AlgoTrading applications - what strategies are really applicable?

[rarefish]

Hi, I'm a software developer working for a company, providing application services for trading and risk management on electronic markets.
Now I'm studying approaches to algorithmic trading, looking to design and implement such a system either as a part of ours or independently.
So, the questions are:
1. What trading strategies and technical analysis methods, according to your experience or opinion, give best opportunities when applied in algorithmic trading?
2. Can Fundamental analysis be really formalized and applied here?
3. What's the state of the art with automatic analyzing the news? Is that possible at all?

Thanks

 

[Adamus]

hi rarefish, welcome to the forum. The thread title looked interesting, but seriously you're asking too much. My internet-surfing brain doesn't judge the answers to your questions to be something I can produce in a reasonable period of time.

I would have to write a ten thousand word essay to answer your first question properly, and I don't know the answers to 2 and 3

Just because I like Prague though, I thought it would be cool to get you talking a bit more

So what is your capitalisation? That would make a big difference. What sort of data do you want to use and what sort of statistical analysis do you want to see to judge your algorithms?

 

[Brabed]

So what is your capitalisation? That would make a big difference.

I disagree, a system design should not take capitalization into account. It simply doesn't matter. A good system should be able to take $100 to $1,000 just as easily as it would take $100,000 to $1,000,000, should it not.

Rarefish, in answer to #1. Any 'strategy' that uses quantitative data at it's core is the best candidate for algorithmic trading. I know that's very general, but think on it. Most typical TA is much too 'analog' for good quant trading. In example, a moving average changes on every tick, much too difficult to manage and it doesn't really 'tell' a good quant system what it needs to know. An algorithmic system needs rock solid quant data such as 'waypoints' of price. What was yesterday's high versus the previous high and low etc.? stuff like that.

#2 and #3 Not really, unless you can find a way to have manual inputs for these things that give a 'bias', but no algorithmic system is going to 'think' through what effect the current NFP numbers, for example, will have on price.

A good quant system will plow through most of this type of stuff anyway. NFP is really the only news trading day that should be avoided like the plague.

I hope this helps answer your questions.

 

[DT]

I disagree, a system design should not take capitalization into account. It simply doesn't matter. A good system should be able to take $100 to $1,000 just as easily as it would take $100,000 to $1,000,000, should it not.

Well given that a decent portion of the systems out there are high frequency I'd say that it matters a lot given that a large portion of the edge in this area is going to be attributable to execution. For a start someone trading in size is going to want to minimise the market impact of their orders across exchanges, ECNs, dark pools etc... this isn't a simple task!

 

[Brabed]

Well given that a decent portion of the systems out there are high frequency I'd say that it matters a lot given that a large portion of the edge in this area is going to be attributable to execution. For a start someone trading in size is going to want to minimise the market impact of their orders across exchanges, ECNs, dark pools etc... this isn't a simple task!

Well, yes, when you are talking about execution, account size does matter, no doubt.

 

[rarefish]

Hi All,
Thanks for your interest and replies, I appreciate sharing your experience.
Now I'll try to be more specific...

Let's assume that order executing techniques like iceberging, smart order routing, etc., are properly used, and we're not going to discuss them here. Let's think of pre-trade analytics only.
The basic information I'm going to use as inputs are the market data history, as well as the order book. In case it's really needed I will consider using 3rd party services providing some extra data on trading or analytics. The system will be used not for trading ForEx-es only, but at least with stocks and derivatives as well.
By now I collected plenty of information on TA as well as FA, and would like to ask:

1) regarding TA... I selected such areas as:
1a) Chart patterns recognition (e.g. H&S, triangles, flags, wedges, etc.)
1b) Candlestick patterns analysis
1c) Trends, support and resistance lines analysis
1d) Trading gaps
1e) Technical indicators analysis

Which ones would you personally use for trading your own money, and which ones you think are not worth wasting time on?

2) regarding FA...
There are indeed quantitative as well as qualitative factors listed in the financial statements. It's rather impossible to get the qualitative factors formalized enough to use them in AT. But it might be useful picking some quantitative parameters, like the history of revenue and expenses, cashflows, etc. They won't be used alone, but they can make TA signals stronger or weaker.
So, is the form and layout of financial statements standard enough to make the automation possible, or are there services that gather this data and provide in some standardized format? What do you think of the idea itself? Is it worth developing?

3) regarding other AT strategies... Which ones you think would really work out?
3a) arbitrage
3b) market making
3c) inter-market spreading
3d) any other you might think are interesting?

4) This question might seem naive, but why do you (or some of you) mind NFP?

I do understand that these are quite a lot of questions for a single message or even for a single thread... But the thoughts in my head need to get ordered better before I continue with my initiatives.

 

[egro1egro]

Hi All,
1) regarding TA... I selected such areas as:
1a) Chart patterns recognition (e.g. H&S, triangles, flags, wedges, etc.)
1b) Candlestick patterns analysis
1c) Trends, support and resistance lines analysis
1d) Trading gaps
1e) Technical indicators analysis

Which ones would you personally use for trading your own money, and which ones you think are not worth wasting time on?

All of them are not worth wasting your time on. Somebody gave you a hint already - "quantitative". None of what you listed is a "quantitative". You might want to look at DSP, TSA, AI and plain old statistics for a start.