System Optimization Fitness Functions

August 14th, 2010, 07:08 PM

flyscalper

I would like to get an optimization type as a Calmar Ratio: Annual Profit / DD max...

For annual profit I would suggest systemPerformance.AllTrades.TradesPerformance.TradesPerDay * 365 * systemPerformance.AllTrades.TradesPerformance.Currency.CumProfit / systemPerformance.AllTrades.Count

(There may be better ways, that's without pulling up documentation or intellisense)

August 16th, 2010, 09:51 AM

Store the date of the first and last trades as follows:

Code

 DateTime start = new DateTime();
            DateTime stop = new DateTime();
            bool init = false;

            foreach (Trade t in systemPerformance.AllTrades.LosingTrades)
            {
                if (!init)
                {
                    init = true;
                    start = t.Entry.Time.Date;
                }
                stop = t.Entry.Time.Date;
                
            }

Then you can determine Annual Profits as (systemPerformance.AllTrades.TradesPerformance.GrossProfit + systemPerformance.AllTrades.TradesPerformance.GrossLoss) / number of years;

flyscalper

Hello,

I would like to get an optimization type as a Calmar Ratio: Annual Profit / DD max, but the problem is that I do not know how to compute the Annual Profit with the systemPerformance variables.

Un saludo,

Pablo

August 16th, 2010, 11:30 AM

Hi guys,

thank you for the suggestions. I will try it very soon. MooreTech, your approach is a little more complicated than the fluxsmith's one. Don't you think it is possible to calculate the way he says?

Un saludo,

Pablo

August 17th, 2010, 07:51 AM

Yes I believe his is a simpler approach. Didn't see his answer before I posted.

flyscalper

Hi guys,

thank you for the suggestions. I will try it very soon. MooreTech, your approach is a little more complicated than the fluxsmith's one. Don't you think it is possible to calculate the way he says?

Un saludo,

Pablo

September 12th, 2010, 09:33 PM

we found that using a total of 10000 totals works well - ie - 50 generations with 200 populations -

MooreTech

Yes I believe his is a simpler approach. Didn't see his answer before I posted.

October 5th, 2010, 08:49 PM

fluxsmith

I'm still trying to find the ideal fitness function. Net Profit / Drawdown as suggested here previously doesn't do it for me, as that ignores consistency and unrealized risk. The attached is the best I've come up with so far. For me the goal is to have a function that will sort to the top the same parameters I would select by hand based on visual analysis of the cumulative profit graph, which ideally would be a smooth progression from bottom left to top right; and when otherwise comparable preferring the smallest stops.

fluxsmith,

Sometimes I wonder if I go in circles, then I realize I haven't, which then makes me wonder even more .. I was using various 'optimization functions' available to NT7, and after a long day and night of jumping from one machine to another VM and back again (and seeing the same results too many times) I thought for sure I was ready for sleep so I did ... in the morning the results were still there ... blessed be: your Quality3.cs is giving me the same results, both on variables inputs AND $$/trades summary that the SQN.cs (

Download: SQN - System Quality Number (Van Tharp)

This is Van Tharp's System Quality Number (SQN) optimizer type for NinjaTrader. Place the SQN.cs file in your Documents\NinjaTrader 6.5\bin\custom\Type directory and then exit and reload Ninja. Here is information from Ragingbull on SQN: The code itself …

caprica

) is producing ...
the only diff is the very nice 'performance' rating supplied by your code (4+ rating in a test of a long only strategy in a period of choppy bearish action)

now I dont know if I should be

or

Any thoughts ? I KNOW the code is different, just wondering if you had noticed the parallel results?

Jon

October 6th, 2010, 12:12 AM

Trader.Jon

...Any thoughts ? I KNOW the code is different [vs SQN.cs], just wondering if you had noticed the parallel results?

They're based on the same principle, so they would often return the same scoring order of parameters. Mine does try to make some adjustments for conservatism, and annualizes the score, which is why on a backtest over less than a year you'll see higher scores on mine (and lower scores on a backtest over more than a year).

October 7th, 2010, 07:35 AM

why would it work better to try to backtest and optimize on a very long time frame - markets change all the time - i find that when i backtest for about 60-90 days and walkforward about 30 days it gives pretty good results overall for the next week or so to live run -
what i am still having a difficult time with is finding the right parameter set count - where a function or system will automaticlly calculate the population number and iterations it needs based on the parameter set count

October 10th, 2010, 12:28 AM

bizman70

why would it work better to try to backtest and optimize on a very long time frame - markets change all the time - i find that when i backtest for about 60-90 days and walkforward about 30 days it gives pretty good results overall for the next week or so to live run -
what i am still having a difficult time with is finding the right parameter set count - where a function or system will automaticlly calculate the population number and iterations it needs based on the parameter set count

I look at markets as having different personalities when they are bullish or bearish, so different parameters or even different straegies may have to be used.

Personally, I feel that it is better to optimize on the most demanding conditions: currently I am working on LONG ONLY(testing in choppy bearish time period) and SHORT ONLY(testing in choppy bullish time period) testing of strategies.

TJ

April 3rd, 2011, 05:49 PM

Before reading this thread I thought the GOOD strategy should have:

- Good profits
- Good win/loss ratio
- Few number of trades

And until now I have been using the attached code.

After reading this thread I've seen I've been simplystic, so I'll use some of your ideas. Also has anyone worked on Expectancy score optimization ?

System Optimization Fitness Functions

Discussion in NinjaTrader

System Optimization Fitness Functions