• C++ Programming for Financial Engineering
    Highly recommended by thousands of MFE students. Covers essential C++ topics with applications to financial engineering. Learn more Join!
    Python for Finance with Intro to Data Science
    Gain practical understanding of Python to read, understand, and write professional Python code for your first day on the job. Learn more Join!
    An Intuition-Based Options Primer for FE
    Ideal for entry level positions interviews and graduate studies, specializing in options trading arbitrage and options valuation models. Learn more Join!

MFE vs. MS Computer Science?

Aerial

Joined
4/2/11
Messages
15
Points
11
Based on what I've read on this forum and WSO, it seems that programming is a very important aspect, if not, the most important skill of someone planning to enter Quant Finance. With this in mind, could someone explain compare/contrast which program is more advantageous - MFE or MS Comp Sci?

From what I know, MFEs are designed more on pricing derivatives at investment banks whereas for algorithmic trading, MS Comp Sci is much better? Is this true or false? Please correct me if I'm wrong, thanks.
 
alain said:
CS has almost no advantage unless you go into Automated Market Making of flow but there are not that many players.


CS will probably have some advantage but it's hard to tell.
Click to expand...

Ah alright thanks. I'm guessing that if one aspires for a career in trading (in general) one should concentrate more on the mathematical aspect (stochastic calc, Differential Eqs etc) and less on the programming aspect (C++, Java C# etc.)?
 
Aerial said:
Ah alright thanks. I'm guessing that if one aspires for a career in trading (in general) one should concentrate more on the mathematical aspect (stochastic calc, Differential Eqs etc) and less on the programming aspect (C++, Java C# etc.)?
Click to expand...

Depends... what kind of trading?
 
Alex Krause said:
Depends... what kind of trading?
Click to expand...

Flow trading vs. algorithmic trading .

I'm just trying to figure out if I should double major in Finance/Math or Finance/Comp Sci. Gut feeling says go with Comp Sci but then again it seems that CS is pretty much useless for BB S&T (all we need to know is Excel, VBA). Just wondering, is it true that algorithmic trading only applies for trading equities? Does algo trading work too for credits and derivatives?
 
Aerial said:
Flow trading vs. algorithmic trading .

I'm just trying to figure out if I should double major in Finance/Math or Finance/Comp Sci. Gut feeling says go with Comp Sci but then again it seems that CS is pretty much useless for BB S&T (all we need to know is Excel, VBA). Just wondering, is it true that algorithmic trading only applies for trading equities? Does algo trading work too for credits and derivatives?
Click to expand...

Stochastic calc and DE are pointless for any kind of trading. The markets don't abide by that sort of math. If you want to do strategy/modeling, you'd be better off studying signal processing, information theory, and machine learning. But you can't really get those sorts of jobs unless you're a PhD. You'd need to start off as a developer and hope you can get lucky and cross over. Without a masters degree even that is pretty much impossible.

The following depends on time horizons, as well as the type of strategies you're talking about. But generally algo trading is in equities, FX, commodities, and derivatives. It's making headway in FX, partly due to t-costs decreasing a lot. Not sure about credits but probably not. The most important necessity is liquidity.
 
Alex Krause said:
If you want to do strategy/modeling, you'd be better off studying signal processing, information theory, and machine learning. But you can't really get those sorts of jobs unless you're a PhD.
Click to expand...

Is a PhD really a requirement to do strats/modeling? I've heard people say you shouldn't do a PhD for career goals, only if you're interested in academia.
 
Of course not, only to do research, and even there you're flexible (you may be able to score a research position with a master's degree, it probably won't be at a big name though). In all other areas of quant you will find a VERY healthy dose of individuals with masters degrees. If you look hard enough, you might even find a few without master's degrees...

And virtually all models you build will be built on top of a massive stochastic base. While all models are inherently inaccurate, to call them "useless" is a hefty overstatement. Especially when you go beyond the simple financial products (i.e. stocks) and approach the more "interesting" (complicated) ones. The only part of finance that stochastics are inapplicable for is super-high frequency (although the definition of the time frame this encompasses varies from financial product to financial product), typically market making or order execution, where the world becomes more logical and deterministic.

If you want to do flow trading, study trading. Go get an MBA in finance from Harvard and godspeed. That seems to be the best way to get into that kind of position. Either that or start off in some other part of the firm on the quant side (i.e. desk strats/risk management) and make your way towards front office within the firm.

Machine Learning et. al. will help you most in high frequency algo. It's not as useful in flow trading.
 
Alexei Smirnov said:
Of course not, only to do research, and even there you're flexible (you may be able to score a research position with a master's degree, it probably won't be at a big name though). In all other areas of quant you will find a VERY healthy dose of individuals with masters degrees. If you look hard enough, you might even find a few without master's degrees...

