QuoteOriginally posted by: katastrofaMarket for exotic derivatives quant is very shallow now. It's better with flow derivatives.flow derivatives quant is glorified IT.

QuoteOriginally posted by: EscapeArtist999QuoteOriginally posted by: katastrofaMarket for exotic derivatives quant is very shallow now. It's better with flow derivatives.flow derivatives quant is glorified IT.There is no glorification in flow trading. The only silver lining is that flow traders will be extinct soon, so that will be positive for comp...

QuoteOriginally posted by: qqqqqIf 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.Eliot Waves seems more like something who has no science background would try to develop rather the opposite ?

QuoteOriginally posted by: ZhuLiAnLots of information here about algo trading. That's very helpfull but does only partially answer the initial question: how a derivative quants (5y+ exp) can move to algo trading? One idea would be to find a part-time MSc involving participating in a algo trading competition and get a good track record.Do an MSc at a top place in statistics or machine learning. Machine learning might be the best bet because it is more of a buzz-word. Get experience of applying it to real world data. Possibly look for some kind of lower paid work at a fund or prop-shop (i.e. prostitute yourself).Google ?Kaggle?.More general comments:There is confusion about what algo-trading means. Most people in the business that I have spoken to would define 'algo-trading' as the sell-side practice of order execution optimisation. Banks (eg. Deutsche, Citi) and some brokers provide this service. Bigger hedge funds also do it in-house. An example of this is VWAP strategies. Sometimes this stuff also involves an element of statistics and game theory. One fund I have looked into works on predicting market depth. The other stuff: Trend following, regression-based strategies, hidden Markov models etc... is really just fancy systematic trading. Yes it uses computers and algorithms, but in some cases it is simply more advanced forms of the stuff that good traders have been doing forever. You only need to read a couple of famous trading books to realize that many of these approaches are not new. Note that many systematic funds are not HFT. Winton has been mentioned. They may or may not be getting into HFT, but none of their advertised products are based on it. The original CTA fund had an average period of something like a month I believe. Cantab can turn over every 15 mins. Not HFT! Even some banks are offering investment products (ETFS) based on systematic trading strategies. About the math: Can be very mathematical or not so much. Its not stoch. Calc. ? true - but there is math beyond that! Important stuff is linear algebra, statistics, probability theory. Crucially, also need to be able to understand how to use data, what is reasonable to use it for and how to avoid over-fitting and other pathologies. This is the part of the skill set that would really distinguish it from a derivatives quant and the part that requires the most experience to acquire.HFT: Maybe stands alone a little, because the edge is not even in any technical factor as in systematic trading, its just speed and information based. Remember, if you are doing things very fast, this really limits the complexity of the analysis. So it would not surprise me if most of this stuff is heuristic. A good thing about getting into systematic trading in particular is that it is likely to be something that carries on into the future, since it is really just 'trading with computers', i.e. quite a broad church. Furthermore, you have the exit strategy of starting your own shop, or even just running your own money. HFT is harder to set up by yourself because the tech and data is so expensive and I think it will eventually get regulated away anyhow. Algo-trading is another business in which you have clients - with all the vagaries that brings.

Neuroguy, which degrees could be obtained part-time and considered good enough to get a foot inside this area ?

Neuroguy, which degrees could be obtained part-time and considered good enough to get a foot inside this area ?

QuoteOriginally posted by: bluetrinQuoteOriginally posted by: qqqqqIf 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.Eliot Waves seems more like something who has no science background would try to develop rather the opposite ?Sometimes people with no science background have decent ideas. Now, I am certainly not a believer in elliot waves. However I think they are interesting because they are based on the idea that the market has a geometry of its own. Indeed eliot waves are basically fractals. I think there is some wisdom in this. I dont believe all the hocus-pocus about exact ratios and retracement levels. But when I watch the market on many time scales I get the feeling that there is structure there which I do not observe if am looking at an autoregressive process. One can also think of decent mechanistic reasons why you might get self similarity in price time-series. The tricky bit is that the generators of the self similarity are additionally transformed in some non-trivial way, which is, I believe, the issue that stops people making money out of it.

