What computer will let me open 40 to 50 demo accounts at the same time to test differents EAs? - page 2
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
I think a few percent of a mediocre CPU should be more than enough for each undemanding EA. But why do you need to do that? If they are your own EAs, can't you be a bit more selective before real-time testing? Suppose you were going to run them on 6 months of live data, you can get to the same situation (barring the manageable errors inherent in backtesting) by backtesting them all on the last 6 months of data. And a lot quicker. Even if you do run them all live, the results won't tell you a lot about how they will behave after that than if you backtested them on the last 6 months of data - apart from things that should not really happen, the main differences should be spread variation and slippage.
Hi, Elroch,
The reason I hope to have many demo and backtesting accounts is to save time. As you know, many EAs, free or commercial, can win short term, but many lose or wipe out accounts long term. Just testing 6 months may not show the real nature of the EAs.
Is it true that backtesting intentionally switches bid and ask prices to make the testing more profitable? If true, who switches it? mt4?
Thanks.
forexjim
Hi, RaptorUK,
Below is a copy of my computer processes. Could you advise why the freezing/crashing?
At the time when you took that screen grab your 17 MT4 terminals were using 63% of your CPU resource . . . and they are using 1.8 GB of your RAM, Internet Explorer is using almost 700 MB of RAM.
You need more RAM that is very clear . . . . if you want to run much more than this in future you will need more CPU resources. Your current CPU isn't bad, Quad core @2.4GHZ, to step up significantly from that you will be looking at an i7 CPU to go even further you would need to be looking at Dual CPU maybe Xeons and spending a lot of money.
One other thing about your current machine, as you are using a lot of CPU resources the CPU will be getting hot, it needs its cooling system to be working very well to avoid overheating, when was the last time you cleaned it inside ?
Hi, Elroch,
The reason I hope to have many demo and backtesting accounts is to save time. As you know, many EAs, free or commercial, can win short term, but many lose or wipe out accounts long term. Just testing 6 months may not show the real nature of the EAs.
Is it true that backtesting intentionally switches bid and ask prices to make the testing more profitable? If true, who switches it? mt4?
Thanks.
forexjim
It is definitely not true that the backtester switches bid and ask prices. 6 months was (of course) just an example period, chosen because I could not imagine anyone forward testing a huge pile of EAs for longer than that without either going live or giving up.
The MT4 backtester is rather crude in that it only uses bid data, but combines this with the current spread for the market, on the grounds that it will have been similar for the entire past (not always a valid assumption). So basically, if the bid was XX at a time in the past, and your expert executes a buy, it will buy at (XX+spread) where "spread" is the current spread.
If you are really intending to test 40-50 of other people's experts (free ones or bought?), the only proviso to my comment about backtesting is that you should only use data from after the date the expert has been created (unless you created it yourself, when you can use any data that you have totally ignored when you were creating it). But the point that if you want to test them for a year, you could wait until you have a year of virgin data (i.e. data since they have been written) and backtest them. Or 10 years. Or whatever. If the experts are expected to get tiny average profits (a few pips a trade), or if they are prone to enter on news spikes, the difference between live testing and backtesting (over the same period) will be greater, but do you really want to use such experts? Back testing on virgin data (with an allowance for slippage and variation in spreads) is a pretty good approximation to forward testing real time with a demo account (and usually thousands of times quicker). A good way to make the spread more realistic is to do backtesting offline and use a utility to set the spread to whatever you want. A couple have been posted in these forums in the past, and should still be there somewhere.
Another important point is that if you test your 50 EAs for several years and find a few of them are profitable and one or two have moderate results (say a few times the drawdown), then statistically your results won't tell you that any of the EAs were profitable for any reason other than luck. If you get much better results from one of your EAs (say profits that are 10 times the drawdown), you can be pretty sure it is not just blind luck, but you cannot infer that the results will continue to be good for the future. This happens a lot when an EA is designed based on recent market conditions (eg a Neural Network that is not retrained). It will be likely to do well for some time, but this performance will be likely to fall, and may eventually become unprofitable.
