• 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!

What does eTrading do in a bank? (eFX, eRates, etc)

1. What does it do? My understanding is they build the platform and logic to sell and auto-hedge vanilla-ish products up to say swaps + optimal execution to minimize costs

2. Do they develop the pricing models or just piggy back on the models to build trading strategies?

3. Any books on the subject from sell side perspective?
 
1. Depends. On some asset classes where you do principal market making (like eFX), you can take discretionary action and ride the pnl. For more standard stuff (like cash equity) in agency trading, you are basically some reporting monkey to jerk the clients off by BS on the benchmark used in TCA.

2. They don't develop pricing models. Wheter to develop pricing models or not depends on non-linear features of the products. They don't even develop pricing models anymore for listed option trading. People are doing data analysis to feed the "projected data" ( div/vol/repo ) into the industry model (SVI for example) just to have a systematic view of the market (across all tenor/maturity/strike). For linear products, there is no other dimensionality so market price is even good enough (lolz)
For pricing models you have to go exotic derivatives, there are always new features....
 
1. Depends. On some asset classes where you do principal market making (like eFX), you can take discretionary action and ride the pnl. For more standard stuff (like cash equity) in agency trading, you are basically some reporting monkey to jerk the clients off by BS on the benchmark used in TCA.

2. They don't develop pricing models. Wheter to develop pricing models or not depends on non-linear features of the products. They don't even develop pricing models anymore for listed option trading. People are doing data analysis to feed the "projected data" ( div/vol/repo ) into the industry model (SVI for example) just to have a systematic view of the market (across all tenor/maturity/strike). For linear products, there is no other dimensionality so market price is even good enough (lolz)
For pricing models you have to go exotic derivatives, there are always new features....
What kind of quant skills do they use then? Is it basically an IT job?
 
What kind of quant skills do they use then? Is it basically an IT job?
Almost none. The interview will be very statistics heavy. But on the job, avg(), std(), left join.
Hmm it is not pure IT job. You still need to think critical about your data. I would say it is more like... data analysis.
Normal IT folk don't possess such analytical skills.

now for exotics, even tho there is pricing models, but the liquidity is dry. so one day you might be no use and end up doing regulatory.. and trust me you don't want to get there... i would rather run asap to a tech firm if this day comes.
 
eTrading roles are closer to Strat than pure Quant. Some banks have 'Straders' who do everything from build the models, do the data analysis, and actually run the book (trade), while others split this into distinct roles: Quants/Strats who work side by side with the book runners (traders).

The quant/strat side of these roles require a three-pronged skill set: Business/market knowledge and intuition specific to your product (at a similar level to the traders/book runners), quant/data science skills, and stellar coding skills (often Python, Java, C++, and sql or kdb/q).
2. They don't develop pricing models. Wheter to develop pricing models or not depends on non-linear features of the products. They don't even develop pricing models anymore for listed option trading.
This is not entirely true, though it depends on the specific role; even linear products have pricing models, which attempt to optimize the price shown to clients in a volatile/liquid market, or try to attract flow that will be most profitable to the desk. Additionally, there is quite a bit of proprietary model development and alpha modeling, depending on the asset class.

While this is quite distinct from a traditional, pure quant role, there is a significant amount of quant work performed day to day, in addition to the data analysis, data modeling, and development.
 
Last edited:
eTrading roles are closer to Strat than pure Quant. Some banks have 'Straders' who do everything from build the models, do the data analysis, and actually run the book (trade), while others split this into distinct roles: Quants/Strats who work side by side with the book runners (traders).

The quant/strat side of these roles require a three-pronged skill set: Business/market knowledge and intuition specific to your product (at a similar level to the traders/book runners), quant/data science skills, and stellar coding skills (often Python, Java, C++, and sql or kdb/q).

This is not entirely true, though it depends on the specific role; even linear products have pricing models, which attempt to optimize the price shown to clients in a volatile/liquid market, or try to attract flow that will be most profitable to the desk. Additionally, there is quite a bit of proprietary model development and alpha modeling, depending on the asset class.

While this is quite distinct from a traditional, pure quant role, there is a significant amount of quant work performed day to day, in addition to the data analysis, data modeling, and development.
If definition of pricing model is some code to determine the best price for market making/quoting then sure. But to be precise (based on industry) I wouldn't really call these pricing models. They are really strat/trading work.

The core of derivative pricing is no-arbitrage. Or to be more precise, portfolio replication (that's how traders think when hedging it).

I would take the statement of significant amount of quant work with grain of salt. Not trying to pick a fight here but I guess it is just a matter of taste but to me it is fcking left join all day lol...

eTrading is all about flow/quote optimization based on market colour. The technique used is honestly completely different than traditional quants. Frankly, your description of the role is pretty damn accurate.

On the three skillsets you mentioned, the 1st one is probably the most important. Often in big shop, these 3 roles are seperated. When that happened, you often see very bad quality of code from folks doing the 2nd mandate, while the guy who implement the execution doesn't know shit about the product/bussiness. This is even worse in my opinion for new grads. The turnover rate is insanely high for this setup but managment doesn't care since they seperate the pieces so small.

FYI i was a former strat myself. And i thank god everyday that i am no longer one :P
 
While this is quite distinct from a traditional, pure quant role, there is a significant amount of quant work performed day to day, in addition to the data analysis, data modeling, and development.
If definition of pricing model is some code to determine the best price for market making/quoting then sure. But to be precise (based on industry) I wouldn't really call these pricing models. They are really strat/trading work.

