I am a quant researcher too developer for QTS Partners, a commodity puddle Ernie (author of this blog) founded inwards 2011. I assistance Ernie railroad train too implement several strategies inwards the puddle too diverse separate accounts. I wrote this article to give insights into a real of import business office of our strategy evolution process: the pick of information sources.
Our principal query focus is on strategies that monitor execution inwards milliseconds too that handgrip for seconds through several days. For example, a strategy that trades to a greater extent than than i currency yoke simultaneously must ensure that several executions accept house at the correct toll too inside a real curt time. Backtesting requires high character historical intraday quote too trade, preferably tick information for testing. Our initial focus was futures too subsequently looking at diverse vendors for the tick information character too quantity nosotros needed, nosotros chose Nanex information which is aggregated at 25ms. This means, for example, that attacker flags are non available. We purchased several years of futures information too laid to work.
Earlier this twelvemonth nosotros needed to update our information too discovered that Nanex prices had increased significantly. We besides needed quotes too trades, too information for to a greater extent than property classes including U.S.A. equities too options.
We looked at TickData.com which has expert information but is real expensive too y'all pay up-front per symbol. There are other services similar Barchartondemand.com too XIgnite.com where y'all pay based on your monthly usage (number of information requests made) which is a model nosotros produce non like. We ended upward choosing QuantGo.com, where y'all direct maintain unlimited access to years of global tick or bar information for a fixed monthly subscription fee per information service.
On QuantGo, y'all acquire figurer instances inwards your ain secure too private cloud built on Amazon AWS amongst on-demand access to a broad arrive at of global intraday tick or bar information from multiple information vendors. Since y'all ain too grapple the figurer instances y'all tin terminate direct whatsoever operating system, install whatsoever software, access the cyberspace or import your ain data. With QuantGo the master vendor information must rest inwards the cloud but y'all tin terminate download your results, this allows QuantGo to rent access to years of information at affordable monthly prices.
All of the information nosotros direct maintain used thence far is from AlgoSeek (one of QuantGo’s information vendors). This information is survivorship bias-free too is just equally provided past times the exchanges at the time. Futures quotes too trades download real speedily on the system. I am testing options strategies, which is challenging due to the size of the data. The information is downloaded inwards highly compressed shape which is thence expanded (by QuantGo) to a somewhat verbose text form. Before the toll split, a twenty-four hours of choice quotes too trades for AAPL was typically 100GB inwards this form. Here is a information sample from the amount Options (OPRA) data:
Timestamp, EventType, Ticker, OptionDetail, Price, Quantity, Exchange, Conditions
08:30:02.493, NO_QUOTE BID NB, LLEN, PUT at 7.0000 on 2013-12-21, 0.0000, 0, BATS, F
08:30:02.493, NO_QUOTE ASK, LLEN, CALL at 7.0000 on 2013-12-21, 0.0000, 0, BATS, F
09:30:00.500, ROTATION ASK, LLEN, PUT at 2.0000 on 2013-07-20, 0.2500, 15, ARCA, R
09:30:00.500, ROTATION BID, LLEN, PUT at 2.0000 on 2013-07-20, 0.0000, 0, ARCA, R
09:30:00.507, FIRM_QUOTE ASK NB, LLEN, PUT at 5.0000 on 2013-08-17, 5.0000, 7, BATS, A
09:30:00.508, FIRM_QUOTE BID NB, LLEN, PUT at 6.0000 on 2013-08-17, 0.2000, 7, BATS, A
These I convert to a to a greater extent than compact format, too filter out lines nosotros don't demand (e.g. NO_QUOTE, non-firm, etc.)
The character of the AlgoSeek information seems to live high. One examine I direct maintain performed is to tape alive information too compare it amongst AlgoSeek. This is possible because the AlgoSeek historical information is forthwith updated daily, too is i twenty-four hours behind for all except options, which varies from 2 days to 5 (they are striving for two, but the procedure involves uploading all options information to exceptional servers --- a pregnant task). Another examine is done using OptionNET Explorer (ONE). ONE information is at 5-minute intervals too the software displays midpoints only. However, past times executing historical trades, y'all tin terminate run into the bid too inquire values for options at these 5-minute boundaries. I direct maintain checked 20 of these against the AlgoSeek information too institute exact understanding inwards every case. In whatsoever event, y'all are gratuitous to contact the information vendors direct to larn to a greater extent than virtually their products. The lastly examine of information character (and of our marketplace model) is the comparing of alive trading results (at i contract/spread level) amongst backtests over the same period.
