Evolving Strategy

Quote from oldtime:

so this system i'm working on periodically reduces size on your biggest loser but it adds to your second biggest loser. It doesn't touch your biggest winner, but it adds to your second best winner.
More...

I actually coded up a money management line similar to this the other day and added it to my system. It modified the equity curve just like a stop loss (cut it by two thirds).

The solution to the market chop is to reverse the trend following signal when volatility is above an arbitrary threshold.
 
Quote from Wide Tailz:

I actually coded up a money management line similar to this the other day and added it to my system. It modified the equity curve just like a stop loss (cut it by two thirds).

The solution to the market chop is to reverse the trend following signal when volatility is above an arbitrary threshold.
More...
I've tweaked it so much it's barely recognizable from my first Saturday morning idea. Not sure if the system is evolving or I am evolving, I started trading it live with very small size and so far have had very encouraging results, but I scrapped the biggest loser biggest winner idea because it added too much during stagnation. I adjusted max winner size and a lower max loser size and another poster gave me a better idea, I call it rehobothing, because that's the posters name that adds based on price rather than time. But so far, the whole thing stays very stable in the chop and then adds pretty aggressively in the trend.

I just wanted something that views a whole portfolio as an evolving strategy.
 
Some of the best ideas I have seen regarding adaptive algorithms came from some of my old (1966) college text books on analog computers that use feedback loops to alter the characteristics of their pre-amps and filters.
 
Quote from oldtime:

at any rate, do any of you Big Cool Number Guys work on strategies that evolve over time? Especially adapting to trending from choppy markets and visa versa?
More...

Just develope a tiered vol-based algo switching management program.

IE:

Tier 1: Activate X strat when IV on the vehicle is <=30%
Tier 2: Activate X.1 strat when IV is 30-40%
Tier 3: Activate X.2 strat when IV is 40-60%
Tier 4: Activate Z strat when IV is 60%+
DEFCON5: Deactivate all
 
Strategy evolution is a noisy business. Here's why: at least when systems for daytrading are concerned (avg 10 trades per day), you may tweak your system for the latest few months so that in backtest it gives better results. But the problem is, a few months is a very small sample, and due to the complexity of some of the systems, it may just be that a few "lucky numbers" are the better (ie more evolved) parameters. That's why it's best to create systems via backtesting a huge amount of data (ie 12 years) and then if needed adjust the system for different volatilities. Better yet, create a system that self-adjusts (changes parameters gradually) as volatility changes and you get the best results that way. So, evolving systems per se is not really possible - this is because to evolve means to use a smaller sample, and a smaller sample is much more prone to curve fitting aka lucky numbers within your parameters.
 
Writing a complete specification of an adaptive algorithm in simple English is the most difficult part of of developing it.

Writing the code is the easy part. :)
 
@oldtime

I think,

there is only one way, what you can be.
A scientist or a trader.

To become a trader, you must have been a scientist, to created Master Trading Plans.
After that, the trader part comes and you must trade your plans.
there is no more researching and fixing your strategies.

NEVER CHANGE A WINNING SYSTEM.

What do you want ? Want make money or want to create hundreds of different systems ? There must be a END someday !!!

I am all through this bullshit. The END was the Best in my trading career.

Yes, its interesting to see how the market does on other ideas, but just for fun --- nothing serious about that.

If you have your Master Trading Plans, you have all you need for your life time.

Trading is about MAKING MONEY, not about creating and improving your strategies all the TIME !!!

Find something easy that works and what you can mental and physical do, without getting crazy ... AND do it, over and over and leverage up, step by step, thats how money is made.

Think about it.
 
Quote from ogarbitrage:

Just develope a tiered vol-based algo switching management program.

IE:

Tier 1: Activate X strat when IV on the vehicle is <=30%
Tier 2: Activate X.1 strat when IV is 30-40%
Tier 3: Activate X.2 strat when IV is 40-60%
Tier 4: Activate Z strat when IV is 60%+
DEFCON5: Deactivate all
More...

Great post.

I actually just tried a 2 tier stop loss based on consecutive losers, to kill some of the drawdowns.

It boosted the 1 year profit by about 20%.

This is where it's at - programming several systems that switch in and out during changing market conditions. Talk about an edge...
 
Quote from HATEtheRisk:



NEVER CHANGE A WINNING SYSTEM.
More...

Agreed, however the parameters might need to be recalibrated from time to time, at the discretion of the scientist in command........

Enough people figured out the 40 day breakout turtle buy signal to fade it. One system was even called "turtle soup".
 
You must log in or register to reply here.
Back
Top