Dont know about part time, but in the uk:UCLCambridgeEdinburghare the best places for machine learning and I know factually that they are specific targets for many quant funds. But getting some skills from somewhere parttime and combining that with some experience/project work could also work fine. Birkbeck is supposed to be good for parttime courses. Dont know if they do ML specifically, but I knew a guy who did a bioinformatics MSc there part time and then got employed by Glaxo. So check out their website.

QuoteOriginally posted by: neuroguyQuoteOriginally posted by: bluetrinQuoteOriginally posted by: qqqqqIf 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.Eliot Waves seems more like something who has no science background would try to develop rather the opposite ?Sometimes people with no science background have decent ideas. Now, I am certainly not a believer in elliot waves. However I think they are interesting because they are based on the idea that the market has a geometry of its own. Indeed eliot waves are basically fractals. I think there is some wisdom in this. I dont believe all the hocus-pocus about exact ratios and retracement levels. But when I watch the market on many time scales I get the feeling that there is structure there which I do not observe if am looking at an autoregressive process. One can also think of decent mechanistic reasons why you might get self similarity in price time-series. The tricky bit is that the generators of the self similarity are additionally transformed in some non-trivial way, which is, I believe, the issue that stops people making money out of it.That seems very sensible and I never meant to imply the opposite, I was only trying to find an explanation about why he could not find quant traders using Eliott waves: someone with a quantative background will formulate the ideal of fractals in a different manner.If you want to go in the specifics of Eliott waves, one problem I have with the people who swear by it is that they have a non systematic way to make their Eliott count and they even change it as the time series develop.Did you have any success implementing this fractal idea ?

No success. Thats my point. It might have some conceptual validity, but there are massive issues with using these ideas. So it does not surprise me one bit if no actual professional traders really use it. In comparison, simple things like regression, strict risk control and relatively large numbers of concurrent positions (as allowed by computational trading) actually make money. This approach is used by many. Another thing that actually probably works is hidden Markov modelling used to discover latent market states. There is speculation that this is what Rentech do. This comes a bit closer to the Elliot stuff, not in terms of actually mathematical construction (this bears no similarity at all) but conceptually: The idea being that there are certain regularities that are masked by non-trivial noise and distortion.

QuoteOriginally posted by: neuroguyDont know about part time, but in the uk:UCLCambridgeEdinburghare the best places for machine learning and I know factually that they are specific targets for many quant funds. But getting some skills from somewhere parttime and combining that with some experience/project work could also work fine. Birkbeck is supposed to be good for parttime courses. Dont know if they do ML specifically, but I knew a guy who did a bioinformatics MSc there part time and then got employed by Glaxo. So check out their website.I went through Birkbeck website quickly.I did not see ML specifically (as you predicted), but saw this degree: http://www.bbk.ac.uk/study/pg2013/stati ... STOPR.html

It seems others have asked this before:http://wilmott.com/messageview.cfm?cati ... id=83162No substantive solution it appears. I know very little about that course or Birkbeck, so my advice would be to contact them and find out what the alumni end up doing. It is possible that the course you cite contains fairly 'dry' stats rather than applied. This is important because the kind of stats quant funds want is stuff that you can apply then and there to some horrible great data set. It also helps if you have some experience of dealing with real-data issues too (eg. missing or outsize values). This is what machine learning is more focussed on, although it also makes heavy use of formal stats. Whatever you do, dont do something half way or for the sake of it. However, since Birkbeck is university of London and is right next to UCL, it might not be that hard to 'audit' some ML lectures at UCL (i.e. just go to them). Also if you have some fire and will, you might be able to to do a project between Birkbeck and UCL. You will need to persuade a UCL supervisor to accept you. I dont see that this is necessarily that difficult as long as you show some passion and interest. UCL also has some specific finance-machine learning cross over as part of the DTC for computational finance: http://fc.cs.ucl.ac.uk/.If you want more info. PM me.