The core of derivative pricing is no-arbitrage. Or to be more precise, portfolio replication (that's how traders think when hedging it).

I would take the statement of significant amount of quant work with grain of salt. Not trying to pick a fight here but I guess it is just a matter of taste but to me it is fcking left join all day lol...

eTrading is all about flow/quote optimization based on market colour. The technique used is honestly completely different than traditional quants. Frankly, your description of the role is pretty damn accurate.

On the three skillsets you mentioned, the 1st one is probably the most important. Often in big shop, these 3 roles are seperated. When that happened, you often see very bad quality of code from folks doing the 2nd mandate, while the guy who implement the execution doesn't know shit about the product/bussiness. This is even worse in my opinion for new grads. The turnover rate is insanely high for this setup but managment doesn't care since they seperate the pieces so small.

FYI i was a former strat myself. And i thank god everyday that i am no longer one :P
While I mostly concur with your points here, there is a matter of semantics. Quant work does not need to be narrowly defined within the confines of derivative pricing. While derivative pricing (and related topics) are a major component of MFE programs, not many quants actively work on derivative pricing (or similar) day to day. The pure definition of a quant is 'an expert at analyzing and managing quantitative data' (as applied to finance), and 'quant work' would be any work related to that. There is certainly a lot of 'quant work' within eTrading roles, but the roles themselves are not 'pure' quant, in that quant skills are just one of several skills utilized (they are 'strat' roles).

There is currently a strong industry trend to merge the above skillsets into single roles, for the very reasons you mention. There is a lot more to the role these days than doing left-joins, especially in eFX and eRates (though maybe interns/grads often get stuck doing much of the 'left-join' type work initially).

I very much enjoy my job in eFX, as it utilizes a good mix of my skills, the market dynamics can be fascinating, and it feels great to have a first-order effect on PnL. Of course, as you said, it's a matter of personal taste; to each their own.
 
Based on some of the job postings I've seen, these positions typically ask for prior experience in one or all of:

1. kdb/q, which seems to be used mainly in high frequency trading

2. Java or python or c++

3. knowledge of the linear products, markets and micro market structure for that respective desk (business knowledge as you say)

4. statistics/econometrics

5. curve construction in the eRates case

6. previous experience in a similar position

I am a FO derivs quant looking into making the switch. Would this be a trivial switch? I think most people could pick up kdb/q if you know how to code. Is such a role (or at least the interview) heavy on stats? Can you get by just knowing say ARIMA/GARCH and linear/logistic regression?

Also, do you have a view on the direction of growth of electronic trading in the coming years, in terms of number of jobs and compensation?

Thanks guys for the replies
 
You are right q/KDB+ is actually not hard. If you are sharp you should be able to pick it up. Learning curve is little bit steep and once you get familiar it becomes muscle memory.

You can read Jeff Borror’s Q for mortals (however it is pretty dry book :) lol)
 
Can anyone give an overview of how you come up with an etrading strategy for fixed income say?
You will know when you get to the job. There is a secret that they don't tell you in text book, which is the proprietary data the bank has. In e-trading, often in RFQ market, you don't really know the price unless you win the deal. Things like this, will give you the access to trade flow data that amateur do not have. When you have those information, certain uncertainty, becomes... certainty, if you know what i mean.

That is just the tip of the iceberg. When you get onto the job, you will know it. Just make sure you can manipulate data efficently and good intuition, which is probably all you need.
 
You will know when you get to the job. There is a secret that they don't tell you in text book, which is the proprietary data the bank has. In e-trading, often in RFQ market, you don't really know the price unless you win the deal. Things like this, will give you the access to trade flow data that amateur do not have. When you have those information, certain uncertainty, becomes... certainty, if you know what i mean.

That is just the tip of the iceberg. When you get onto the job, you will know it. Just make sure you can manipulate data efficently and good intuition, which is probably all you need.
This doesn't really explain
 
ok a simplest example (just hypothetical), lets say current last traded EURUSD is 1.25. Imagine it is a RFQ venures..
A client submit an order, and you post your quote.
Assume every one is positng quote as 1.24/1.26.
You decide to go more aggressive, 1.245/1.255 per se. (just hypothetical)

You get hit.

Now you are the only one with access to the trade flow, which is $1.245, while everyone else is in the dark because what they know is that EURUSD > 1.24 (or whatever price they posted).

During the day, you build yourself a pricing curve with all these trade flow data (imagine you have won lots of these). And you take on it.

Just some idea, you gotta be creative on what you want to do after.
 
ok a simplest example (just hypothetical), lets say current last traded EURUSD is 1.25. Imagine it is a RFQ venures..
A client submit an order, and you post your quote.
Assume every one is positng quote as 1.24/1.26.
You decide to go more aggressive, 1.245/1.255 per se. (just hypothetical)

You get hit.

Now you are the only one with access to the trade flow, which is $1.245, while everyone else is in the dark because what they know is that EURUSD > 1.24 (or whatever price they posted).

During the day, you build yourself a pricing curve with all these trade flow data (imagine you have won lots of these). And you take on it.

Just some idea, you gotta be creative on what you want to do after.
What's a pricing curve?
 
Top