How accurate is back test? (For Past Result)

[richardmarcus1509]

Hi,

Can anyone tell me how accurate is back test for past result (Not future result). I know bast performance do not guarantee future result.

So, here i am asking about a past result.

I made a back test for EA (Oct 2020) and the result is in negative but the developer of EA said me back test provide false result and for him October 2020 was profitable with same EA which given a negative result with back test.

Can anyone clear my doubt?

Backtesting is accurate. I do not see it would show a negative result without the indicator producing negative result in reality.

 

[Andy89]

Is back test reliable and accurate to check past results? Is it possible if used different historical data or counted transaction costs in different ways?

 

[Pharaoh]

Backtesting is accurate. I do not see it would show a negative result without the indicator producing negative result in reality.

Half right. If an EA or other system fails backtesting, the chances of profit in a live account are virtually zero.

On the other hand, a profitable test (even using the highest quality data) is not an assurance of what will happen when faced with real world market conditions. Slippage and requotes don't happen in backtests. Market conditions in backtests are fixed. The real market changes over time.

So, the real use of a back test is not to determine profitable vs not-profitable. The real use is to filter out "definitely not profitable" from "has the potential to make profit."

The next step after a backtest on high quality data shows good potential is to try the system out on a VERY small live account. Make sure to say your goodbyes to the money you deposit into the test account. You are sending it into a dangerous environment and it may not come back.

 

[Adolphus]

I would say it is not that accurate because backtest only shows how good an EA could be. The backtests mostly show if a trader can worth risking a small amount of capital on a forward test.

 

[Amina El Mehdi]

As all i heard about it it`s 50:50. Can be useful but not absolutely accurate. Don`t know from personal experience.

 

[Vinagre]

I would say that it is incomplete.

You have to go further on the analysis not only backtesting based on the past data.

Use a monte carlos simulation too. Also test your returns with a econometric solution/technique named GARCH.

Try also to test your results using a Fat Tail Distribution. (See Nassim Taleb book).

Good luck!

 

[ele020]

There is no point to doubt the back testing report of an EA. Go with the EA only if your back testing is sound enough.

 

[Amina El Mehdi]

Go with the EA only if your back testing is sound enough.

And even then it`s not 100% sure.

 

[Donovan.CC]

I do not think it is accurate all the time. Sometimes the strategies that perform well in the past may fail to do well in the current trade.

 

[fxoday]

Backtest result maybe will be depending on historical data, if use accurate historical data it may be the result also accurate, but because in backtest not involved slippage, disconnect problem, etc, good result backtests no guarantee will give the same result in real trading.