M1 modeling quality and backtests
I guess it mostly depends on your EA, what accuracy you would need.
I am one of the few around, who thinks, that using data compression with dataloss, and trying to decompress from that lossy compressed "file" can be dangerous sometimes. This is what really the method does, when you make different period bars from price data (this is compression with data loss), and you simulate ticks from the bar data.
I experienced, that by using tick data (using data without loosing anything from it), I got drastically better accuracy, with fast EAs. There can be situations, where the limits of tick simulation only turns out on the long run. I mean, that if there is a certain type of bar, where ticks simulated from M1 data are built up in a way, to confuse your strategy (eg. multiple entries/exits on a bar), you won't even spot it if you watch it for a week or two, bit on the long run, these situations repeat themselves, thus you get false results at the end. But again, it can be EA dependent.
Check these topics as well it MT4 forums:
"MQL4: Made $ 258 663.50 from $3000 using Problematic Backtesting by just buying at open (Fractal ZigZag)"
"MQL4: Indicative vs "Real" quote data..."
I am one of the few around, who thinks, that using data compression with dataloss, and trying to decompress from that lossy compressed "file" can be dangerous sometimes. This is what really the method does, when you make different period bars from price data (this is compression with data loss), and you simulate ticks from the bar data.
I experienced, that by using tick data (using data without loosing anything from it), I got drastically better accuracy, with fast EAs. There can be situations, where the limits of tick simulation only turns out on the long run. I mean, that if there is a certain type of bar, where ticks simulated from M1 data are built up in a way, to confuse your strategy (eg. multiple entries/exits on a bar), you won't even spot it if you watch it for a week or two, bit on the long run, these situations repeat themselves, thus you get false results at the end. But again, it can be EA dependent.
Check these topics as well it MT4 forums:
"MQL4: Made $ 258 663.50 from $3000 using Problematic Backtesting by just buying at open (Fractal ZigZag)"
"MQL4: Indicative vs "Real" quote data..."
I don't think your friends are all that MT4 savvy. If they were, they would know that 25% is the highest MQ that can be achieved in MT4. This does not mean that your backtest is inaccurate. It simply means thats MT4 can't look inside the bar on 1 min data since 1 min is the smallest time incriment available. To prove this to you, change your backtest model to Control Points rather than Every Tick and backtest on the 1 min. Your results should be the same.
I don't think your friends are all that MT4 savvy. If they were, they would know that 25% is the highest MQ that can be achieved in MT4.
Haha, wow! You've just made a complete fool of yourself, mister. I suggest you hit some forums on this one. 90+ quality is nothing special. So again, before you try to go "guru" on us all, I'd suggest you double check your sources.
Now, just to make your life more interesting and try to kick-start your brain:
1. http://rapidshare.com/files/17631376/how-get-modelling-quality-90-when-testing-ea-mt4.pdf.html (taken from forex-tsd.com)
2. "25% is the highest MQ that can be achieved in MT4"? OK, a random example: "MQL4: RobotPowerM5 meta4V12" . Modelling quality 42.06%
3. "25% is the highest MQ that can be achieved in MT4"? Another proof-of-concept, taken from the russian Code Base section, "MQL4: Самообучающийся ЭКСПЕРТ(lsv)" :
Now you see the part I highlighted? It reads: "Modelling quality 88.18%". And this is not the maximum, 90+ is nothing special, one just needs to know how to do it. And contrary to what you're trying to claim here, higher quality DOES matter.
Regards,
D.
Diamond,
I meant that 90% was the highest MQ that can be achieved ON THE 1 MIN TIME FRAME. I thought that was a given.
Haha, wow! You've just made a complete fool of yourself, mister. I suggest you hit some forums on this one. 90+ quality is nothing special. So again, before you try to go "guru" on us all, I'd suggest you double check your sources.
Now, just to make your life more interesting and try to kick-start your brain:
1. http://rapidshare.com/files/17631376/how-get-modelling-quality-90-when-testing-ea-mt4.pdf.html (taken from forex-tsd.com)
2. "25% is the highest MQ that can be achieved in MT4"? OK, a random example: "MQL4: RobotPowerM5 meta4V12" . Modelling quality 42.06%
3. "25% is the highest MQ that can be achieved in MT4"? Another proof-of-concept, taken from the russian Code Base section, "MQL4: Самообучающийся ЭКСПЕРТ(lsv)" :
Now you see the part I highlighted? It reads: "Modelling quality 88.18%". And this is not the maximum, 90+ is nothing special, one just needs to know how to do it. And contrary to what you're trying to clain here, higher quality DOES matter.
