Darren Cook

570 reputation

312

bio	website	dcook.org/work
	location	Tokyo, Japan
	age
visits	member for	1 year, 7 months
	seen	May 21 at 23:29
stats	profile views	78

I'm director at QQ Trends, a company that solves difficult data and software challenges for our clients. (We often have freelance projects, so get in touch if interested.) (And, of course, please do get in touch if you have interesting challenges that you would like our world-class experts to work on!)

Typical work: doing fun stuff with data (fixing, mining, etc.), web sites (front and back-ends), trading strategies. Research: trading strategies, computer go, machine translation, understanding context, AI search algorithms. Languages: C++, PHP, R, javascript, and many more.

I'm British, living and working in Tokyo for 15+ years. Human Languages: English, Japanese (fairly fluent, 1 kyu), some German, Chinese and Arabic.

	570 reputation	bio	website	dcook.org/work	visits	member for	1 year, 7 months
312 badges		location	Tokyo, Japan		seen	May 21 at 23:29

summary answers questions tags badges favorites bounties reputation activity

52 Comments

suggestions reviews revisions comments badges posts accepts all

Sep 4	comment	ROC: difference between discrete and continuous? BTW, re: PerformanceAnalytics, "most of the functions assume arithmetic returns". I now notice user508's comment at quant.stackexchange.com/a/1082/1587 says "PerformanceAnalytics defaults to log returns." :-(
Sep 4	comment	ROC: difference between discrete and continuous? Thanks Joshua. The article was also very useful in pointing out the terminology, which was where a lot of my confusion had come from! (discrete == simple == arithmetic) (continuous == log == geometric)
Aug 22	comment	Is this a common variation of sharpe ratio? Thanks; now I know what I'm looking for I see R also has functions for `DownsideDeviation`, `SemiVariance`, etc. I didn't ask in the other question as: a)it was off-topic; b) you'd already expressed your bias, and I wanted to hear arguments against ;-)
Aug 22	comment	Drawdown calculation for strategies @Freddy BTW, I started a question about your comment on Sharpe Ratio: quant.stackexchange.com/q/3986/1587
Aug 5	comment	Why use a column database for tick/bar data? @JL344 Re the "Row-oriented databases just aren't designed to handle a huge number of tables as effectively as a huge number of rows in one table." comment, it seems this is a good point. MySQL uses one file per table, so it should be fairly unlimited. However see (serverfault.com/q/83438/87322) where meta queries can become notably slow once you have thousands of tables, apparently. (This is a MySQL bug that is fixable, rather than a flaw with the row-oriented concept.)
Aug 5	comment	Why use a column database for tick/bar data? @JL344 One table per symbol of OHLCV data in a row-oriented DB is very different from the column-oriented DB. In the former they are are stored in row order; in the latter they are stored in column order. In the former appending a new row requires appending, say, 44 bytes, to a single file; in the later it requires appending 8-12 bytes to each of 5 separate disk files. That is their advantage; the gist of Q1 and Q2 is for this usage scenario do column-oriented DBs bring any other advantage?
Aug 5	comment	Why use a column database for tick/bar data? @JL344 You didn't mention the row-oriented approach of one table per symbol (see the UPDATE in my question). Is there a reason people are not using that approach?
Aug 3	comment	Why use a column database for tick/bar data? @Freddy Thanks for the update, sorry I only just saw it. It sounds like you are confusing NoSQL DBs (which, to me, means MongoDB, CouchDB, Cassandra, Redis, etc.) with column-oriented databases? Or do you regard them all as NoSQL solutions?
Jul 17	comment	Are there any standard techniques for adding realistic synthetic microstructure noise to a price series? @Serg Is this (using the difference between `future` and `current`) any better than using the difference between `previous` and `current`? I.e. your answer implies that if the market is about to get more volatile (bigger jumps between prices) then there is more noise in `current`. Is that an observable phenomena?
Jul 10	comment	Why use a column database for tick/bar data? @chrisaycock By "aggregate" do you mean things like turning ticks into 1m bars, 1m bars into hourly bars, etc.? And/or do you mean making moving averages and other more complicated indicators? And/or something else?
Jul 10	comment	Why use a column database for tick/bar data? Thanks @chrisaycock I had read one of those in my hunt yesterday, but the first link I'd missed and it was very informative (I'm still working through the linked 85-page PDF, but that looks useful too). I've added more information to my question to explain why I don't feel the linked-to answers fully answer my question.
Jul 9	comment	Why use a column database for tick/bar data? Another question: I usually think in terms of one table per symbol (instrument/contract). Reading between the lines of your answer, when you have a column-oriented DB do you keep all symbols in one table?
Jul 9	comment	Why use a column database for tick/bar data? Thanks for the reply Freddy. When you say column DBs are all about Key/Value, what is the key? When I think tick/bar data, the key is a datestamp, but that is a row-oriented concept, isn't it?
Apr 16	comment	Free intra-day equity data source Answering my own question, Google Finance keeps 15 days of intraday data, 3 years of EOD daily data (I couldn't find an authoritative source, but a few blogs gave these numbers).
Apr 15	comment	Free intra-day equity data source Thanks, that is a useful source. How far back does the minute data go back? I tried with one stock, requesting 720d, but it only went back 18 days (March 26th to April 13th). Is there a way to specify start and/or end date and make multiple requests to get it further back?
Apr 11	comment	Free intra-day equity data source @user508 Great tip, thanks! I've just been playing with the IBrokers package, and I was indeed able to get minute bars for a day a year ago (for a Japanese equity). It also worked for an expired Mini-Nikkei futures contract for one day a year ago. OTOH, I couldn't get minute data for even a week ago for EUR/USD.
Mar 7	comment	Is there an open source alternative to Reuters Kondor+? Incidentally, anyone have a reliable link to a page about Kondor+ ? Reuters website keeps giving me 404s, even when following links in their own search results!
Feb 9	comment	Why is C++ still a very popular language in quantitative finance? This is a good point: I've been forced the write something in C# the past month because an API was only available in C#. They claimed C++ was supported, but gave no usage examples, and basically refused to answer C++ related questions.
Feb 9	comment	What concepts are the most dangerous ones in quantitative finance work? What didn't you like about the Wired article? It made a complex subject quite readable I thought; were their any factual errors, or was it just too dumbed down for your taste? (Incidentally the T. Mikosch paper was published Nov 2005; he must have felt vindicated come 2008 :-)
Feb 9	comment	How to remove the risk element from a set of fixed rate mortgage offerings? Thanks for the interesting answer Brian, though it left me confused: if there is no risk premium involved, why do the fixed rates go up the longer the term? Or do you mean there is nothing for me to remove and it is "pure" prediction about the way this bank thinks the BOJ benchmark rate will go? E.g. that it will go from 0.00% to 0.425% in the next year.