Q&A: Quantitative Analyst/Trader - Career Path, Technical Topics, Education

Sell-Side typically wants to go buy-side but not the other way around. It's also what the industry sees

It really depends who you ask and how happy they are with their jobs. Buy side typically has much fewer jobs than sell side so what happens is most people start on sell side and end up wanting to transition to buy side due to better pay, more meritocratic environment, smaller teams, more impact, reputation, etc. Buy side people don't transition as much to sell side because the thought process is totally different.

But I know a bunch of guys who built long and successful careers on the sell side. It depends on your personality and how good you are. Those who work their way up on sell side roles to MD or C-suite tend to have extremely strong interpersonal skills with clients and understand how to bring in business and money for the firm. Those who are geniuses in their respective areas or are the best traders end up taking their skills to the buy side and measure their success in terms of P&L or strategy. 

The best way to look at buy side verse sell side is to look at it from a quant standpoint: risk neutral probabilities (Q-quant) verse physical probabilities (P-quant). Without getting into the details (I recommend this paper to understand the differences: http://talus.maths.usyd.edu.au/u/UG/SM/MATH3075/r/Meucci_2011.pdf), sell side live in Q measure and as a result focus on market making and pricing to make a spread - they are directionally neutral. Buy side live in P and care about real world probabilities to forecast market moves and make a profit. The style of trading and thinking is totally different. Whatever world you prefer to live or think in can give you a clue as to what you may like.

My pleasure! Sounds like interesting work but I cannot speak much to agency or execution. I am on the sell side so focused on market making quant work only, i.e. building and maintaining and validating derivative pricing models which the traders use to price and make a trade. But I will give my 2 cents.

I think the idea of execution trading is a dying breed. The trading world has already shifted tremendously over the last few decades from outcry to pure electronic trading. Cash equity is fully automated. Futures markets proliferated immensely and can give your average investor exposure to just about anything. Even bond markets are even trying to shift and venture into electronic trading (https://www.marketwatch.com/story/electronic-trading-in-u-s-corporate-b…). Eventually everything will be automated and there will be no need for sell side trader thanks to people like you working on the execution algos.

Now that's not to say it's not in demand now. It's a valuable business especially for a bank or buy side firm doing high frequency trading. Theres plenty of pennies to be made and on a million dollars of notional it's an easy 10k in P&L off 1 trade. With consistent flow its a very profitable business. The emergence of the bond market can really pave the way for more execution strategies I would think.

Moreover the strategies that focus on optimal execution are also highly sought after in the buy side. A lot of it now is focused on machine learning. Reinforcement learning algorithms can help understand market participant behavior and react to new information. Neural nets are great for extremely high dimensional data like limit order book bid ask data. As ML becomes more widespread, so too will execution algorithms. Predicting other market participants behavior has been around since the beginning of trading but with algos it can be detected much easier. For instance, is your competitor pinging the market with hidden orders, trading in 48 hour intervals, chopping their orders into several individual orders through fragmentation, etc. Accurate predictions can help anticipate where markets move before they do and in order to do that you have to understand execution very well. This might not be something going away anytime soon.

Overall, I think execution strategy is hot right now and will continue to grow with the emergence of new tools in ML/AI. A lot of jobs are hiring in this area. But eventually there will only be room for the execution quant/strategist, not the trader themselves.

There's no silver lining trick. To be a great quant or trader you need to consume. By consume, I mean learn everything and read a lot of market articles or research papers. 

To be a good quant there are, what I like to call, 4 pillars: math, statistics, computer programming and finance. No one starts of strong in all four areas. Everyone will have one they prefer over the other. It's about how well you develop your weaknesses that make you good. I had a math background and sucked in coding but now I can price exotics in Python using simulation or other fancy stuff. Granted, I did a masters but the idea is you can learn quickly within a few years. The key is you don't need to be an expert at any one, just good enough to break down the problem into multiple parts and use the tools from each pillar. 

The quick and dirty answer to your question is focus on the weak pillar while being good at your strong pillar. And pick up a bunch of projects you can find. Or make some up (brownie points for creativity). For example, before my masters I learned how to code and implemented a stock prediction algorithm in Python using polynomial interpolation of stock prices. Looking back, it's a pretty garbage technical indicator but it conveyed interest in the field as I needed to collect data, come up with a strategy, understand math techniques to implement it, and backtest the results.

There are also a bunch of resources to code (see my below answer to midnightcowboy's question).

