Successful Automation?

inCom · Dec 31, 2005

Quote from Norm:

Hello,

I would like to hear from people <snip>

Thanks in advance,
Norm

Hi Norm

My system is composed of 484kb of C++ source code for core trading algorithm, broker API communication (IB) and low level stuff. I wrote a library which is a clone of WealthLab's WealthScript trading model (have no fear Cone, I've no intention to be the next WL competitor!).

On top of that I run my front-end, written in Microsoft Access, composed of roughly 35k lines of VB code amounting to about 3MB.

Using the whole system, which took my several years to develop and still is, of course, under development, I can write, backtest, monitor, evaluate and run multiplie strategies. The system is running in auto mode for about a year and a half now. In this period my return has been around 50% annualized.

Aside from development, which I enjoy because I'm a professional programmer but which was a great effort anyway, the most difficult thing is finding truly profitable strategies.

In the last couple of months I stopped running the system because the strategies I was using were becoming unprofitable. Even if I'm not (yet) a rich man, and despite my trading equity is small, I've learned enough to know when to limit losses and swallow the pride. They were all EOD pullback systems for equities. I'm now testing other systems, one of them being a customized version of the "dummy" pattern discussed in another thread on this board.

However, one of my objectives still is to learn how to become profitable trading "by hands", while keeping on developing my ATS.

Good luck and good trading to all.

-GS

ct3232 · Jan 2, 2006

Hi Norm

Yes, most definitely it is possible (and most likely preferable) to use an automated system.

I developed my own relatively simple system using Easy language - approx 35 lines of code. It is a variation of a volatility breakout system. It has been running fully automated for just over a year now. So far, approx 55% win ratio on 152 trades.

The key to trading an automated system is to have the discipline to not override the systems entries and exits. Its very tough to watch the market go against your position and not want to do something about it.

For me, the reason to go fully automated was one of time. Having a full time career and living on the West Coast, makes it difficult to trade "hands on". I personally prefer to trade the currency futures markets and not ES. Using an automated system allows for trades to occur at all hours of the day and not just when I'm at the screen.

As GS says, the hard part is to develop a profitable strategy in the first place. FWIW, I think its relatively easy to develop good entries. The tough thing is designing good automatic exits. This is where I spend 100% of my tinkering time.

CT

SideShowBob · Jan 2, 2006

CT,

Would you mind mentioning what you trade -- which securities (stocks, futures, etc)? Is it a single security, a basket of stocks?

SSB

ct3232 · Jan 3, 2006

SSB

The only future I have traded in the past year is the British pound futures contract. I find this has more volatility than the Euro, whiich means more trades and like currencies in general, tends to trend enough to squeeze out a few points profit.

The electronic "pit' allows for trading during the nighttime hours, when a lot of the larger moves seem to occur. I trade using the hourly charts and have generally been quite happy with Tradestations fills.

CT

SideShowBob · Jan 3, 2006

CT,

Thanks for the info. I spent a lot of time playing with the Euro and the Swiss Franc -- I didn't even look at the Pound. I guess I should take a look

SSB

satchel · Jan 15, 2006

Just began adding matrices to the front-end. Had an idea last week on how to evaluate setups across timeframes based on mkt conditions. Then being able to multiply instrument matrices against certain set criteria matrices for signal processing or even a quick row vector. For visuals, got them thrown into grids. Down the road I hope to be able to dump the forward-test matrix data into manual backtesting (it is a mystery to me why there are no decent backtesting packages - they all sit on top of some barely adequate charting package - you have to roll your own and that can take months).

Another 1,000 lines code added since last posted. A little bit of C++ and VB, dream it build it!

Cesko · Jan 15, 2006

Quote from satchel:

Just began adding matrices to the front-end. Had an idea last week on how to evaluate setups across timeframes based on mkt conditions. Then being able to multiply instrument matrices against certain set criteria matrices for signal processing or even a quick row vector. For visuals, got them thrown into grids. Down the road I hope to be able to dump the forward-test matrix data into manual backtesting (it is a mystery to me why there are no decent backtesting packages - they all sit on top of some barely adequate charting package - you have to roll your own and that can take months).

Another 1,000 lines code added since last posted. A little bit of C++ and VB, dream it build it!

I just trade. Feel inferior though!
They build automated trading systems so they don't have to do the trading, then they hire people to babysit that(automated) system. Since I do the trading I think I will need a babysitter for myself too (cute one).

Futures_Man · Jan 16, 2006

Ez language is ok for retail guys , any CTA or Hedge fund R&D
will never use it, as much as you try you will be limited to the script (not just for ez language).

take a crack at Strategyrunner's API its C++ based (not for beginners) , but it is event driven , and the major advantage is once you compile your Systems (DLL) you deploy it on a server (in an FCM or FX provider)

Many hedge funds and CTA use it for their blackbox trading portfolios , again this is not for the average retail Joe

rickty · Jan 16, 2006

Quote from Futures_Man:

take a crack at Strategyrunner's API its C++ based (not for beginners) , but it is event driven , and the major advantage is once you compile your Systems (DLL) you deploy it on a server (in an FCM or FX provider)

Recently I looked at SR closely for possible deployment of my trading strategies. The biggest problem I have with them is that even if you do all your own programming, SR still needs to verify your code (at a cost to you (of course)) before you can deploy it.

Also I like the flexibility of being able to tweak the code when needed without it again having to be re-verified.

Richard

Futures_Man · Jan 16, 2006

Quote from rickty:

Recently I looked at SR closely for possible deployment of my trading strategies. The biggest problem I have with them is that even if you do all your own programming, SR still needs to verify your code (at a cost to you (of course)) before you can deploy it.

Also I like the flexibility of being able to tweak the code when needed without it again having to be re-verified.

Richard

Yes, that is True, (though I suppose if you do large volumes I know they can be flexible on the verification costs)

Though still from an execution standpoint and flexibility and Power of the API I think that for high volume complex blackbox systems it a good solution (for above retail level users)