Completely different results from same expert - page 3

Some tips as I have been fighting the wretched inconsistency of results in the tester as well but am now able to get consistent results. I was getting inconsistent results with no changes but simply pressing the "Start" test button after completion of an earlier run. There appears to be more than just the "Spread" at work. The following is a bit of a pain but I get consistent results.

1) Download the historical data and redownload until you perform at least one recalculation.

2) In the navigator, delete your demo account. This will disconnect your session and keep your results consistent when restarting MT4 (just don't login or recreate the account).

3) Check the spread for your symbol, using simple script with mode_spread= MarketInfo(Symbol(),MODE_SPREAD); Print("MODE_SPREAD=",mode_spread). If your broker uses fixed spreads, you may not be impacted by spread changes. I just like to ensure the spread is a reasonable value for the symbol pair. If not, I can adjust my results slightly to reflect a reasonable spread or I can reconnect, check the spread, and disconnect again.

4) Run your tests. As long as you do not reconnect/relogin your results should be consistent.

Good luck.

I tried this procedure today. What I found was I needed to be connected to a server in order to import data or to run a backtest. Without being connected, neither would work. So I'm not sure how this (reasonable sounding) idea of working offline can be made to work.

Another more specific discovery (which doesn't come close to explaining my observations) is that the EURUSD data I downloaded recently had a huge gap in it (over a week), despite being described as "high quality". I have informed the provider.

@pianoman59, I'm not quite sure what you mean by your first suggestion. I import data that I have previously downloaded from an independent data provider (and decompressed). What are you suggesting repeating?

If you need reliable data: http://eareview.net/tick-data should be your first choice. It's a pain to get it working, but afterwards you have quality live data

OK, thanks, I'm using data back before 2009 at the moment so I should be OK.

I presume you are doing that to keep the recent data untainted for later analysis when you have something that looks really good? Even then, though, I am a little skeptical of how useful your results will be. My impression is that EURUSD has been very different in recent years to the entire period before. Major economic factors may be the root cause why it has had a greater tendency to trend and (at least it seems to me) more profitable to trade as a result. I have seen many examples when backtesting, of systems failing to work at all until the last couple of years and then doing great.

At the moment I'm mainly debugging and testing my code, so I just need decent data without gaps and mismatches.

I take it you mean this data ?

I've checked, and the data was indeed replaced later on August 1st with one without the gap.

That true ! My programmer once told me that sometime lagging happen a lot in demo account make it instable 

SO it best that if you test result is good, try to test it on real account with small among first before use it for real