Regards,
D.
I meant that 90% was the highest MQ that can be achieved ON THE 1 MIN TIME FRAME. I thought that was a given.
I don't think your friends are all that MT4 savvy. If they were, they would know that 25% is the highest MQ that can be achieved in MT4.
Haha, wow! You've just made a complete fool of yourself, mister. I suggest you hit some forums on this one. 90+ quality is nothing special. So again, before you try to go "guru" on us all, I'd suggest you double check your sources.
Now, just to make your life more interesting and try to kick-start your brain:
1. http://rapidshare.com/files/17631376/how-get-modelling-quality-90-when-testing-ea-mt4.pdf.html (taken from forex-tsd.com)
2. "25% is the highest MQ that can be achieved in MT4"? OK, a random example: "MQL4: RobotPowerM5 meta4V12" . Modelling quality 42.06%
3. "25% is the highest MQ that can be achieved in MT4"? Another proof-of-concept, taken from the russian Code Base section, "MQL4: Самообучающийся ЭКСПЕРТ(lsv)" :
Now you see the part I highlighted? It reads: "Modelling quality 88.18%". And this is not the maximum, 90+ is nothing special, one just needs to know how to do it. And contrary to what you're trying to clain here, higher quality DOES matter.
Regards,
D.
Diamond,
I meant that 90% was the highest MQ that can be achieved ON THE 1 MIN TIME FRAME. I thought that was a given.
I meant that 90% was the highest MQ that can be achieved ON THE 1 MIN TIME FRAME. I thought that was a given.
Whatever might be evident for you does not necessarily mean that it's so for other people. "25% is the highest MQ that can be achieved in MT4" << that doesn't say anything whatsoever about a timeframe, it refers to MT4 in general. Anyhow, terribly sorry for any confusion then.
Diamond,
I meant that 90% was the highest MQ that can be achieved ON THE 1 MIN TIME FRAME. I thought that was a given.
I meant that 90% was the highest MQ that can be achieved ON THE 1 MIN TIME FRAME. I thought that was a given.
Whatever might be evident for you does not necessarily mean that it's so for other people. "25% is the highest MQ that can be achieved in MT4" << that doesn't say anything whatsoever about a timeframe, it refers to MT4 in general. Anyhow, terribly sorry for any confusion then.
If you read the rest of my post, I was clearly referring to the 1 min time frame, which was the entire topic of this thread to begin with: 1 MIN TIME FRAME.
But I guess some people need things spelled out for them.
You must be know a lot about the backtester if you believe that percent. It can be manually set in the fxt header, so it doesn't mean anything. What does, is only the quality of data...
The link you gave here only makes the backtester easier to use, and more secure, but the data quality (M1 data) is the same, and the tick generating algorithm is the same.
90% would be true, if 4.4 tick happened average per minute, since only open/close/high/low of the M1 bar is accurate, the rest is not. Have you seen situations, where a bit more than 4 ticks happen in a minute? Well that 1 minutes are all inaccurate. That is everything, but 90%.
The link you gave here only makes the backtester easier to use, and more secure, but the data quality (M1 data) is the same, and the tick generating algorithm is the same.
90% would be true, if 4.4 tick happened average per minute, since only open/close/high/low of the M1 bar is accurate, the rest is not. Have you seen situations, where a bit more than 4 ticks happen in a minute? Well that 1 minutes are all inaccurate. That is everything, but 90%.
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
But I am puzzled by the fact that when I put it to forward testing and it seems to be doing almost exactly as the backtest. I thought that historical data modeling quality could play a major role. But apparently, for the M1 timeframe, the modeling quality is not a big deal because it is the lowest timeframe possible (except if we consider tick data).
However, I am still a bit skeptical about this simply because all my online friends -also MT4 savvy programmers- are urging me to enhance the modeling quality to 90% for the M1 frame. I'm not sure I need to do that. Do you think there is any need to have such high modeling quality for this smallest timeframe?
Of course, I also read this topic:"MQL4: One-Minute Data Modelling Quality Rating" , which explains that 25% modeling qualit for M1 is the furthest I could -or need to- get.
Thanks for your clarification in advance.