Pour utiliser le chat, veuillez vous connecter.
Retour aux contacts

Backtest in the past is good, in the past is bad. What's happening?

Aug 21 2016 at 07:06
posts 32
Hi guys!

I have an EA and it works without none indicator or fundamental decision. The EA is purely statistical (NO martingale). And it seems work very very well!

See the backtest since 2013 until today:

Initial deposit = 10 000 USD
Fixed lot = 1.00

But, I have a problem, this graphic is common in all pair of currencies. The EA always make too much money, but, in the last three months, it doesn’t gain money. I don’t why! The number of positions in the last three months is equal the number in the last 3 years. In others words, this EA lost money with spread.

Why!?!? Would be the last thee mounts the real performance of this EA? Would be that database is with bug? Would be that is possible or impossible reproduce the past performance in the present moment?

Someone here has had the same problem?
Ivan (StoneHeart)
Aug 21 2016 at 21:23
posts 131
One of the possible reasons is NEGATIVE SWAP.
 It is accumulated daily to any open trade, but You see this result only on close !
If You have many trades opened for more than 1 day
 It is normal when You have Positive Profit, to get Negative Swap, and final Negative result.
Watch more detailed to the orders closed by tester at the end of test.
 (comment in journal like this : 'Tester: order #xxxx is closed')

One-other reason is a bit complex for me to explain, I will try.
If Your account currency is USD , and You trade EURUSD. You Profit is in USD.
Many people use PIPS to calculate PROFIT , but if price made some bigger move (form 1.10 to 1.15)
the price of 1 pips at 1.10 is higher than price of 1 pips at 1.15 (USD is cheaper at 1.15)
This way, if you have 100 pips loss at 1.10 , and 100 pips profit at 1.15,
the result is 0 PIPS, but it is NEGATIVE in money.
Simply EA must calculate MONEY, not PIPS.
I hope You understand me.
Small GAIN, big WIN
Aug 22 2016 at 00:17
posts 325
n/a model quality. Means your backtest is likely broken, should always have 90% if you get the data from history, or 99/99.9 if you use one of the tick backtests.
Veuillez vous connecter pour commenter .