How do you achieve a qualitative leap in market analysis? There is an option: - page 2

New comment
 
My memory expectation is precisely over 20. I have not got to the Expert Advisor yet. 1000 points and 800 points - just round numbers to clearly represent the ratio (I just did not want to focus on it initially). I think the normalization algorithm for History Center is probably very simple. Of course, first we must analyze a lot of different quotes to get an idea of possible nuances. I haven't done that yet.
 
Mathemat:

Renat, is it possible to solve the mystery of the quote normalisation algorithm just a little bit? I'm not asking you to give me the sources of the quotes. It's just that I still don't understand what normalisation is.

You probably think that "normalisation" is some complex process which modifies quotes. In reality, it is just filtering quotes without changing their price levels, but with correction of tick volumes. In other words, outright fraudulent quotes are simply thrown out.

Unfortunately, traders are looking at the quotes from 2000 and completely forgetting that spreads and Instant Execution were big before, which resulted in a noisier flow. We have not artificially cut the range of price changes, as it would inevitably lead to very bad results. We limited ourselves to filtering (although at first we wanted to normalise it, but then we gave up).

Trading in 2000 and 2006 is a huge difference for over-sensitive experts. As sophisticated traders in 2000 kept repeating "choose serious targets so as not to depend on a few pips of execution error", so they do now. But beginners cannot pass up the captivating opportunity to adjust to every tick, especially when there is a tester at hand who can calculate everything easily.
 
getch:
My memory expectation is precisely over 20. I have not got to the Expert Advisor yet. 1000 points and 800 points - just round numbers to clearly represent the ratio (I just did not want to focus on it initially). I think the normalization algorithm for History Center is probably very simple. Of course, first we must analyze a lot of different quotes to get an idea of possible nuances. I haven't done that yet.
If you had published a full report at once, instead of verbal excerpts, the answers would have been clearer. But you haven't.
 
Renat, does MT4 MultiTerminal plan to support several simultaneous accounts from different brokerage companies? If you have written about the history of creation of History Center, it is clear that the real-time approach cannot be implemented, it may be done with a lag of a minute or two. I think that the simultaneous estimation of quotes of several brokerage companies will give more objective view, than from one. And this objectivity cannot be obtained using an anti-filter on quotes of one brokerage company. More on the report later.
 

Symbol GBPUSD (Great Britain Pound vs US Dollar)
Period 1 Minute (M1) 2004.07.01 00:02 - 2006.09.30 00:00 (2004.07.01 - 2006.09.30)
Model All ticks (based on all smallest available periods with fractal interpolation of each tick)
Parameters Lots=0.1;
Bars in history 774050 Simulated ticks 3824199 Modeling quality 25.00%
Initial deposit 10000.00
Net profit 14098.27 Total profit 20119.03 Total loss -6020.76
Profitability 3.34 Expectation of winning 30.38
Absolute drawdown 49.61 Maximum drawdown 859.73 (3.44%) Relative drawdown 3.44% (859.73)
Total trades 464 Short positions (% win) 232 (65.52%) Long positions (% win) 232 (68.53%)
Profitable trades (% of all) 311 (67.03%) Loss trades (% of all) 153 (32.97%)
Largest profitable trade 296.44 losing deal -638.91
Average profitable deal 64.69 losing trade -39.35
Maximum continuous wins (profit) 17 (760.70) Continuous losses (loss) 5 (-152.70)
Maximum Continuous Profit (number of wins) 760.70 (17) Continuous loss (number of losses) -859.73 (3)
Average continuous winnings 3 Continuous loss 1

Files:
strategy05tester.zip  24 kb
 
getch:

Bars in history 774050 Ticks simulated 3824199 Simulation quality 25.00%

Yes, expectation is 30 pips, not pips. But the modeling quality is not close to 90%. What objectivity can we talk about when estimating the strategy?

Try to test it on a larger timeframe. Instructions on how to reach 90% in testing are available in the forum thread mentioned in the comments to Hendrik (see The Championship) that helped me much. The link to the forum is http://www.forexfactory.com/forexforum/showthread.php?t=8820. But you have to register there to download the ModellingQuality.pdf file . Also, here in the Articles you can find very good advice on how to use the tester.
 
The quality is maximum. Read here: 'Assessing the quality of modelling of minute data'.
 
getch:
The quality is maximum. Read here: 'Assessing the quality of modelling of minute data'.
This is the maximum possible quality for a given test TF, but it's not maximum in principle. Well, if you are happy with it, then it's up to you...
 
To make the number 90% appear, you can do the same as in this article (link below). It won't make any real difference.
http://www.lightpatch.com/forex/mt_yahoo/Obtain%20modeling%20quality%20of%2090%25%20when%20testing%20EAs. pdf.

There is an option to work with real tick data described here:
http://www.lightpatch.com/forex/mt_yahoo/Backtesting%20with%2099%20percent%20modelling%20V1. doc

Tick data:
http://rapidshare.com/files/1333387/EURUSD1_0.zip
 
Well yes, the first article is ModellingQuality.pdf. And the tick data... why do I need someone else's? And I'm not going to test strategies where entry and exit are often on the same minute candle. Extrapolating the results of testing such strategies is suicide, even if it is 99%.
12345678910
New comment