The information offerings direct maintain latterly expanded dramatically amongst to a greater extent than information partners too forthwith include historical information from (QuantGo claims) "every telephone commutation inwards the world". I haven't verified this, but the add-on of elementized, tagged too scored intelligence from Acquire Media, for example, volition permit us to backtest strategies of the type discussed inwards Ernie's latest book.
So far, nosotros similar the system. For us, the positives are:
1. Affordable Prices. The argue that the toll has been kept relatively depression is that master vendor information must live kept too used inwards the QuantGo cloud. For example, to access years of U.S.A. information nosotros direct maintain been paying
Five years of U.S.A. Equities Trades too Quotes (“TAQ”) is $250 per month
Five years of U.S.A. Equities 5 infinitesimal Bars $75 per month
Three Years of U.S.A. Options 1 infinitesimal bars $100 per month.
Three Year of CME, CBOT, NYMEX Futures Trades too Quotes $250 per month
2. Free Sample Data. Each information service has gratuitous demonstrate information which is actual existent historical information where I tin terminate select information from the demonstrate engagement range. This allowed me to sentiment too move amongst the information earlier subscribing.
3. One API. I direct maintain i API to access dissimilar information vendors. QuantGo gives me a coffee GUI, python CLI too diverse libraries (R, Matlab, Java).
4. On-Demand. The mightiness to select the information nosotros desire "on demand" via a subscription from a website console at whatsoever time. You tin terminate select information for whatsoever symbol too for simply a twenty-four hours or for several years.
5. Platform non proprietary. We tin terminate purpose whatsoever operating organisation or software amongst the information equally it is beingness downloaded to virtual computers nosotros fully command too manage.
Because all this is done inwards the cloud, nosotros direct maintain to pay for our cloud figurer usage equally well. While cloud usage is continuing to drib rapidly inwards toll it is nonetheless a variable cost too it needs to monitored. QuantGo does render or thence real-time billing estimates too alarms y'all tin terminate preset at dollar values.
I was at starting fourth dimension skeptical of the restriction of non beingness able to download the information vendor’s tick or bar data, but thence far this hasn't been an number equally inwards do nosotros exclusively demand the results too our derived information sets. I'm told that if y'all desire to purchase the information for your ain computers, y'all tin terminate negotiate direct amongst the private information vendor too volition acquire a discount if y'all direct maintain been using it for a land on QuantGo.
As nosotros purpose the windows operating organisation nosotros access our cloud computers amongst Remote Desktop too at that spot direct maintain been some latency issues, but these are tolerable. On the other hand, it is a large payoff to live able to start amongst a relatively modest virtual car for initial coding too debugging, thence "dial up" a much larger car (or grouping of machines) when y'all desire to run many compute too information intensive backtests. While QuantGo is latterly launched too is non perfect, it does opened upward up the the world of the highest institutional character information to those of us who produce non direct maintain the information budget of a Renaissance Technologies or D.E. Shaw.
===
Industry Update
(No endorsement of companies or products is implied past times our mention.)
- A novel site for jobs inwards finance was latterly launched: www.financejobs.co.
- A novel software package Geode by Georgica Software tin terminate backtest tick data, too comes amongst a fairly rudimentary fill upward simulator.
- Quantopian.com forthwith incorporates a new IPython based query surround that allows interactive information analysis using infinitesimal grade pricing information inwards Python.
Workshops Update
My side past times side online Quantitative Momentum Strategies workshop volition live held on Dec 2-4. Any reader interested inwards futures trading in full general would produce goodness from this course.
===
Managed Account Program Update
Our FX Managed Account plan had an unusually profitable calendar month inwards October.
===
Follow me on Twitter: @chanep
Tidak ada komentar:
Posting Komentar