What are the steps to move from derivative quant to algo quant, in a bank or hedge fund? Is that easy to leverage the skill set? What are the main skills apart from languages like R, database ones, etc. and time series analysis, etc. ? Any good introduction papers/books on the subjects (not academical)? Any example of people doing the move after 3-5y of experience in derivatives? Currently many banks are building fixed income algo trading platforms, equity being used for many years already...

First step - forget mathematics.

What about statistics?

Only very basic. Algo trading is more about common sense - how to reasonably code some sorcery of technical analysis: support, resistance, channels, candles, flags, Elliot waves and all that poetry. To be honest - there are trading policies based on mathematical notions like correlation, mean reversion and price momentum but they are used in a not very orthodox way.

So all the claims that algo trading is more "scientific" than derivatives in the risk-neutral paradise are BS?

There are many flavors of "algo trading", which one are you interested in? If it is fixed income flow trading in a bank it is one thing. But there are also high-frequency prop firms that trade stocks, the ones that trade FX, hedge funds that trade everything, CTAs that trade futures, from low to high frequency. The single most important skill is working with data: getting it somewhere, processing, filtering and testing whatever ideas you have. If you are developing production systems then you need to know multithreading, FIX or other communication protocols, market rules, asset class specifics etc.

QuoteOriginally posted by: katastrofaSo all the claims that algo trading is more "scientific" than derivatives in the risk-neutral paradise are BS?It is definitely more demanding. You must really know programming, not just simple translating formulas into C++, you must understand trading policies. Maths is there last and least. And of course you're rewarded for that, quant's package is peanuts these days, what we all know very good. I'm thinking of academia.

I am more interested in arbitrage (e.g. FX spot trading) rather than execution trading. I guess a good exercise would be to backtest technical signals on historical data using a language like R. Then i could play with real-time prices using a broker (demo account). My point was more about the way to get an interview and move. Things to add on the resume like small projects like above. Cases of people who did the move.

QuoteOriginally posted by: ZhuLiAnI am more interested in arbitrage (e.g. FX spot trading) rather than execution trading. I guess a good exercise would be to backtest technical signals on historical data using a language like R. Then i could play with real-time prices using a broker (demo account). My point was more about the way to get an interview and move. Things to add on the resume like small projects like above. Cases of people who did the move.Stat arb is hard to get into because as an industry it is quite small if measured by number of people. For example, $1 billion fund can have only 1-2 researchers. Everyone is looking for "star traders" with proven Sharpe Ratios >3 on hundreds of $mil capital. If you are not one of them there is a strong demand for developers.Playing with R and trading your own account is useful for learning but will probably not get you an interview.

QuoteOriginally posted by: GamalQuoteOriginally posted by: katastrofaSo all the claims that algo trading is more "scientific" than derivatives in the risk-neutral paradise are BS?It is definitely more demanding. You must really know programming, not just simple translating formulas into C++, you must understand trading policies. Maths is there last and least. And of course you're rewarded for that, quant's package is peanuts these days, what we all know very good. I'm thinking of academia.That's not what I meant. I've been told, by people who should know it well, that algo is more "scientific", because you're theorizing less and sticking to data and empirically testable hypotheses more (it's more or less orthogonal to how much maths you use) then in the exotic derivatives space (how can you test that your model is describing the dynamics of a 30-year old range accrual TARN correctly?). But now that you've mentioned technical analysis, Elliott waves and so on, you've kind of thrown what I've been told in doubt.

In algo trading you use all trading policies known from the "macro" world. Some are more scientific, some less.Look at the way traders talk. They've got precise trading algorithms but have serious problem in describing them. To extract all that from their minds and code it smartly - it's a real challenge. But still it's not maths related.

QuoteOriginally posted by: katastrofaQuoteOriginally posted by: GamalQuoteOriginally posted by: katastrofaSo all the claims that algo trading is more "scientific" than derivatives in the risk-neutral paradise are BS?It is definitely more demanding. You must really know programming, not just simple translating formulas into C++, you must understand trading policies. Maths is there last and least. And of course you're rewarded for that, quant's package is peanuts these days, what we all know very good. I'm thinking of academia.That's not what I meant. I've been told, by people who should know it well, that algo is more "scientific", because you're theorizing less and sticking to data and empirically testable hypotheses more (it's more or less orthogonal to how much maths you use) then in the exotic derivatives space (how can you test that your model is describing the dynamics of a 30-year old range accrual TARN correctly?). But now that you've mentioned technical analysis, Elliott waves and so on, you've kind of thrown what I've been told in doubt.If you can rigorously test your Elliot wave it is fine. It is an empirical discipline, if you can set up your experiment correctly and your theory fits, why not use it? Not that I know of quant traders using Elliot waves.

QuoteOriginally posted by: GamalIn algo trading you use all trading policies known from the "macro" world. Some are more scientific, some less.Look at the way traders talk. They've got precise trading algorithms but have serious problem in describing them. To extract all that from their minds and code it smartly - it's a real challenge. But still it's not maths related.Systematic macro is a different kettle of fish - bear in mind that the data is far less plentiful than price data and far less frequent. I would not have a hard time believeing that funds which are not huge would use some variant of technical analysis profitably.

Playing with supports and candles is quite common.Algo trading is different from derivatives pricing - in BS world we rather disclose models, algo funds keep everything in secret, so to be honest we do not know much and mostly guess.

I'm not sure where you got the idea that algotrading is about trying to formalize technical analysis but this is not generally correct. At many/most statarb hedge funds its more about sophisticated types of pairs trading where you try to exploit very short term price discrepancies between correlated/cointegrated instruments. As well as this statistical modeling, much of the research focuses on microstructural issues such as estimating the probability of trades being successfully executed, estimating transaction costs, etc. The math is very different from what is done in derivatives pricing, and is more about statistical modelling and data analysis; a PhD in statistics or machine learning from a top university is a standard (but not the only) way of getting an interview at a quant fund.