Is back-testing EA's fair enough?

[5of7]

If you really want to develop your own EA's, I have to say that MT4 is probably the worst platform possible to do it on.

It may be okay for executing the strategy, but for developing and backtesting, it's the pits.

There are much better tools out there to develop on. Then when you have something you feel confident in, port it to MQL to run on MT4.

-Trader 5of7 of TheCollectiveFX-

 

[5of7]

The order of development should always be:

1. Backtest.

2. Forward test in demo.

3. Forward test with 1 micro.

4. Setup a live account trading 1 micro fixed, with enough capital to support the largest historic drawdown, plus some.

5. Only increase trade size with the strategy's own profits. It should earn the right to trade larger size.

If you increase trade size arbitrarily, you are really just throwing dice.









-Trader 5of7 of TheCollectiveFX-

 

[Pharaoh]

Good order to put things in. I'd add to test with nanolots if you have a broker that allows one to trade that small.

 

[MikKi]

It seems that You do not trust EAs. I have seen many kind of 'holy Grails' too but I believe that good EAs can be developed.
I like the above mentioned order. There are brokers which provide cent-based accounts, it might be a better choice than nano-lots.

 

[Steven_Smith]

It is very much dangerous to believe just in backtesting... for example the system maybe already optimized for the definite time period and on backtesting chart it will be working just fine, but in real life - it can easily lose your money.

 

[MikKi]

It is very much dangerous to believe just in backtesting... for example the system maybe already optimized for the definite time period and on backtesting chart it will be working just fine, but in real life - it can easily lose your money.

Recently I found 2 robots that were extremely curve fitted (unfortunately I don't remember the names). They used the most extreme method: when backtesting they used a custom data file with the exact times when to open and close trades They were quite profitable in backtest but had no real trading logic... The 2nd one was a bit more sophisticated: it opened MT4's tick data file and preread the tick data so that it was a perfect holy grail scalper: every trade was successful and didn't last longer than a minute

So yes, be very careful when backtesting. And use your head: strange results (all trades last the same time, have the same profit, no losses, etc-etc...) usually mean strange trading logic. If you have access to the source code of the robot it is always a good idea to take a look at it even if You don't understand the whole logic.

 

[TromS]

For what time period those systems were optimized?

 

[MikKi]

For what time period those systems were optimized?

If You're asking me then it doesn't matter. The whole point of the systems I mentioned was *not* to have any trading logic. The only thing those 'experts' do is knowing the past and knowing the exact entry and exit points where trades were good.

If You know the past You can create a perfectly performing strategy for that time.

 

[Stanly]

backtesting is a very doubtful thing. there are a lot of tricks, which can be done by EA developer to make you see excellent backtesting results