The most important thing especially for trading or research is to read a bunch of news. You need to understand what's happening with markets across all asset classes (equity, rates, commodities, credit, etc.) and also the political/macro landscape. Understand what catalysts move markets, drive prices, and provide or dry up liquidity. Be able to hold up a conversation with someone who asks you to pitch them a stock or tell you whats happening with the markets and where you see SPX, 10 year Treasury, CDS spreads, etc. in 6 months. Taking additional certifications like CFA can really help distinguish you can help you nail down the finance pillar. 

I will give you a concrete example of the 4 pillar process. For instance, a client wants to trade an interest swap. The first thing you need to understand is what the instrument actually is (finance part). Understand the contract itself and the implications it has on the people who trade the contract. What are they using the product for? To hedge or speculate on interest rates? How do the components of the instrument work, e.g. the fixed and floating legs of the swap or direct components like Libor rates. Next understand the pricing. It's a lot of mathematics to find the NPV of a swap. Some contracts use simple linear mathematics (forwards) to solve and other require fancy stochastic differential equations to understand (variance swaps). Next, you'll need programming to implement the pricer in practice and to numerically solve for a price (in this case the fixed swap rate that makes contract value 0 at initiation) implementing any formulas, algorithms, simulations, etc. C++ will be used for the speed in pricing and Python will be used for any additional analysis you might need. Finally, the statistics side will be used to understand how the contract behaved historically (e.g. time series of return data) or wanting to understand how to decompose what the factors that drive yield curve movements (parallel, steepness, and curvature) through a PCA analysis. This was a very detailed example but you get the point at how much learning you'll need to do.

It depends on the firm and what languages are used. For my bank we use mostly Python and C++. Python is used for any analytical tools such as computations, working with data, or plotting out charts. C++ is used more for the pricing models and any simulations done (pretty much anything that requires a lot of computing power, loops, etc.). Basically, we have a library where all the code is stored. I contribute to this library by maintaining code that's already in there or adding new code. For instance, we have scripts that have different pricing methods (Black Scholes, Hull White, etc.). We have various instrument classes with their respective features (e.g. interest rate swaps, Eurodollar futures, equity option, etc.). We have modules that provide exact calendar or holiday dates so that when pricing you have an accurate representation of time. There are miscellaneous scripts for mathematical tools like root finding algorithms, spline calculations and so on. The library is dense and has a lot of moving parts. It's your job to understand how to connect with what you already have to not reinvent the wheel and do something someone else has already coded. Automation is a huge focus across banks right now on making their processes more efficient through code. Github is used constantly for peer editing code and pushing/pulling changes other people might make to your code.

Advice: first learn the basics of coding like an IDE, packages, variables, conditional statements, functions, and so on from a website that covers the basics of a language. Then start doing small problems (Google "Python exercises") or Hackerank problems. Hackerrank shows up a lot on quant/engineering interviews especially for Goldman Sachs. I don't recommend reading books or watching other people when learning to code. What you have to do is think of a problem and try to implement it yourself in code. For instance, I come from an Excel background before my masters but Python is 10x better and more efficient. It won't break if you have more than 10k unique line items. A problem you would do in Excel, like make a computation if a condition is met can easily be done in Python. If you have no idea how to start, for each question you have use Stackoverflow and someone else probably had the same issue applied to a different problem. You translate their solution back to your problem and make the code work. Eventually doing this enough times, you become really good at debugging - the most important thing to learn when coding. The key is to find the solution on your own instead of asking someone else always how to fix your bugs. 

I started in risk management as a model risk quant after my masters. I don't know where in risk you'll be working or what your educational background is, but yes, the comments you have heard about middle office are generally true but not always. In general, if you can get a quant job right out of your bachelor's and you have no other outstanding offers or think you will anytime soon then don't think twice and take it. If you have your bachelor's and have worked in any other BO or MO role and transitioning to quant from something less appealing, take it. It will help lead you to the more sexier roles you might be looking for.

I'll elaborate on each point:

1) "Middle office is dull." Not necessarily true. I know many quants in MO that are extremely happy with the work and stay in their roles for 15+ years. The models are constantly changing and you have access to a wide variety of models across the firm (derivative pricing, forensic, ML) and across asset classes for capital market models (EQ, FX, IR, comm, credit, bonds, ABS, xVA, etc.). Regulatory landscape is changing and some people like that interaction or constant feedback. However, for the most part, if you are more trading oriented are like following market news and coming up with ideas you'll find it boring after a year or two. That's not to say the work won't be valuable for you to eventually transition - heck it happened for me, I'll be joining a fixed income research group.

