Questionable Strategy Tester Results.

Quoting wackena

Disliked

Attached are several backtest reports with astonishing reults over time ranges from 6 weeks to 5 years. The most astonishing is the 5 years gbpusd backtest.The EA uses money management to set lot size. On first long term backtest, it stopped in the middle due to excessive lot size. I contacted IBFX and was advised that they limit lot sizes to 50 for Mini and 100 for Standard accounts. I set max lot size to 90 and retested. On $1,000 deposit, end profit is over $17,000,000.

I tested on IBFX, FXDD and NF with similar results. I also test "every Tick" and "Control Point" models and again similar results.

Many distrust backtest results and this may be why. On 2006-09-19, I started EA on $5,000 demo on the 4 major pairs. So far, the account is down $453.

Any comments would be appreciated.

Wackena

Ignored

Quoting philmcgrew

Disliked

Wackena,

A couple thoughts come to mind:

Did you include commissions, spreads, and slippage into your testing? In my experience that can account for up to 30% of your equity.

You can not run a 5 year test and then declare a system dead after 2 days. You simply don't have enough data to tell one way or the other.

While I didn't study every curve in detail, it looks to me like your strategy did better a couple years back than it does now. It may have already been fading a bit.

Your strategy was trading 100 lots in the end. I am assuming that you are trading less than that. I would suggest that you eliminate the position sizing algorithm and test a single contract. This will give you an idea of where the system is working and where it is not. Position sizing can do a wonderful job of helping you through poor performance but it's not useful in doing a raw evaluation of a strategy.

When done properly backtesting can tell you a lot about your strategy and the possibility of future performance.

Best wishes,
Phil

Ignored

Quoting wackena

Disliked

Phil,
Thanks for the comments. The slippage is set to 3 in the EA. I didn't know how to set commissions and spreads in the Strategy Tester module. Please advise.

I agree on snap judgements on demo performance. Attached is Excel file that shows that over time there are good periods and losing periods. Even last month of Sept. has loss. click the Chart tab in the excel file.

I have tested fixed lost size and win/loss ratio is the same.

Again thanks,

Wackena

Ignored

Quoting witchazel

Disliked

another thing to note, is that until the modeling quality says 90% your test is not accurate. the longer the timeline of you test the better its going to look because there is less and less data. this is especially true on interday trades.

you might look to importing accurate data into mt4

Ignored

Quoting wackena

Disliked

Attached are several backtest reports with astonishing reults over time ranges from 6 weeks to 5 years. The most astonishing is the 5 years gbpusd backtest.The EA uses money management to set lot size. On first long term backtest, it stopped in the middle due to excessive lot size. I contacted IBFX and was advised that they limit lot sizes to 50 for Mini and 100 for Standard accounts. I set max lot size to 90 and retested. On $1,000 deposit, end profit is over $17,000,000.

I tested on IBFX, FXDD and NF with similar results. I also test "every Tick" and "Control Point" models and again similar results.

Many distrust backtest results and this may be why. On 2006-09-19, I started EA on $5,000 demo on the 4 major pairs. So far, the account is down $453.

Any comments would be appreciated.

Wackena

Ignored