i want to know how many people use exogenous timeseries. by that i mean everything but price and volume as the basic ingredient. please note before answering that every indicator derived out of price is not such another time series, since it is price based. thnx.
You wrote "Price/Volume", what's the degree of the data you're referring to??? Bid/Ask Data? Intermarket data? Market Sentiments like TICK, TRIN? Anyways, let's start with you. Do you use it and how/what do you use it?
price only so far. and i must admit that i still have a human trader (at my broker) executing my orders. so i am by strict defintion not a real ATS ... i trade on 5min bars, which enables me to manage more money without going into the process of building my own vwap machine. i am running systems on daily data as well and tried to incorporate exogenous tine series there first, but it did not turn out attractive enough. now i am doing two things on intraday: first move done to higher frequency, since my current systems are just slightly positive for the year so far, which i find embarrassingly poor given these distortions and ranges we see. so i am moving down and start order routing in a first version with a matlab connection to IB on a 25k account. plus want to integrate exogenous stuff on intraday. and the current state of the poll is highly motivating in this respect ... . and the thread has fulfilled its purpose for me. thnx everyone.
i see "exogenous" as everything that does require additional information than the trading activity of the instrument at hand. so price, volume, bid/ask and so forth are coming from trading this instrument. everything that requires information on other instruments (single stocks, options or completely different instruments) i would call exogenous.
Matlab to me seems rather slow; how short of a time period bar do you think you can go to when using Matlab? By the way, what software are you using to interface Matlab to TWS? Richard
Thank you for your clarification. My answer is yes. I trade global equities, futures, FX and some options. I am required to used "exogenous" data under portfolio, system and trade-by-trade basis. 80% of my trading is automated. In terms of portfolio, I have a grand portfolio "model" that monitors my overall exposure. In terms of systems, I have both outright and stat. arb (pairs) trading. In terms of trade/trade basis, I have smart routing (multiple liquidity vendors/broker/dark pool) and algo. order models (VWAP, TWAP, Pounce... etc. etc.). I have a few other things going on but you get the idea. The other 20% is my own human aspect. I focus on the "non-computational" aspect of trading. I, myself, is the best pattern recognition and discrete AI out there, monitering underlying risks that my programs can't catch. And of course, I'm both the programmer and system developer for my trading. Until a few years back, I was a fully automated trader. But now, it's a matter of having a clear cut balance between my "expert system (computer)" and my "AI (human)".
Very interesting... How do you manage the AI/human function? Do you apply that kind of decision making within predefined parameters / specific situations? cheers, es175
its name is TWSlink. someone within my team is doing it, maybe he will post details ... our other stuff on trading is all C++. but C++ is currently a horrible bottleneck here. and the matlab-IB is just to get our feet wet in ground below 5min. will be very simple stuff to begin with. i want to know about slippage, order types - you know, all that basic stuff. we never ran anything fully automatic. so we are admittedly a little overexcited. the "virgin"-effect; no matter how bad the lover, it is first time ...