2) "Pay is worse." Also a misconception but in general is true for further career progression. Some banks start you off pretty well in quant risk ($110-130$K) which can be attractive out of even a masters program. However, the salary increases in MO will be around 2-3% from my experience and bonuses depending on firm can be anywhere between 10-30%. So yeah, you won't be making the Q-millie or half-millie in a year but that's not to say you can't be happy with a $150-$200K all in for your first few years of working. Also,  this salary can be more valuable than a comparable banking salary as your work life balance is significantly better and the team and culture in quant risk are usually extremely chill (expect anywhere between 30-50 hours and even the lightest weeks working from home I have put in 20 or less). Really can't complain from a work life balance standpoint.

3) "Less career advancement." Also not necessarily true depending on your grind and background, but generally true for the masses who become passive from the pay and work life balance I mentioned earlier. Some people are generally happier and rather not move. 2 years in my job and I was getting a bunch of FO desk quant interviews multiple times a week, a couple of trading interviews (IR swaps, EQ options, FX options desks) and research (just the one in IR). Not sure how hot this job market is going to last though. It's all about how you approach your job interviews and tailor your resume. Mine was geared more toward trading, I actively keep up with the markets and trade my own PA, and have made progress towards the CFA. So it all depends on your will and drive. It'll be somewhat hard to get out of MO but not impossible. And in some cases like mine, even help you to transition to FO.

Best of luck!
 

Can you be a bit more specific on what you mean on my opinion on traditional traders? Traders who come from a non-quant background are generally very smart and have skills that some quant traders might not have. It really depends on the product and industry.

No clear cut answer on this but I would say getting to PM is not necessarily easier for a quant trader, in fact, I think it's the contrary. Anything more quant related tends to be more narrow focused and they tend to be siloed in their expertise whereas as general traders usually have a broad skillset that encompasses external factors related to a portfolio like better understanding of macroeconomics or traditional fundamental analysis.

QTs have a better edge on quant fund PM roles obviously and expertise on more technical products (derivatives related or high frequency). Think Two Sigma, Susquehanna, DRW, Citadel, IMC, etc. whereas your traditional traders would most likely be rise to PMs at asset managers, pension funds, endowments, and so on and manage a more standard portfolio (e.g. a 60/40 EQ/FI, target date fund, etc.). This is a generalization, of course, and it all depends on the person; there can always be overlap in skillsets. 

My pleasure. I am here as a resource. Congrats on the new role, that sounds very interesting. Can you send the link to your forum?

From what I understand about structuring, it is more geared to tailoring a product to a client's desired payoff or exposure. So with respect to volatility, you might could be looking at anything like options, variance/vol swaps, forward vol, vol of vol, VIX/vol ETNs, or any structured product related to the volatility of an instrument or index. For instance, they client may want a payoff exposure related to the volatility of an inflation rate, how do you come up with such a product? Clearly, CPI is not a tradeable asset so instead you use things that are tradeable on the market (most of the products structurers deal with are OTC). Could be a combination of a delta-hedged inflation swaption with some other product that removes external risk factors and your job is to understand how to construct such an instrument and match the desired payoff and hedge the risk throughout the trading horizon.

I will say although these systematic strategies might not be directionally related, getting a thorough understanding of vol products could be desirable for a fund. Pure market makers (Citadel, IMC, etc.) don't care about directional risk on a non-low latency level so it won't matter for them and in terms of exotic vol products; don't think it would be that useful for them as they focus more on basic equity options (think exchange-traded). However, for HFs involving vol strategies you might be hired in a risk capacity to hedge positions and can eventually transition into a trader, however, the skillset will be different and the road to get there might be tough (see my answer on going from sell side trading to buy side trading - same things apply here about needing experience in managing money and taking risk).

My advice to you is get as much exposure to your new position as possible and talk with traders and see what you actually like. Even though you come from a structuring role, you can be a good "sales person" in an interview and pitch your knowledge of trading strategies within the products you work with. Once you are a skilled professional with a few years of work experience and you have expert product knowledge under your belt in a desired and in specialized/technical products, what school you went to and your GPA doesn't matter anymore but rather your ability to crush interviews because you are an expert in your field and have the proof to back it up based on previous work you've accomplished or reputation from a top team at a large BB.