Somthing wrong? autom strat comes back with profit factor 2.20/2261 trades

[BataviaTrader]

I was wondering if I am probably overseeing something. I have developed an automated strategy for ES and tested it on historical data of about 3 years. It came back with a profit factor of 2.20 (2261 trades). To me, it is pretty good, if not too good to be true. Am I missing something maybe? I would say 3 years of data with this amount of trades should give a trader some edge right?

Anyone with a lot of knowledge about testing automated systems? for what it's worth, I use Sierra Chart.

Looking forward to your input.

Thank you and best regards
BT

 

[malaguti]

I was wondering if I am probably overseeing something. I have developed an automated strategy for ES and tested it on historical data of about 3 years. It came back with a profit factor of 2.20 (2261 trades). To me, it is pretty good, if not too good to be true. Am I missing something maybe? I would say 3 years of data with this amount of trades should give a trader some edge right?

Anyone with a lot of knowledge about testing automated systems? for what it's worth, I use Sierra Chart.

Looking forward to your input.

Thank you and best regards
BT

so now forward test it, that will tell whether it was too good to be true
good luck with that one

 

[Liquid validity]

Have you made an allowance for or included in the backtest:
1. comms
2. spread
3. slippage

Is it good quality tick data you are using?

If no to all of the above and its based on tight stops / targets,
its probably flawed.

If yes to all the above points, may be worth forward testing.

 

[BataviaTrader]

Have you made an allowance for or included in the backtest:
1. comms
2. spread
3. slippage

Is it good quality tick data you are using?

If no to all of the above and its based on tight stops / targets,
its probably flawed.

If yes to all the above points, may be worth forward testing.

I have included commissions, $4,02 for a round turn. Also, Target 2,25 points Stoploss 3,25 points (trailing), which is reasonable imho .

Data comes from IB, my broker.

If I may know, how would you include spread/slippage ?

BT

 

[Liquid validity]

I have included commissions, $4,02 for a round turn. Also, Target 2,25 points Stoploss 3,25 points (trailing), which is reasonable imho .

Data comes from IB, my broker.

If I may know, how would you include spread/slippage ?

BT

Sounds ok on the comms and data front then.
Backtesting with spread is platform dependent, Ninja doesn't currently do it.
I'm not sure, but I think Sierra does (you'll have to google that one).

As for slippage, most platforms include a setting for allowing average slippage per trade.
Say 1 tick.

Really though, with spread and slippage, only real way to be sure is live forward
testing with chump change.
Forward testing on sim will include spread, but slippage will not be included.

As comms, spread and slippage can kill something that otherwise appears
OK, its not something to avoid doing.
Its a pain for sure, with your trade frequency, I would forward test for
1 year (750 ish trades).

Thats just my personal view on forward testing, larger live sample size is safer.
Ultimately its your choice how soon you hit it with proper size.

Also look into monte carlo and bootstrap testing of backtest results
to see if its a statistical anomaly / curve fit etc.

 

[BataviaTrader]

Sounds ok on the comms and data front then.
Backtesting with spread is platform dependent, Ninja doesn't currently do it.
I'm not sure, but I think Sierra does (you'll have to google that one).

As for slippage, most platforms include a setting for allowing average slippage per trade.
Say 1 tick.

Really though, with spread and slippage, only real way to be sure is live forward
testing with chump change.
Forward testing on sim will include spread, but slippage will not be included.

As comms, spread and slippage can kill something that otherwise appears
OK, its not something to avoid doing.
Its a pain for sure, with your trade frequency, I would forward test for
1 year (750 ish trades).

Thats just my personal view on forward testing, larger live sample size is safer.
Ultimately its your choice how soon you hit it with proper size.

Also look into monte carlo and bootstrap testing of backtest results
to see if its a statistical anomaly / curve fit etc.

Hi Liquid validity,

Thanks for your respons. That was very valuable to me. I have not discovered yet the spread function in SC, but I will look futher into this.

For the slippage I would reckon 1 tick, so my max gross profit per trade would be 2 points. I am looking into trading 1 contract first, building sufficiënt equity to trade 2 contracts, and take it from there.

I have just started forward trading this week, so I let it run for a while. I also found that trading only during RTH has proven to come out with a higher profit factor. So I leave the aftermarket- hours out of my charts. Additional advantage is the required margin is lower for trading intraday.

How would a larger size be safer in your opinion? I would think doubling contracts is doubling the risk?

I have take a peek on monte carlo and bootstrap testing, which is a complete new phenomenon to me, looks interesting, but it's gonna take a scoop from my wallet I noticed :cheesy:


thanks and regards
BT

 

[Liquid validity]

How would a larger size be safer in your opinion? I would think doubling contracts is doubling the risk?

Yes correct.

I was referring to a larger sample size of live forwards tested trades.

As far as trade size goes, as its ES contracts you will be at $12.50/tick minimum
for live testing.
I would test on sim first for a while.

 

[numbertea]

While developing my trading system I went through a similar situation. If your trades happen during times of extreme volatility, as mine used to, on the ES, then projecting slippage amount is difficult for simulation if not impossible. Maybe for your system the trouble simulating ends there, but for me I also determined risk levels and funding based on entry price and profit expectancy so it made much more trouble than I could handle with any accuracy since I was placing many trades each day. I found that the best way was to spend some trading money on my education and trade 1 contract until profitable. Quickly I saw that simulation is great to test logic but not profitability of a system. Now I trade only a couple times a day and it suits me. I'm always trying for total automation but don't have it entirely yet. I trade ES and SPY depending on which I deem as the leading indicator for the moment. My system is in Cpp and I use Lightspeed.