QuoteOriginally posted by: gianluca19QuoteOriginally posted by: spv205As with any field, you need many more people implementing theories than doing research.The large volumes of quants required was for programming/implementing new payoffs etc ( in existing pricing library).its not to say that people aren't encouraged to do their own research/ give presentations etc, but as catastropha says the majority of quant work is programming.you then have headquants presenting their research (derman, piterbarg, balland etc, and that hasn't really changed)- which may or may not ever be implemented within the bank.the reality is that there are just not that many new models required.gianluca19: they don't need Phds in maths- just smart people. Put it another way - they don't care whether your phd is in algebraic topology or stochastics. Its just a filter to get smart people with minimum effort- ie without interviewing 1000 people.what i don't get is why they need smart people to do a programming job...probably because they need to understand what they are implementing then...because mistakes are expensive.pay 25k/p.a. more to a smart guy who's probably not gonna make a huge mistake is a better deal than saving these 25k/p.a. and hiring some guy who might be sloppy. Are Phd automatically smarter? No. But it's a pretty good and really cheap filter.

Quotewhat i don't get is why they need smart people to do a programming job...probably because they need to understand what they are implementing then...This is a very interesting question. You voice what most upper management/non-programmers feel about programming but won't say openly. Smart people who are good programmers have several qualities that not so smart programmers do not have-1. Willingness and confidence to explore the unknown. Just yesterday we wanted to build a good user-interface on our data mining application. We found a ready made one in Ruby on Rails. None of us knew it. I managed to install it with a few hiccups and got our app hooked to it. Another person dug into the code and changed it to work as we wanted it to work. None of us knew Ruby or Rails before. Non-smart people do not do that. They say I know Java or .NET and freeze. Which is why I look for Pyhon experience these days. Those guys are super flexible.2. Pay attention to why they are doing what they are doing. They are egoistic enough to question you and this leads to constructive feedback and better solutions.3. Programming is hard. The difference between a smart programmer and a dumb programmer is huge. When you have team of quality programmers the results are obvious. As Gates used to say "If you hire A people, they go our and hire other A people and the organization grows. If you hire B people they will hire C people and then you are fucked over time".

QuoteOriginally posted by: gianluca19QuoteOriginally posted by: spv205As with any field, you need many more people implementing theories than doing research.The large volumes of quants required was for programming/implementing new payoffs etc ( in existing pricing library).its not to say that people aren't encouraged to do their own research/ give presentations etc, but as catastropha says the majority of quant work is programming.you then have headquants presenting their research (derman, piterbarg, balland etc, and that hasn't really changed)- which may or may not ever be implemented within the bank.the reality is that there are just not that many new models required.gianluca19: they don't need Phds in maths- just smart people. Put it another way - they don't care whether your phd is in algebraic topology or stochastics. Its just a filter to get smart people with minimum effort- ie without interviewing 1000 people.what i don't get is why they need smart people to do a programming job...probably because they need to understand what they are implementing then...Interesting viewpointDo you consider programming to more menial than let's say, maths? Or that it is easier? What no one here is mentioning is s/w design, maintenance and refactoring. Because no new models ==> less greenfield projects ==> more maintaining existing systems.

QuoteOriginally posted by: CuchulainnQuoteOriginally posted by: gianluca19QuoteOriginally posted by: spv205As with any field, you need many more people implementing theories than doing research.The large volumes of quants required was for programming/implementing new payoffs etc ( in existing pricing library).its not to say that people aren't encouraged to do their own research/ give presentations etc, but as catastropha says the majority of quant work is programming.you then have headquants presenting their research (derman, piterbarg, balland etc, and that hasn't really changed)- which may or may not ever be implemented within the bank.the reality is that there are just not that many new models required.gianluca19: they don't need Phds in maths- just smart people. Put it another way - they don't care whether your phd is in algebraic topology or stochastics. Its just a filter to get smart people with minimum effort- ie without interviewing 1000 people.what i don't get is why they need smart people to do a programming job...probably because they need to understand what they are implementing then...Interesting viewpointDo you consider programming to more menial than let's say, maths? Or that it is easier?Probably there are different levels of programming...but I remember when i used to program (i did the informatic olympiads) the hard thing was to imagine the reasoning, the mathematical stuff behind the problem you wanted to solve.then the programming part was just about memorizing the syntax and writing a code without errors.i'm not saying that this is easy, but if you have got the idea, then you will implement it, somehow. while if you don't have the right idea you will never finish.

