of "Fully automated, no supervision or intervention required".



Along the first definition of black box, a "Grey Box" is where you know some of the systematic strategy's workings and you may change some of the parameters. A simple example would be some sort of filter where the filter algorithm is proprietary but you may adjust some parameter for frequency or length or whatever.



In the sense of the second notion of "Black Box", "Grey Box" then means "not quite automated trading strategy - some intervention required"



Finally, "White Box" in the first sense means that you know everything about a systematic trading strategy. I.e. the algorithm is known to you.



In the second sense (don't think people actually really use this), "White Box" would imply a trading strategy that requires tons of human decision making, i.e. is not a systematic trading strategy at all.



Note that my White Boxes are Black Boxes to you and vice versa.

[sasquatch]

Feel free to give me a shout regarding possibilities. I highly recommend you brush up on your coding and knock the rust off before you start looking, though. C++ is typically the preferred language, but C (and some others) is okay, too.

[Omega]

what about matlab?

[electrical]

FDAX, thanks again. Is there someplace I can read about "hunting" in the sense that you mentioned it? I searched on Google with the keywords `hunting "assisted trading"' and came up with nothing.

[aaron]

I completely agree with the first paragraph of FDAXHunter's reply, but have a different definition of "quantitative trading." Since people use the term in lots of different ways, I can't claim my definition is correct, but I think it's closer to what most people mean.



FDAXHunter defined it by the inputs, which had to be hard. But I could say, "I buy stocks with 'X' in the ticker symbol." That's a hard input, but there's no quantitative aspect. More reasonably, technical analysis often uses hard inputs but highly subjective non-quantitative reasoning from those inputs. On the other hand, I might assign values to all kinds of subjective inputs like "company culture," "management honesty" and "growth potential." If I worked those inputs into a quantitative model, I'd call that quantitative trading.



So I think quantitative trading can take hard or soft inputs, I define it by the analysis of those inputs. There can be subjective factors added, but to me the key is the subjectivity is restricted to abstract factors, not prices or trading decisions. A true quantitative trader operates in model space, not transaction space.



For example, a quantitative strategy might involve a parametrization of the volatility surface of some underlying. A computer program uses market prices to generate a best fit surface, then selects a low-risk, high-return portfolio of derivatives under the assumption that the best fit surface is in fact correct, both in the sense that it predicts movements in the underlying and that the market will revert toward it. In my definition, a quantitative trader can tweak the fit, and also impose constraints on the recommended portfolio based on ideas about which aspects of the fit are more reliable, and what type of corrections are most likely. But if the trader tweaks the inputs or outputs of the model, she is not a quantitative trader, despite using a quantitative model to support her trades.

[FDAXHunter]

[b]electrical:[/b] [i]I searched on Google with the keywords `hunting "assisted trading"' and came up with nothing.[/i]

That does not suprise me. In fact, if you did then that would mean we have a leak here. It's all very proprietary 'n stuff. Sorry Sad



I'm not in agreement what aaron said.



Saying "I buy stocks with 'X' in the symbol" is indeed very quantitative. It's stupid, and it's not going to make you money (no more or less than approximately random selection), but it's in fact a completely systematic strategy. You just happened to pick an example algorithm that's non-sensical.



The distinguishing factor, in my humble opinion, to Quantitative Trading is the extent of how much subjectivity is included in the analysis.

Aaron mentions Technical Analysis. On the one hand we have completely systematic technical strategies ("Buy the list of the 20 best performers over the last 3 days and sell the list of the 20 worst performers over the last 3 days" which is incidentally, from a systematic point of view no different than buying all the stocks that start with 'B').

On the other hand with have a traditional chartist that, while maybe using some quantitative input such as historical prices, transforms those in a completely unstructured and non-deterministic way.



That is what I meant with "trading based heavily or entirely" on quantified information.



So there is a "Grey Zone" (like the grey boxes mentioned below) where you're not totally systematic, but I'd still say you're quite quantitative.



But like I said: different people, different definitions. You can't argue that much about definitions. It's just important that when you argue, that all parties are using the same definition. Which incidentally, doesn't happen all that often.

[DrTarr]

[b]Aaron >[/b] [i]I define it by the analysis of those inputs. There can be subjective factors added, but to me the key is the subjectivity is restricted to abstract factors, not prices or trading decisions.[/i]



[b]FDAX >[/b] [i]The distinguishing factor, in my humble opinion, to Quantitative Trading is the extent of how much subjectivity is included in the analysis.
[/i]



If we limit or call "hard information;" as information or attributes of or related to a financial instruments price which are quantifiable with a reasonable capacity to predict the future price or its movements. 



Then quantitative trading:  Trading based on the results or determined outcome from a reasonable mathematical analysis of hard information. 



Then break it done into; Algorithmic, Systematic, Assisted etc.