Can anyone who has been through the process point me towards a good set of skills to study before going for stat arb interviews? My understanding is that interviews for stat arb positions are very different from other quant interviews. For example, I've heard (maybe incorrectly?) that Black-Scholes and option pricing questions usually don't come up stat arb interviews. General tips are welcome too, of course.If a similar question has been posted before and gotten an informative response, would someone please point me to it?My background:- I'm American.- I recently finished a PhD in physics (high energy experiment) from an Ivy League university.- I have taken a job as a postdoc in the same field at a different Ivy League university.- I'm looking at statistical arbitrage because it seemed like the field in quantitative finance that was closest to my current work: i.e. analyzing large datasets and drawing statistical conclusions.So far, I've been brushing up on:- C++ algorithms (I also have a background in Python, but I hear that C++ is the standard in finance now)- Probability and Bayesian statistics- Logic and brain teaser questions- General math: quick arithmetic, integration, etcThanks for any help or insight.

Start looking at the recent thread 'Interview with big Quant hedge funds '

QuoteOriginally posted by: neuroguyStart looking at the recent thread 'Interview with big Quant hedge funds 'Thanks for the prompt reply. That's an interesting thread.I'm primarily interested in a perspective on stat arb interviews in particular.

- BudFox2013
**Posts:**23**Joined:**

I would suggest a couple of intermediate books to break the ice: "Statistical Arbitrage" written by Andrew Pole, which a good book, not too technical but at the same time it does give you a few formulas that help you understand what stat arb does, why it was so successful in the past and why it stopped working. Another good reading is the book "inside the black box", which is good to tile the ground, as an introduction to the field but it is not technical at all. Also give a look at the book "The Quants", written by Patterson, which tells how stat arb was born and how it has evolved since then. Concerning the stats that you need to step into the field probably you should read "Stanford elements of Statistics and Data Mining", which is what people in algo trading mean when they refer to the field of statistics. Yet the classic book "Time Series Analysis" by Hamilton should be on your bookshelf as a general reference to what can be called "classical statistics".

- katastrofa
**Posts:**10260**Joined:****Location:**Alpha Centauri

Isn't G-research doing statistical arbitrage?a

QuoteOriginally posted by: katastrofaIsn't G-research doing statistical arbitrage?aWell thats why I referred him to the other thread.I have not seen any firms parading around saying 'hey, we do stat arb', but to a certain degree this is what all statistical (or 'scientific') quant strategies are all about.This assumes that we permit some elasticity in the definition of stat arb.Statistics -> Uses 'statistics' (driving in the rear-view mirror)Arbitrage -> A 'pattern' or strategy that we like so much that we think it is a dead cert. (So we will call it 'arbitrage')I am not sure that things are as ultra specific as Data seems to be anticipating.

Last edited by neuroguy on May 4th, 2014, 10:00 pm, edited 1 time in total.

Quote I am not sure that things are as ultra specific as Data seems to be anticipating.OK, fair point. Maybe I'm assuming too much. Here's what I've observed:- Lots of quantitative finance interview prep books want interviewees to know about: * Options/derivatives pricing * math/stats * code/algorithms * brain teasers ("you have a fox, a chicken, a bag of corn, etc...")- However, when I talk to my friends who've gone from from physics to finance (N=3): * None of them had to answer questions about pricing financial products. * All of them had to answer the other types of questions.- It happens that my friends all went to stat arb positions, but maybe that's just a red herring.Maybe I should draw on the broader experience of people in this forum and ask:1) "Is it common for people who come from academia with PhDs in non-financial fields to be asked about pricing financial products in their interviews?"and, more generally:2) "If you participated in an interview with someone with a PhD who came from an academic background (either as interviewer or interviewee), what kind of questions were asked?"Thanks

QuoteConcerning the stats that you need to step into the field probably you should read "Stanford elements of Statistics and Data Mining", which is what people in algo trading mean when they refer to the field of statistics.Thanks for the tip. I couldn't find a book with that exact title. Is this the book you're referring to?http://statweb.stanford.edu/~tibs/ElemStatLearn

