Lessons Learned From Profitable System Development

[funky]

for me, i have found the biggest problem when you find a new system is the tendency to 'trade in a vacuum'. you are so enamored with the accuracy of some good examples, that you are completely ignoring all the other variables.

when i started using the floor pivots and observed and learned how the floor is swinging price, checking liquidity on both ends, and applied that into a strategy that i had developed, my trading improved dramatically. when you start to know during the day what has happened to price since the opening bell, you will understand and anticipate the correct moves. this will not only improve your strategy, but also your accuracy and selection will be beneficial to you, as you spend less time trading.

I currently employ 2 countertrend trades each morning, at times where i can count on meaning something.

 

[TriPack]

Quote from acrary:

Good systems:

- are easy to trade. They don't have rules like buy on open of next bar using 1 min. data. They give you enough time to actually get fills similar to the tested results.

- don't try to capture profits in all market conditions. Every system I have is targeted to a specific market behavior.

- have a edge superior to random entries and exits.

- are stable. Market behaviors with a edge happen with varying frequency. However they should be both frequent enough and with sufficient size moves so as to warrant taking prudent risks.

- are not curve fit. Use entries/exits versus random entries with same holding period and a chi square test to avoid fooling yourself.

Great comments acrary. I'm glad you stopped by this thread. From reading your past posts I know you have a lot of insights to offer on this subject. A couple of questions:

What are the categories of systems that you trade, how do you break down market conditions into classifications? What category of market condition are we in right now?

 

[0008]

Quote from acrary:

Good systems:

- are easy to trade. They don't have rules like buy on open of next bar using 1 min. data. They give you enough time to actually get fills similar to the tested results.

True! However, it depends on the market. Sometimes you can enter the market without rush. But you need to be quick in other situations. Of course, you can decide taking the signals or not but this may lead to cherry pick, a very dangerous practice.

 

[acrary]

Quote from TriPack:



Great comments acrary. I'm glad you stopped by this thread. From reading your past posts I know you have a lot of insights to offer on this subject. A couple of questions:

What are the categories of systems that you trade, how do you break down market conditions into classifications? What category of market condition are we in right now?

My production models are in two broad categories: money flow and breakouts. The money flow is one model based on inflows/outflows to mutual funds with a 1 week hold timeframe.
The breakouts are split into many sub-categories like (volatilty breakouts, opening range breakouts, trend continuation breakouts, daily reversal breakouts, mid day continuation breakouts, time of day based breakouts, trend day breakouts). I've tested just about every other kind of system I can think of. In all I track the results of about 25 models on a monthly basis. I'm trading against 8 of the models (which I found is all I can handle). The easiest way I've found to come up with ideas is to take a blank piece of paper and draw two or three bars. Come up with all the different ways the bars can show up and classify the description yourself. For instance, two rising bars could be a continuation, or if the first bar is smaller than the second, it could be a volatility breakout.

So far this year volatility breakouts have been the best and most consistent performer for me. Trend day breakouts have also done very well. Opening range breakouts didn't have much followthrough earlier this year. Continuation and reversal breakouts have done well, but there haven't been many of them.

 

[EliteThink]

acrary, are you only running your models on index futures?

 

[acrary]

Quote from EliteThink:

acrary, are you only running your models on index futures?

That's all these models are run on. I have another system setup with a diversified commodities portfolio and separate models. I haven't seen much interest in talking about physicals on this board so I don't post anything about them.

 

[EliteThink]

Interesting, thanks for the response. Many well known index models are crashing this year, do you have any thoughts there?

 

[acrary]

Quote from EliteThink:

Interesting, thanks for the response. Many well known index models are crashing this year, do you have any thoughts there?

There's been a dropoff in both trendiness and volatility this year.
That's usually bad news for profits.

Here's something that shows the trendiness and volatility for the past 5 years, last year, and this year. It can be used for comparing one period against another but not for saying things like trendiness dropped 20% this year compared to last.

 

[EliteThink]

Excellent acrary, thanks.

Back to tripack's topic. My system designing has traditionally been keeping a system in as much as possible and managing losses risk. Part of this is because I've found it quite difficult to quantitatively define an environment with enough time to take advantage of the market characteristics.

 

[TriPack]

5) If the results look too good to be true, double and triple test everything. If it still looks too good to be true, forward test it. It is usually about that point when you figure out why it looks better than it performs.