At the time when you took that screen grab your 17 MT4 terminals were using 63% of your CPU resource . . . and they are using 1.8 GB of your RAM, Internet Explorer is using almost 700 MB of RAM.
You need more RAM that is very clear . . . . if you want to run much more than this in future you will need more CPU resources. Your current CPU isn't bad, Quad core @2.4GHZ, to step up significantly from that you will be looking at an i7 CPU to go even further you would need to be looking at Dual CPU maybe Xeons and spending a lot of money.
One other thing about your current machine, as you are using a lot of CPU resources the CPU will be getting hot, it needs its cooling system to be working very well to avoid overheating, when was the last time you cleaned it inside ?
Hi, RaptorUK,
Many thanks for the analysis. Would it be simpliest to buy another two computers just like the one I have? Would the price of three computers be higher or lower than the Xeons computer?
True, I never knew I needed to open the computer to clean the cooling system. I will do it one of the weekends when the forex market is off.
Many thanks.
forexjim
It is definitely not true that the backtester switches bid and ask prices. 6 months was (of course) just an example period, chosen because I could not imagine anyone forward testing a huge pile of EAs for longer than that without either going live or giving up.
The MT4 backtester is rather crude in that it only uses bid data, but combines this with the current spread for the market, on the grounds that it will have been similar for the entire past (not always a valid assumption). So basically, if the bid was XX at a time in the past, and your expert executes a buy, it will buy at (XX+spread) where "spread" is the current spread.
If you are really intending to test 40-50 of other people's experts (free ones or bought?), the only proviso to my comment about backtesting is that you should only use data from after the date the expert has been created (unless you created it yourself, when you can use any data that you have totally ignored when you were creating it). But the point that if you want to test them for a year, you could wait until you have a year of virgin data (i.e. data since they have been written) and backtest them. Or 10 years. Or whatever. If the experts are expected to get tiny average profits (a few pips a trade), or if they are prone to enter on news spikes, the difference between live testing and backtesting (over the same period) will be greater, but do you really want to use such experts? Back testing on virgin data (with an allowance for slippage and variation in spreads) is a pretty good approximation to forward testing real time with a demo account (and usually thousands of times quicker). A good way to make the spread more realistic is to do backtesting offline and use a utility to set the spread to whatever you want. A couple have been posted in these forums in the past, and should still be there somewhere.
Another important point is that if you test your 50 EAs for several years and find a few of them are profitable and one or two have moderate results (say a few times the drawdown), then statistically your results won't tell you that any of the EAs were profitable for any reason other than luck. If you get much better results from one of your EAs (say profits that are 10 times the drawdown), you can be pretty sure it is not just blind luck, but you cannot infer that the results will continue to be good for the future. This happens a lot when an EA is designed based on recent market conditions (eg a Neural Network that is not retrained). It will be likely to do well for some time, but this performance will be likely to fall, and may eventually become unprofitable.
Hi, Elroch,
Many thanks for the detailed advice.
Honestly, I am not sure what info I can draw validly from backtesting, demo and live testing and what info cannot be drawn from each. Take, for the example, this result from a backtesting:
1) Modeling quality = 25%, does it mean nothing is valid?
2) Mismatched chart errors = 0, does it mean good?
3) Profit factor = 5.75, does it mean thie EA is profitable?
4) Maximal drawdown = 36018, does it mean if the initial balance is over 36018, there will not be a margin call?
5) Backesting results have not taken spreads into considerations?
Thanks for your advising.
forexjim
Hi, RaptorUK,
Many thanks for the analysis. Would it be simpliest to buy another two computers just like the one I have? Would the price of three computers be higher or lower than the Xeons computer?
True, I never knew I needed to open the computer to clean the cooling system. I will do it one of the weekends when the forex market is off.
Many thanks.
forexjim
Make sure you do it with the power off but the power cord still plugged in so the PC is earthed. Use a soft, clean paint brush and clean and vacuum out all the dust from the fans and the CPU heat sink/fan.
Regarding new computers, you will really need to look at local prices, I am in the UK, prices elsewhere are very different from here . . . you need more RAM for you r current PC, if it will take it.