QuoteProbably there are different levels of programming...but I remember when i used to program (i did the informatic olympiads) the hard thing was to imagine the reasoning, the mathematical stuff behind the problem you wanted to solve.then the programming part was just about memorizing the syntax and writing a code without errors.i'm not saying that this is easy, but if you have got the idea, then you will implement it, somehow. while if you don't have the right idea you will never finish.Programming for olympiads is a different sort of challenge from industrial grade programming. There you have a problem and you have to write a program for it in the language you know and no one is ever going to read it or use it beyond the challenge period. Industrial grade programming has several aspects to it which make it challenging-1. You do not have clear requirements and they change in the middle of tasks. 2. You know your project will survive you and your staff, meaning it will be be used in ways you cannot imagine now. So you have to develop your programs such that the next generation will both understand it as well as be able to expand on it. Which means ease of understanding and extensibility are more important than extra-smart routines3. You have to constantly choose between developing your own, using a third party library or using open-source and simply modifying it (without violating the license agreements)4. There are administrative aspects like source control, issue management, changing team members, etc5. And many more....If I said something along the lines why do you need smart traders? After all it is all about buying low and selling high. Those 3 months I did trading I successfully brought Garmin for 60 and sold it for 65 a few times easily making $3000 in 3 months on a portfolio of $30k. How does that sound? But unfortunately that is what a lot of people think about professional programming. Until they try their hand at it in the real world...

Surprisingly I know a lot more good mathematicians than programmers. One good one I ran into barely knew SQL, which is why I was brought into the project. In the end I ended up doing logistic regression for him as well in R as his solution to using categorical variables was creating a fancy 0,1 matrix for each value of the categorical value of a variable. It would have taken a week as we had large datasets and the categorical variable was a zipcode of united states. I defined the variable as a factor in R and just like that R did it internally and we had the logistic model in 1 hour. It helps to be a good programmer and learn how to program in several good languages. Or else you are an artist without tools. All good fantasy but nothing to show for it.

QuoteWhat no one here is mentioning is s/w design, maintenance and refactoring. Because no new models ==> less greenfield projects ==> more maintaining existing systems.Cuch don't depress them any further! Quants aren't good because they are great programmers. just as there are no new models, there are no new architectures- they are slotting code into an existing code library. adding a new payoff in quantlib is not the same as designing quantlib.

QuoteOriginally posted by: gianluca19QuoteOriginally posted by: CuchulainnQuoteOriginally posted by: gianluca19QuoteOriginally posted by: spv205As with any field, you need many more people implementing theories than doing research.The large volumes of quants required was for programming/implementing new payoffs etc ( in existing pricing library).its not to say that people aren't encouraged to do their own research/ give presentations etc, but as catastropha says the majority of quant work is programming.you then have headquants presenting their research (derman, piterbarg, balland etc, and that hasn't really changed)- which may or may not ever be implemented within the bank.the reality is that there are just not that many new models required.gianluca19: they don't need Phds in maths- just smart people. Put it another way - they don't care whether your phd is in algebraic topology or stochastics. Its just a filter to get smart people with minimum effort- ie without interviewing 1000 people.what i don't get is why they need smart people to do a programming job...probably because they need to understand what they are implementing then...Interesting viewpointDo you consider programming to more menial than let's say, maths? Or that it is easier?Probably there are different levels of programming...but I remember when i used to program (i did the informatic olympiads) the hard thing was to imagine the reasoning, the mathematical stuff behind the problem you wanted to solve.then the programming part was just about memorizing the syntax and writing a code without errors.i'm not saying that this is easy, but if you have got the idea, then you will implement it, somehow. while if you don't have the right idea you will never finish.In real-life industry, the situation is much much different. There you build systems with 100 people, 3000 customers and the s/w must last 20 years. And make a profit."memorising syntax" I have never heard that before. Strange..

Quotei'm not saying that this is easy, but if you have got the idea, then you will implement it, somehow. while if you don't have the right idea you will never finish.Indeed!You never get the right idea correct first time (and maybe never). That's software products. Rel-life requirements are never nicely scoped, as in exams or the lecture hall, for example.