QuoteOriginally posted by: DataQuote I am not sure that things are as ultra specific as Data seems to be anticipating.OK, fair point. Maybe I'm assuming too much. Here's what I've observed:- Lots of quantitative finance interview prep books want interviewees to know about: * Options/derivatives pricing * math/stats * code/algorithms * brain teasers ("you have a fox, a chicken, a bag of corn, etc...")- However, when I talk to my friends who've gone from from physics to finance (N=3): * None of them had to answer questions about pricing financial products. * All of them had to answer the other types of questions.- It happens that my friends all went to stat arb positions, but maybe that's just a red herring.Maybe I should draw on the broader experience of people in this forum and ask:1) "Is it common for people who come from academia with PhDs in non-financial fields to be asked about pricing financial products in their interviews?"and, more generally:2) "If you participated in an interview with someone with a PhD who came from an academic background (either as interviewer or interviewee), what kind of questions were asked?"ThanksThis is also discussed in the other thread.1) I think standard brainteasers of the type 'how many hedgehogs would fit on the pinnacle of the empire state building' are not commonly asked to PhD quant-types, especially not on the buy side. This is more the kind of thing that newcomers to IB are asked. If you have someone with a PhD talking to someone with a PhD it should usually be possible to find better things to talk about. So generally I would anticipate that you would be asked more specific/academic problems than just 'brainteasers'. Although anything is possible of course.You might be asked about financial products either to gain some understanding of how widely you have read or within the context of a problem where the expectation is that you will be able to solve it with minimal domain knowledge. For example if I tell you something about the statistics of an instrument I then ask you something about that instrument. You dont necessarily have to know finance to answer such questions. Interviewers will be much more interested in what you can bring with you rather than whether you know stuff about finance that you can read-up on later if neccesary.2) I have never been asked silly puzzle type questions (by which I mean goats, boats, manhole covers, men in bowler hats, light switches etc...). I have always been asked more specific mathematical type questions. Some have been about dice or about coin flips... but always related to specific mathematical methods or ideas. Personally in interviews I will always ask questions with a specific mathematical basis (to which I either know the answer, or have bothered to work out, in writing, before hand). So usually questions are to do with stats or probability. Some of those questions will be tailored to what I see on the CV. Some might be less related, but knowable. The idea is that if it is a good candidate and they are truthful on their CV then they should be able to answer them and we will have a jolly old time.

QuoteOriginally posted by: DataI'm looking at statistical arbitrage because it seemed like the field in quantitative finance that was closest to my current workIt sounds like you are plenty smart or technically qualified to do the work. I would be looking instead at whether this is a guy who will fit in with our lifestyle or working culture, who will be happy and energetic doing the specific tasks we are asking of him, and whether the markets will really hold his interest.A guy with your background, who also demonstrates a curiosity and energy for what people are doing in trading, would be very attractive. So you might avoid asking questions like how late do people at the office usually work? And avoid saying things like I am surprised, that sounds pretty crude and simplistic, I imagined people were using some more elegant and high-level statistical methods to solve these sorts of problems.And avoid quoting specific trading strategies you have heard about. Since the people at a specific group may have a specific way of thinking about things, and be biased against people who think about things in a way that seems stupid or pop-culture to them. You want to be a teachable, trainable person, not a know-it-all.As to the Bayesian and C++ stuff, that sounds good. It may be there are smart particle physicists who are challenged when it comes to thinking in that way. If you are not interested in programming, I think it is hard to fake it though this is an area where you could study the popular interview challenges. So far as Bayesian thinking, you could go through examples to make sure you at least understand what it is.I like the popular challenge of suppose there are two envelopes, and one contains double the amount of money in the other. If someone gives you one, and then offers you the option to trade it for the other, should you take it? In theory, you stand to gain 100% or lose only 50%, so you should always switch, right? Or it goes something like that.

- katastrofa
**Posts:**10260**Joined:****Location:**Alpha Centauri

Also remember that not every statistician is a Bayesian and there are valid arguments for both ways (Bayesian and frequentist) of thinking about probability.

QuoteOriginally posted by: katastrofaAlso remember that not every statistician is a Bayesian and there are valid arguments for both ways (Bayesian and frequentist) of thinking about probability.P(bayesian|statistician)=P(statistician|bayesian)P(bayesian)/P(statistician)

- katastrofa
**Posts:**10260**Joined:****Location:**Alpha Centauri

Mind your priors!

I wanted to give a big "thank you" to everyone who answered. This was extremely useful.

- katastrofa
**Posts:**10260**Joined:****Location:**Alpha Centauri

QuoteOriginally posted by: farmerQuoteOriginally posted by: katastrofaAlso remember that not every statistician is a Bayesian and there are valid arguments for both ways (Bayesian and frequentist) of thinking about probability.P(bayesian|statistician)=P(statistician|bayesian)P(bayesian)/P(statistician)I and my colleagues at work are Bayesians and we are collectively exhausted... TGIFa