Back Test Cumulated Profit

[pistolpete]

Ok,
I am using the Divergence pattern that isitpossible put on the strategies section of the NT forum. I am running FAS a triple barreled bull financial ETF. I am running a .015 stop loss pct and a .01 trailing stop pct on 100 shares. I am starting it at 930am to 4pm. The test is starting at 6/1/2009 to 8/10/2009 Let me know if you need anything else.
Thanks

[NinjaTrader_JoshP]

pistolpete,

I ran that strategy with those settings on a 1min series for FAS on my end and my results do not look like yours at all.

[pistolpete]

I am running 5 minutes.

[NinjaTrader_JoshP]

Ran on 5mins.

I suggest you guys try reinstalling your NinjaTraders to see if that helps.

1. Backup any thing you guys want to preserve (File -> Utilities -> Backup)
2. Uninstall NT6.5
3. Delete My Documents\NinjaTrader 6.5 (You will lose everything from chart templates, data, indicators, strategies, etc.)
4. Clear web browser cache
5. Reinstall NT6.5

[NinjaTrader_JoshP]

Please be sure you guys are comparing currency vs currency. Total Net Profit is always given to you in currency. To switch Cumulative Profit to currency you need to do it in the dropdown menu in the top toolbar of the Strategy Analyzer.

[KrisS]

It may be a while before I encounter the scenario in backtesting again, but here is an example from strategy performance. It is clear that the accumulated profit is a simple running total of the column - incorrectly adding the percent that results from each portion of the execution. To calculate this properly it needs to respect the relative percent of shares in the execution. As an alternative, you could calculate for the total order rather than the multiple executions in the order (probably more appropriate).

[KrisS]

In addition to the example I highlight below, I think I may have determined why my sim account is growing strangely, and why my other statistics are a mess. Much seems to relate to the accumulated profit calc error and there may be other issues as well - see attached. My sim account grew $30K in one day, but the real net was much less. My profit factor shows over 6, but is about 1.68 in reality.

To reproduce the accumulated profit issue, all you have to do is sim trade an account size (really a position size) large enough to result in 2-3 or more separate executions per order.

This seems to be an important issue to fix in order for me (and others) to have any confidence in strategy testing.

[NinjaTrader_JoshP]

They use this equation.

PRODUCT(1 + profit / entry price) of all trades - 1

where profit is defined as (exit price – entry price) for long trades and as (entry price – exit price) for short trades, quantity is defined as the number of contracts traded, and point value is defined as the monetary conversion of each point (e.g. 100 for currency pairs).

[KrisS]

Josh,

In my post #22 example below, do you believe the accumulated profit of 245% in two days is correct?

I appreciate the formula, but it appears to be calculating by execution rather than per order (sum of executions per order).

I guess I am looking for some sort of acknowledgement that there are some opportunities for improvement in upcoming releases. ;-)

[pistolpete]

I agree.
Because of this flaw I don't trust any number that the system generates that is calculated using more then 100 shares. If the formula is that simple why not fix it?

[NinjaTrader_JoshP]

The formula is accurate by executions. It has to be by execution because executions can fill at different prices causing differences in profit/loss values for those executions.

Thank you for the suggestions though. We will add it to the list of future considerations.

[pistolpete]

If I am trading 1000 shares of XOM and I have a 2% return on the trade but since NT breaks everything up then multiplies the results as individuals it is documented as 10% return. Can you not see the problem with this? You're formula might be mathematically correct but that does not mean it logically makes sense. I think NT developers need to start thinking as traders not as programmers. There are a number of flaws within this system that need to be addressed.

[NinjaTrader_JoshP]

Suggest you not use percentage mode and just use currency.

[forerx]

Suggest you not use percentage mode and just use currency.
And then try again...

[KrisS]

The formula is accurate by executions. It has to be by execution because executions can fill at different prices causing differences in profit/loss values for those executions.

As of today I have a reported acccumulated profit of 272%. You can see in my earlier examples that the execution approch leads to an incorrect accumulated profit %, incorrect profit factor, incorrect monthly projected profit, etc.

I guess my point is that you can't do this by execution as it presently is. You have to calculate the total net of the executions for each order, and use the order round-turn net for the cumulative profit. I do this in Excel for real-world values of varying quantity/price executions on each end of the round turn and the cumulative profit is correct.

Never before have I been so eager to make software work and pay well for it, but there are so many issues. After thousands of hours of CPU time in backtesting, it is only the hope of fixes in NT7 that keeps me from throwing in the towel. We appreciate your support.