Complex Backtesting

[Doji3333]

I'm developing a strategy that includes a main entry trigger and multiple additional triggers/exits.

For example, I may want to buy 1 or 2 contracts, then add an additional contract or remove an additional contract as determined by my code. The additional contracts may continue to fire triggers over and over depending on the trend.

Then, possibly reverse short and start the entire process over again.

At the end of the day, everything needs to go FLAT.

I'm running into an issue where the backtest report is rather confusing.

Can this type of development be done with EFS and backtesting?

I'm using Strategy.ALL when I'm trying to force the system to go flat before trying to initiate a new main trigger or EOD close function.

Are there any tricks to doing this and are the backtest results correct? I can post screen shots and other info, but not the code yet.

[Doji3333]

Some images

OK, I see there have been lots of views of this thread, but no replies. So I thought I would try to post some examples.

First, I've developed my code to only trade between 630 and 1305 on the ES. I force a GO FLAT at 1306 with a "strategy" call to close "Strategy.ALL". So there should never be a situation where any trade carries over from one day to another.

Now, I know the strategy backtesting system can be picky with regards to entering and exiting at specific prices. The prices have to be available on the bar you are trying to execute the order on otherwise the order is not executed. Thus, I'm thinking my problems are with how/where I'm trying to execute these orders and not with the backtest system.

Any, these charts show what I'm trying to accomplish and the strange backtesting results I'm getting. I think I'm going to reduce my charts to 2 days and see if I can track down the problems.

B

[Doji3333]

BT results

[Doji3333]

I figured it out..

First, there appears to be an error in the Strategy.DEFAULT return. I've run numerous tests with this setting set at ONE and the return appears to be -2. Don't know why?? So I quit using that function and created my own contract settings for the tests.

Second, the problem was exactly as I reported earlier - trying to fire orders at wrong prices and at prices that don't exist on the bar. Thus, garbage in, garbage out.

Now on to more testing.

B

[Doji3333]

Here are the end results of my hard work over the last few days..

This is a 2.5 week test of a 3 minute ES chart with my new system (targeting only 1 pt per trigger).

[Doji3333]

Here is the exact same tested data with a 2.75pt target. Notice the number of trades fell to half that of the original 1pt test. Accuracy stayed about the same and net results were about the same.

Now to turn it into an automated trading system... weeeee.

[jgr]

Doji3333

I enjoyed your posts on this. I am starting to develop my own strategy too, not as sophisticated as yours, and your experiences are helpful.

For instance the Strategy.DEFAULT return. I hope the developers fix that although I don't anticipate using it, at least yet.

You have been doing this a while no doubt and know that if you start winning consistently with your strategy, the people whose money you are taking will change what they are doing to not loose so much and then you will have to change yours too. It becomes a bit of a game.