Do banks control modeling quality?
LL
> Do banks control modeling quality?
No - it is an MT4 function
> Is there any filter (used by the banks or MT4 platform) that makes the modeling quality go from 90% to 25% on certain dates?
No
> I used M5 frame and the EA was pretty successful on the old data (25% quality). But it stops being successful and begins to lose when I test it closer to modern data.
> Long research, but I found the dates of switch. In February 2009, the switch was Dec 18/2008 (about two months). In April 2009, the switch was Feb 19/2009 (about two months). > In June 2009, the switch was April 9/2009 (about two months again).
The dates you describe correspond with general pattern changes in the market
Your EA is not successful in all market patterns, as indeed most are not
Try and find a filter to improve or suppress trades during adverse market action
Variations in modelling quality happen any time for a given installation of MT, but this is clouding the bigger picture of variations in EA behaviour in relation to the market
Good Luck
-BB-
If you were able to fill all the holes in the data, making it all 90%, you'd eliminate modeling quality and find that your EA still had a shelf life or was over-fitted to certain data.
CB
LL
> Do banks control modeling quality?
No - it is an MT4 function
> Is there any filter (used by the banks or MT4 platform) that makes the modeling quality go from 90% to 25% on certain dates?
No
> I used M5 frame and the EA was pretty successful on the old data (25% quality). But it stops being successful and begins to lose when I test it closer to modern data.
> Long research, but I found the dates of switch. In February 2009, the switch was Dec 18/2008 (about two months). In April 2009, the switch was Feb 19/2009 (about two months). > In June 2009, the switch was April 9/2009 (about two months again).
The dates you describe correspond with general pattern changes in the market
Your EA is not successful in all market patterns, as indeed most are not
Try and find a filter to improve or suppress trades during adverse market action
Variations in modelling quality happen any time for a given installation of MT, but this is clouding the bigger picture of variations in EA behaviour in relation to the market
Good Luck
-BB-
Thank you, BarrowBoy!
I truly appreciate the answer.
However, if it was that simple, I would not even start asking at the Forum.
The fact is that first I was thinking like you - "this is the market conditions." That was the first thought - natural thought. I checked everything - ups and downs, waves, etc. There was not much of a difference in the market. Then all of a sudden, in two months, the EA easily went through that day (Dec 18) and continues to win - steadily. But there appeared another break-down point (Feb 19).
This time I was patient and methodical. I was checking it every day (same dates, same settings), and sure enough, in a couple of months the EA went through that stumbling point too. Now it was winning all over till the April 9, and was losing after.
So market conditions do not matter.
For the final check, I changed the back testing dates around April 9 (starting earlier/closer and ending closer/further), and the quality of modeling started to change, as well. From 25% it went up. That made it obvious - it was the switch - done by the human being or equipment.
So you say it is not the bank.
The MT4 function does it.
The next question is the following: if it is the MT4 function, can it be turned off and on?
What does this function do to the data? How does this function operate to make the precise data (90% modeling quality) look imprecise (25% modeling quality). Does it work only on historic data? Or can it do the same with the flow of data on the move?
Can you or anybody else make a filter to help the newbee like me?
Thanks again.
Warmly,
Lifelover
If you were able to fill all the holes in the data, making it all 90%, you'd eliminate modeling quality and find that your EA still had a shelf life or was over-fitted to certain data.
CB
Hi Cloudbreaker!
Thank you.
It may sound strange, but I like the 25% quality. At least I see the EA winning.
:)
Lifelover
Hi Cloudbreaker!
Thank you.
It may sound strange, but I like the 25% quality. At least I see the EA winning.
:)
Lifelover
Hmm. The words head and sand come to mind. I do hope you're joking.
No matter whether modeling quality is 25% or 90% it is still "modeling". In other words ticks have been invented where no ticks exist. There may be absolutely no similarity between the modeled ticks and what happened in reality. Even when the modeling has been based upon M1 data.
Now, depending upon how your EA makes its decisions you may find that:
- A certain EA may test fine on 25% and work similarly, making money with live ticks.
- Another EA may test fine on 90% and work completely differently, losing money with live ticks.
The key to deciding what (if any) modeling quality will produce accurate results, is to completely understand the impact, based upon an understanding of exactly how the particular EA works.
CB
Hmm. The words head and sand come to mind. I do hope you're joking.
No matter whether modeling quality is 25% or 90% it is still "modeling". In other words ticks have been invented where no ticks exist. There may be absolutely no similarity between the modeled ticks and what happened in reality. Even when the modeling has been based upon M1 data.
Now, depending upon how your EA makes its decisions you may find that:
- A certain EA may test fine on 25% and work similarly, making money with live ticks.
- Another EA may test fine on 90% and work completely differently, losing money with live ticks.
The key to deciding what (if any) modeling quality will produce accurate results, is to completely understand the impact, based upon an understanding of exactly how the particular EA works.
CB
Thanks, CB!
>In other words ticks have been invented where no ticks exist. There may be absolutely no similarity between the modeled ticks and what happened in reality. Even when the modeling has been based upon M1 data.
Wow! That is a nice explanation. It means that the phrase "Every tick (the most precise method.....)" in the model selection on the tester is misleading (and many of us, including me, fall for it!!!).
You are right. There is no tick information in the History. So ticks that we see on the tester are "fake" - or "invented," or not real - whatever we decide to call them.
>- A certain EA may test fine on 25% and work similarly, making money with live ticks.
- Another EA may test fine on 90% and work completely differently, losing money with live ticks.
Furthermore, this means that modeling (no matter how good it is) is still not good for the real market (live ticks). Right?
Still, the question stands: (whether I am joking :) ;) or not - there is a glimpse of truth in every joke) can there be an instant converter of market reality (live ticks) into the "fake modeling reality (fake ticks)" for the EA to make a right decision (because it wins in the modeled world).
Let me give you an analogy. You look like an aviation man on the logo. So the analogy is from aviation. A pilot may not see the enemy planes (too far) or the landing strip (too cloudy), but the monitor image (a computer modeling the reality in real time) allows the pilot to make a decision for launching a missile or for landing. So the pilot (or autopilot in modern planes) can ACT in real life, i.e. land the plane, while using the "fake model."
Thanks again, CloudBreaker, for breaking my clouds of ignorance in the Forex World.
LL
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
Do banks control modeling quality?
Or is it the MT4 finction?
I used M5 frame and the EA was pretty successful on the old data (25% quality).
But it stops being successful and begins to lose when I test it closer to modern data.
Long research, but I found the dates of switch. In February 2009, the switch was Dec 18/2008 (about two months). In April 2009, the switch was Feb 19/2009 (about two months). In June 2009, the switch was April 9/2009 (about two months again).
Is there any filter (used by the banks or MT4 platform) that makes the modeling quality go from 90% to 25% on certain dates?
Can anybody explain what is happening?
Thanks