Benchmarks for mechanical trading systems?

@all,

I'm developer of mechanical trading systems (currently a german Dax Index system http://www.zentrader.de/html/trading_systems.html ).

My benchmarks are systems of www.kursalarm.de (german), www.futurestruth.com , www.striker.com and www.collective2.com .

Are there other websites with transparent system results useable for benchmarks?

Thanks,
zentrader

zentrader22 said:

@all,

I'm developer of mechanical trading systems (currently a german Dax Index system http://www.zentrader.de/html/trading_systems.html ).

My benchmarks are systems of www.kursalarm.de (german), www.futurestruth.com , www.striker.com and www.collective2.com .

Are there other websites with transparent system results useable for benchmarks?

Thanks,
zentrader

Click to expand...

I don`t consider C2 results to be transparent or useable. Having traded quite some systems there I found my results never to mirror the published ones (aside from the question of poor fills due to lack of liquidity in some underlyings). Some results seemed to be tweaked quite a bit.

The most important question -not answered on your site either- : How long has the system been forward tested? How sound was the backtesting methodology?

As you seem to be a specialist regarding testing, this is astonishing even more.

Regards

Hittfeld

Hittfeld said:

I don`t consider C2 results to be transparent or useable. Having traded quite some systems there I found my results never to mirror the published ones (aside from the question of poor fills due to lack of liquidity in some underlyings). Some results seemed to be tweaked quite a bit.

Click to expand...

@Hittfeld,

You are right. I also can't use C2 as I've reported in my performance comparison paper on page 9 ( http://www.zentrader.de/perfCompare_e.pdf ). Even the information on Striker isn't enough. So remain kursalarm (german) and futurestruth (usa). This is the reason, because I am still searching benchmarks or other websites supporting such comparisons.

The most important question -not answered on your site either- : How long has the system been forward tested? How sound was the backtesting methodology?

Click to expand...

Forward testing? For example I've tested my system also using a monte carlo simulation stress test. So not only the 1,512 historical trades generate the system report, but also the simulation result of additional 151.2 million trades (!).
http://www.zentrader.de/zenTSPerformance_e.pdf (page 12)

bye,
zentrader

Hi

Just looking at your system results.

You advertise a > 100% per year return (start with 20,000 end with about 220,000 after 8 years)

However, looking at the equity curve you seem to be compounding your results which means the annual returns are actually only about 35% per year.

Am I correct and if I am is this simply an error on your part born of incorrect analysis or are you trying to mislead us?!

Zen Trading System (Dax Index)...

sr100m said:

Hi

Just looking at your system results.

You advertise a > 100% per year return (start with 20,000 end with about 220,000 after 8 years)

However, looking at the equity curve you seem to be compounding your results which means the annual returns are actually only about 35% per year.

Am I correct and if I am is this simply an error on your part born of incorrect analysis or are you trying to mislead us?!

Click to expand...

@sr100m,

the example is based on a CFD-Trading with a leverage of 10 CFD contracts and a trading capital of 20,000 Euro.

On average the system won (2000-2009) > 24,000 Euro / year, so the average performance for every year is > 120% (always based on the unchanged trading capital of 20,000 Euro, which is sufficient for is scenario of system trading concerning the DDs and the transaction fees).

bye,
Volker

sr100m said:

Hi

Just looking at your system results.

You advertise a > 100% per year return (start with 20,000 end with about 220,000 after 8 years)

However, looking at the equity curve you seem to be compounding your results which means the annual returns are actually only about 35% per year.

Am I correct and if I am is this simply an error on your part born of incorrect analysis or are you trying to mislead us?!

Click to expand...

you made a mistake regarding the compounding...

100% a year doesn't transform 20k into 220k after 8 years... 100% a year in 8 years is actually 2.56 Million ... big difference eh?:cheesy:

Ah! I see now. You are using a fixed number of contracts on every trade. I tend to think in terms of fixed risk position sizing and was assuming that you were increasing the number of contracts as account value went up. (as you quote on your site “We do not see the things, how they are, but as we are.”)

My comment above may have questioned your integrity. I apologise.

Keep it simple - use the some index of your asset class as the benchmark :smart:

zentrader22 said:

Forward testing? For example I've tested my system also using a monte carlo simulation stress test. So not only the 1,512 historical trades generate the system report, but also the simulation result of additional 151.2 million trades (!).
http://www.zentrader.de/zenTSPerformance_e.pdf (page 12)

bye,
zentrader

Click to expand...

No, he means forward testing in the real market, which is probably the most important test - but you already knew that, right?

At the very least, you have a perfect opportunity to use a 'realtime' replay of the period from the end of your backtest to now.