Hi all,
When I test my EAs on MetaTrader 4, profits are OK during normal Forex trading hours (Sunday pm to Friday afternoon). When I test the exact same EAs on the weekends (same parameters, same start/end dates), profits are down to 1/10th of what I see if I run tests during the week. I’m using a demo acct on MB Trading.
How could this be? This is backtesting!
Robert
Check the spread. It has probably been widened to be twice or 3 times the normal week day spread. In Strategy Tester, the current spread is used instead of the expected spread.
This provoked a thought about a question I had asked before, but didn’t get an answer - -
Since I am unexperienced at ‘winning’ - I was wondering, if I create an EA -that when run over a 1 yr period shows about a 10% return;
my question(S) become - or my thought process is:
It’s a stupid EA -if it produces 10%, if “I” simply implement some better money management, or simply watch for price action to tell me it’s a no go, and I exit some trades early:
while maybe, if TP is almost hit, I move TP out some more while moving a SL tucked behind a little S+R.
Did I just UNDO the whole point of an EA?
Did I just Enhance it to win 20% vs. 10%?
I guess what I am asking , what is reasonable and just?
I guess part of it was too – the EA I used was suited to work when price is ranging, so if I notice price is trending - skip 5-8% (that’s a random number) of trades - then that 10% has to go up.
Thank you for the reply!
I’m not sure I understand you. So during the week, the current spread is used whenever you test? That would make sense, since the historical data doesn’t contain the spread. But then, on the weekends when the market is down, what’s the expected spread? Just some default number in MT4?
-Robert
Yes and no (I think!). I don’t see the problem with doing what you do when using an EA. But if you want to test a particular stragegy, then I’d leave it alone. If you mess with the trading, you’ll never know how the strategy would have worked if you didn’t.
Did I just Enhance it to win 20% vs. 10%?
In the future, you won’t know…since you might not act the same way when moving SLs, TPs, etc.
I guess what I am asking , what is reasonable and just?
I guess part of it was too – the EA I used was suited to work when price is ranging, so if I notice price is trending - skip 5-8% (that’s a random number) of trades - then that 10% has to go up.
Man, you’re loosing me here!
Don’t worry about it, MT4 is crappy when it comes to backtesting.
Sorry for being confusing.
I guess, what I am asking, if you created an EA, that generates XX% per year. lets say 10%,
Do you say to yourself - WOW - I’ve done it! now put it live!
OK - Great start, it produces 10%, lets put it live then work on tweaking it!
OR on the other side of positive
10% from an EA over 1 year - doesn’t mean much, Lets see if I can get it to 20% before I even THINK about going live.
I hope that clears it up a bit -
Your spread will be the last Friday tick spread which is usually 3x to 10x higher than normal. (Ex: EURUSD normally 1.2 - last weekend was 3.6 for my backtesting) - I use this higher spread to stress-test my EA’s. If they can make money with this superhigh spread applied to EVERY trade then it’ll be more profitable in live trading.
Hate to burst your bubble, but even 20% annually isn’t very good in back testing. There is a huge discrepancy between back, demo and real money testing. The reasons are spread widening, requotes and disconnects. I doubt if 20% in back testing would be break even in real trading. I don’t know what the magic number is and would not like to guess because there are other factors such as trade frequency and risk:reward that have to be considered.
PERFECT answer - all I am looking for is understanding what I’ve got . and that answer explains it very well. Thanks.