MetaTrader 4 Build 600 with Updated MQL4 Language and Market of Applications Released - page 42

Donald R Isbell Jr
224
Donald R Isbell Jr  
For practical purposes, nothing in <metatrader>/experts/ is seen or used by build 600+. It might as well not exist since it only causes confusion. That's my two cents.
Matthew Todorovski
12900
Matthew Todorovski  
...MetaQuotes should: a) require all vendors to submit 99% accurate test results of their EA b) prevent any vendor selling their EA if they cannot do so. I am so sick of seeing 90% or even 25% test results - how anyone can even consider this is adequate to sell their product is beyond belief!
Ian Venner
2397
Ian Venner  
disbellj:

Is there a manual that can walk us through and show us what changes, including examples? This would really help us developers. Then a separate manual showing the indeed new stuff. Because we not only have one hurdle but two. We have the hurdle of making sure our old code works with the new build. Then we have an additional larger hurdle that will take longer time of learning new functionalities from MQL5. Let's not kid ourselves: MQL4 above 509 is an entirely different language from MQL4 build 509 and before. A much more complex language than MQL4 is. And for developers like me who have ONLY CODED in MQl4, we don't know what classes and the other new terms mean. We don't necessarily code other languages or for Microsoft.

Kindest regards,

Don


The docs have been updated, I think the F1 help in metaeditor has the most recently updated information, there is a section about the new version and compatability. I'm not sure it is the last word on all of that though
Simon Gniadkowski
17241
Simon Gniadkowski  
bluepanther:
...MetaQuotes should: a) require all vendors to submit 99% accurate test results of their EA b) prevent any vendor selling their EA if they cannot do so. I am so sick of seeing 90% or even 25% test results - how anyone can even consider this is adequate to sell their product is beyond belief!
I take it you are talking about modelling quality ? how do you propose to get better than 90% using MT4 alone ? how do you get better than 25% on M1 ?
Ian Venner
2397
Ian Venner  
bluepanther:
...MetaQuotes should: a) require all vendors to submit 99% accurate test results of their EA b) prevent any vendor selling their EA if they cannot do so. I am so sick of seeing 90% or even 25% test results - how anyone can even consider this is adequate to sell their product is beyond belief!


It is not beyond belief when you understand what those values mean.

To summarize, the modelling quality of a backtest is calculated based on how many of the bar states used were derived from real prices. Obviously on a 1 minute chart backtest, only the HLOC of each bar can be considered real prices, the rest of the ticks are artificially modelled by the strategy tester. On higher timeframe backtests, HLOC from the lower timeframes are included so there are a lot more "real" prices used per bar in the backtest. Therfore modelling quality of tests on higher timeframe charts get a higher modelling quality value.

graziani
2198
graziani  
RaptorUK:
I take it you are talking about modelling quality ? how do you propose to get better than 90% using MT4 alone ? how do you get better than 25% on M1 ?


99% on M1 and higher TFs is achievable with tick-data backtest using ... ah, sorry, no advertising. ;)

Ian Venner
2397
Ian Venner  
graziani:


99% on M1 and higher TFs is achievable with tick-data backtest using ... ah, sorry, no advertising. ;)

It is but is it really neccessary ? if you run a backtest on 1 minute chart data, strategy tester models all the ticks that would have been neccessary to create that 1 minute bar. It might not do that in the exact same order that they would have been recieved on live trading, but really when you think about it, is tick modelling vs tick data going to make that much difference to the results ? Even with high frequency scalping it is debatable. I think the way modelling quality is evaluated is a little unfortunate as it will always give testing on 1 minute charts that 25% value and the impression of a very poor quality test when really it probably deserves a higher value.
Simon Gniadkowski
17241
Simon Gniadkowski  
graziani:


99% on M1 and higher TFs is achievable with tick-data backtest using ... ah, sorry, no advertising. ;)

Can't do tick based Strategy Tester runs using MT4 alone. Please re-read my question
Simon Gniadkowski
17241
Simon Gniadkowski  
SDC:
It is but is it really neccessary ? if you run a backtest on 1 minute chart data, strategy tester models all the ticks that would have been neccessary to create that 1 minute bar. It might not do that in the exact same order that they would have been recieved on live trading, but really when you think about it, is tick modelling vs tick data going to make that much difference to the results ? Even with high frequency scalping it is debatable.
If your strategy relies on the real action between real ticks then yes it is necessary, synthetics ticks won't cut it. Why shouldn't Users have the option to use real tick data if they want ? even if MetaQuotes think it's crazy or unnecessary, why shouldn't they facilitate it using the users own tick data ?
graziani
2198
graziani  
RaptorUK:
Can't do tick based Strategy Tester runs using MT4 alone. Please re-read my question


yes you can. tickdata .fxt is generated once and reused on every backtest.