And virtually all models you build will be built on top of a massive stochastic base. While all models are inherently inaccurate, to call them "useless" is a hefty overstatement. Especially when you go beyond the simple financial products (i.e. stocks) and approach the more "interesting" (complicated) ones. The only part of finance that stochastics are inapplicable for is super-high frequency (although the definition of the time frame this encompasses varies from financial product to financial product), typically market making or order execution, where the world becomes more logical and deterministic.

If you want to do flow trading, study trading. Go get an MBA in finance from Harvard and godspeed. That seems to be the best way to get into that kind of position. Either that or start off in some other part of the firm on the quant side (i.e. desk strats/risk management) and make your way towards front office within the firm.

Machine Learning et. al. will only help you if you want to go into high frequency algo... it will do nothing for you if you want to do flow... not all trading is created equal ;)
Click to expand...

He was referring to strats/modeling in buy-side algo trading. Assuming we are talking about high frequency here, 99% of these guys have PhDs. That is a fact. You can start off as a developer, impress the researchers, and if you get really lucky be given your own book or a research role. However, this is rare and you shouldn't enter the position with this intention because you will almost certainly be disappointed.

Stochastic models are not relevant for trading. Please tell me any trading strategy you've heard of that uses stochastics (and no, a "stochastic oscillator" is not related to stochastic theory). Regardless of frequency, traders don't like to make bets on randomness...
 
Alex Krause said:
Please tell me any trading strategy you've heard of that uses stochastics.
Click to expand...

Fixed income strategies where you need to model interest rates, credit strategies, mortgages. All those will use stochastic calculus. Keep in mind the equity world is nothing in comparison to the fixed income world when it comes to notional.
 
alain said:
Fixed income strategies where you need to model interest rates, credit strategies, mortgages. All those will use stochastic calculus. Keep in mind the equity world is nothing in comparison to the fixed income world when it comes to notional.
Click to expand...

What percentage of the high frequency space does this make up?
 
Fair enough, "flow" trading typically refers to non-automated trading (as Alain mentioned, there are few automated players), although I suppose execution algorithms are often used for at least some of it. I personally assumed that when the OP mentioned "flow" trading he meant what is stereotypically meant by "flow" trading, not an algorithmic implementation thereof (since these are rare compared to actual flow traders at IBs).

As far as a strategy that uses stochastics - Avellanada and Lee's paper is based off of a stochastic mean reverting process. While their execution methodology is deterministic, it attempts to take advantage of excessive deviations of the process.

In the sell side, I have seen information presented that stochastic optimal control may be advantageous. Although to be fair, since I'm not very interested in algo trading, I never bothered to investigate further.

Also, a stochastic oscillator is indeed related to stochastic theory. You are assuming the underlying asset has a characteristic with a certain distribution, and you are measuring the deviation of that distribution from your expected path for it. This is, to the best of my understanding, the basis of mean reverting strategies. Now, algo traders might be doing this without paying much respect to what it is that they are assuming, but that does not make the underlying that they are trading deterministic. Unless of course this is HFT and they have an algo with a sharpe of ~100 (i've heard those exist... but very rare and probably generate little alpha).

Basically any kind of model out there assumes some sort of process with some sort of distribution. Because if there is no stationarity, there is no point in modeling something - your model will provide no insight. And, of course, outside the high frequency space, asset price moves are not deterministic. Actually supposedly nothing about assets is deterministic. If anything was, there would be no markets. Prices would be eternally "efficient".
 
Alexei Smirnov said:
As far as a strategy that uses stochastics - Avellanada and Lee's paper is based off of a stochastic mean reverting process. While their execution methodology is deterministic, it attempts to take advantage of excessive deviations of the process
Click to expand...

In pairs trading (pretty much arbed out at this point by the way) s-score offers a marginal (if any) improvement over z-scores and this use of stochastics is trivial. For high frequency: signal processing, information theory, machine learning. Signal processing, information theory, machine learning.

Alexei Smirnov said:
Also, a stochastic oscillator is indeed related to stochastic theory. You are assuming the underlying asset has a characteristic with a certain distribution, and you are measuring the deviation of that distribution from your expected path for it. This is, to the best of my understanding, the basis of mean reverting strategies. Now, algo traders might be doing this without paying much respect to what it is that they are assuming, but that does not make the underlying that they are trading deterministic. Unless of course this is HFT and they have an algo with a sharpe of ~100 (i've heard those exist... but very rare and probably generate little alpha).
Click to expand...

So by this logic, anything with a distribution must be a stochastic process. I'm not sure this proof would hold up to mathematical rigor. Even if that's true, compared to the difficulty level of the strategies we are talking about, this is trivial. You're putting way too much weighting on it. By the way, if the market behaved by stochastic theory there would be no arbitrage, by definition. In fact, profitable trading strategies are entirely deterministic (in expectation). That's the point! ;)
 
Back
Top