My first fully automated system: are these statistics sound enough for you?

O

ourson

Hello everybody.

As a newbie, I do not actively participate in this forum as I still have much to learn before being in a position to give some advice.

I am writing today in hope of getting some constructive feedback on the statistics of my first automated trading system. It has been backtested (no optimization performed) over the last 10 years. Since I plan to start with a small account (USD 10,000), the statistics are based on 1 future contract only. The money management is included in the automation and as you will see in attached graph, I am limiting my daily loss at around USD 400.

I still need to find some good filters to improve some key statistics but still I want to know what is your first feeling as for the robustness of this system.

Thanks a lot for your feedback.

Will.
 

Attachments

Last one... Thanks once again.

Performance All Trades Long Trades Short Trades
Total Net Profit 131094 31634 99460
Gross Profit 835462 393206 442256
Gross Loss -704368 -361572 -342796
Commission 29836 14936 14900
Profit Factor 1.19 1.09 1.29
Cumulated Profit 567.28% 55.96% 327.86%
Max. Drawdown -12.79% -10.11% -13.10%
Sharpe Ratio 0.48 0.18 0.52

Start Date 1/1/1999
End Date 5/15/2009

Total # of Trades 7459 3734 3725
Percent Profitable 45.88% 45.02% 46.74%
# of Winning Trades 3422 1681 1741
# of Losing Trades 4037 2053 1984

Average Trade 0.03% 0.01% 0.04%
Average Winning Trade 0.37% 0.36% 0.39%
Average Losing Trade -0.27% -0.27% -0.26%
Ratio avg. Win / avg. Loss 1.39 1.33 1.46

Max. conseq. Winners 12 13 11
Max. conseq. Losers 16 16 15
Largest Winning Trade 7.26% 5.00% 7.26%
Largest Losing Trade -0.72% -0.55% -0.72%

# of Trades per Day 1.97 0.99 0.98
Avg. Time in Market 113.8 min 105.4 min 122.1 min
Avg. Bars in Trade 22.6 21 24.3
Profit per Month 1.54% 0.36% 1.18%
Max. Time to Recover 702.01 days 701.85 days 955.01 days

Average MAE 0.20% 0.20% 0.19%
Average MFE 0.36% 0.34% 0.37%
Average ETD 0.33% 0.33% 0.33%
 

Attachments

Your'e not going to get much feedback on the "robustness" of your entire system because all you are really showing us is the cumulative profit performance report. However, based on the equity graph, it would appear that you have averaged @50% per annum. Beats most fund managers by a mile.

Also, there seems to be most of your accelerated profits between 1/07 and 5/09 which leads to a question.

Have you been compounding your equity in trades to account for this acceleration?

The equity graph is rising during both bullish and bearish periods, and that in itself is a sign of an adaptable and robust system.

I'm sure you realize that backtested hypothetical results should make you hopeful , not giddy.
 
To stevegee58:

You are right, slippage is key and I will try to integrate this.

To Big AAPL:
No there is no compounding but 2004-2006 were poor years for this system.


What type of elements should I bring here for experienced traders be in a position to better assert the "robustness"?
 
Quote from ourson:
...
More...


my suggestion is forward test it for 3 months.

if the net result for the period is similar to the historical quarterly return, then you might start trading it live with small size. then, after one year if the profitability stays there you can scale up accordingly.

i wouldnt spend time retouching the strategy as you said but rather forward testing it. the more you fine tune it, the more the chances are it will blow up soon. that is, the risk of failure is proportional to the amount of complexity/curve fitting built into it.

let me clear with what my understanding of curve fitting -- in essence the act of backtesting a strategy is pretty much a curve fitting action. that is, by backtesting a strategy, you are trying to get reassurance that is a valid forward looking plan. if the backtest fails to show that, you'll move on to another strategy until the backtest says what you want to hear. beware that, as long as you backtest, you are CURVE FITTING a strategy.

curve fitting is extremely dangerous because it leads to false premises and assumptions regarding the overall risks involved. for instance a complex arbitrage strategy involving multiple instruments that has been yielding 50% per annum with 5% DD's many years is more prone to disaster than a simple instrument mechanical trading system that has been returning the same 50% per annum but with DD of 50%. why? because the former has been yielding abnormal risk adjusted returns and thus it is likely to return to its statistical mean, while the latter has been yielding close to normal returns and thus is less likely to suffer a nasty and unexpected DD.
 
Quote from asap:

my suggestion is forward test it for 3 months.

if the net result for the period is similar to the historical quarterly return, then you might start trading it live with small size. then, after one year if the profitability stays there you can scale up accordingly.

i wouldnt spend time retouching the strategy as you said but rather forward testing it. the more you fine tune it, the more the chances are it will blow up soon. that is, the risk of failure is proportional to the amount of complexity/curve fitting built into it.

let me clear with what my understanding of curve fitting -- in essence the act of backtesting a strategy is pretty much a curve fitting action. that is, by backtesting a strategy, you are trying to get reassurance that is a valid forward looking plan. if the backtest fails to show that, you'll move on to another strategy until the backtest says what you want to hear. beware that, as long as you backtest, you are CURVE FITTING a strategy.

curve fitting is extremely dangerous because it leads to false premises and assumptions regarding the overall risks involved. for instance a complex arbitrage strategy involving multiple instruments that has been yielding 50% per annum with 5% DD's many years is more prone to disaster than a simple instrument mechanical trading system that has been returning the same 50% per annum but with DD of 50%. why? because the former has been yielding abnormal risk adjusted returns and thus it is likely to return to its statistical mean, while the latter has been yielding close to normal returns and thus is less likely to suffer a nasty and unexpected DD.
More...

Bingo!

What (s)he said.
 
As you mentioned, your system bled for 2 years straight before coming back with a vengeance. Are you prepared to sit on your hands & not pull the plug the next time this happens? I think you should consider waiting until you have 20K & trade an additional non correlated instrument.
 
You must log in or register to reply here.
Back
Top