Quote"memorising syntax" I have never heard that before. Strange..He probably means programming constructs - "While,for,do-while" loops, "global,local,static variables", "recursion", "basic structures like arrays,lists, trees,graphs (maybe)". Those are enough to solve to classroom problems. Most people cover that in their classes in school and think that is all there is to programming. This person just voices what most people think about programming. I remember once when a problem showed in huge application and one of the MD's said on a call publicly "Why did the coders not think of something like this while coding?". Truly a classic. That is why self-respecting programmers do not stick it out in IT in finance.

QuoteOriginally posted by: capafan2Quote"memorising syntax" I have never heard that before. Strange..He probably means programming constructs - "While,for,do-while" loops, "global,local,static variables", "recursion", "basic structures like arrays,lists, trees,graphs (maybe)". Those are enough to solve to classroom problems. Most people cover that in their classes in school and think that is all there is to programming. This person just voices what most people think about programming. I remember once when a problem showed in huge application and one of the MD's said on a call publicly "Why did the coders not think of something like this while coding?". Truly a classic. That is why self-respecting programmers do not stick it out in IT in finance.yes that is what i meant + remembering a lot of functions and what they do...never had a programming class..i studied C++ on my own when i was younger.what i don't understand, then, is the difference between a quant and a programmer...

Quotewhat i don't understand, then, is the difference between a quant and a programmer...A quant is just a very good programmer whose primary domain is developing/maintaining quantitative software modules in the domain of finance (for this forum). It is a multi-disciplinary field where you need to be a very strong programmer, a pretty good and rounded mathematician and have good grasp of the financial domain you work in. Each of the skills is important but the first two you need to be bringing to the table, the last one you pick up. Conversely a Fund Manager needs to be bring the last skill to the table and he can learn ad-hoc programming skills in excel eventually.It is like a trader who also needs to be an accountant of sorts. We live in a world where people need to be multi-disciplined. Some skills like programming are basic but basic does not imply easy or trivial. I do not know what you do for a living or if you are still a student but if you find yourself in a field where software is developed you will appreciate that the difference between a good and a average "software engineer" is huge. It is unfortunate that this fact is not very well respected. One reason for that happening is that hiring decisions are short-sighted. We need Ruby on Rails, Java, .NET , Python developers to do a specific task. But instead you should be saying we need a very good development staff managed by a developers. But I would not hold my breath on that happening on a mass scale. Finance IT tends to be good at this. They just are not good at respecting the role in the context of the organization.

QuoteIt is like a trader who also needs to be an accountant of sorts. We live in a world where people need to be multi-disciplined. Some skills like programming are basic but basic does not imply easy or trivial. I do not know what you do for a living or if you are still a student but if you find yourself in a field where software is developed you will appreciate that the difference between a good and a average "software engineer" is huge. It is unfortunate that this fact is not very well respected. One reason for that happening is that hiring decisions are short-sighted. We need Ruby on Rails, Java, .NET , Python developers to do a specific task. But instead you should be saying we need a very good development staff managed by a developers. But I would not hold my breath on that happening on a mass scale. Finance IT tends to be good at this. They just are not good at respecting the role in the context of the organization.Incidentally, academia could use some improvements in the research process (with developing software being a fundamental part thereof) as well, it's not the 19th century anymore (and even with that said, one thing that hasn't changed is that irreproducible research doesn't have much to do with scientific method):http://www.nature.com/news/2010/101013/ ... ryQuoteDue to an error caused by a homemade data-analysis program, on page 1875, Geoffrey Chang and his colleagues retract three Science papers and report that two papers in other journals also contain erroneous structures.

QuoteOriginally posted by: spv205QuoteWhat no one here is mentioning is s/w design, maintenance and refactoring. Because no new models ==> less greenfield projects ==> more maintaining existing systems.Cuch don't depress them any further! Quants aren't good because they are great programmers. just as there are no new models, there are no new architectures- they are slotting code into an existing code library. adding a new payoff in quantlib is not the same as designing quantlib.You imply that quantlib was designed by good programmers, which is not what I hear usually.

The best manager I worked for, at 75 was still coding the best sw libs our group had ever seen. He was also fluent in 7 languages and had seen his prior research been granted with a Nobel prize shared with Prof. Prigogine, even so, we would still hack us youngschmucks to smeetherens! Prof. Gregoire Nicolis was also one of the most humble individuals and also, the only "genius" I ever met. Then, in finance I met truly outstanding people of all ages, backgrounds, however none like Nicolis.