|
/ | Forum |
  |
BarrowBoy
2008.10.10 14:01
|
|
c0d3
2008.10.10 21:54
BarrowBoy wrote >>
c0d3 > Can you please expand on this, the details of why there is a difference between demo server and real server. > What exactly do you mean when you say that the real market feeds are lot more ragged than any demo server. Simply hitch up an EA on a live account and the same EA on a demo account of that broker. The EAs will have different numbers of trades and different success ratios. EA's that enter mid-bar will show very significant differences, but even those that trade on 'last closed bar' will show variance The main differences are that:- - Response times vary more on a live server due to varying load - Spread doesnt vary on a demo server - The feed to the demo servers comes off the live servers not off the banking system itself - so is 'smoothed' a bit .. You can argue the contributing factors but the fact is that they are significantly different. An EA can succeed on a demo account yet fail on a live one :( Just look at an M1 chart on a live account and compare alongside the demo chart for OHLC differnces, gaps & general... spikiness :eek: -BB- So this basically means, that the backtesting/forward testing on a demo is garbage data? I know that backtesting is gargabe, but forward testing, i dont know if i should Laugh out Loud, or cry as i scratch my head. |
|
c0d3
2008.10.11 03:25
BarrowBoy wrote >>
c0d3 > Can you please expand on this, the details of why there is a difference between demo server and real server. > What exactly do you mean when you say that the real market feeds are lot more ragged than any demo server. Simply hitch up an EA on a live account and the same EA on a demo account of that broker. The EAs will have different numbers of trades and different success ratios. EA's that enter mid-bar will show very significant differences, but even those that trade on 'last closed bar' will show variance The main differences are that:- - Response times vary more on a live server due to varying load - Spread doesnt vary on a demo server - The feed to the demo servers comes off the live servers not off the banking system itself - so is 'smoothed' a bit .. You can argue the contributing factors but the fact is that they are significantly different. An EA can succeed on a demo account yet fail on a live one :( Just look at an M1 chart on a live account and compare alongside the demo chart for OHLC differnces, gaps & general... spikiness :eek: -BB- Basically you can use backtesting/forward-testing on a demo-account to make sure that your complete mechanical trading system is properly executed, and just not monitor profit/loss as this data is useless. Also, if i'm going to go live, I should pick a broker that supports microlots, and test the EA with a mini account on microlots. |
|
BarrowBoy
2008.10.11 22:15
I'm not saying forward testing on demo accounts is useless, far from it - but everyone should realise that it is not 100%, particularly on EA's that open mid-bar. This is not nuclear physics, so not so many absolutes - which it sounds like you are after :) Let us be aware of a certain lack of... consistency but not to be beaten down by it! Never forget we should have a traders head on for planning the strategy, and put our coding (scientific) chip back in for writing the EA. Many outstanding coders dont trade so well, and all traders I've had contact with are hopeless at defining the scope or spec for an EA... Ha - got a bit OT :) Back on message - if you are picking a broker for your first live EA, certainly go for the smallest lotsize available. Make sure they support MT, and have the account enabled for EA's (most are EA-disabled by default) Be aware of the brokers policy on spreads, commissions, rollover (aka swap) policy, etc And good luck -BB- |
|
KenL
2008.10.17 22:20
BarrowBoy, I registered on PipCop and noticed you were online last night. Its a great site and I'm very interested in their testing. Have any robots survived their testing yet? All I could find were the current tests and the failures. |
