My Partner and myself have been playing with developing a trading robot and we have had mixed results.
Have it running on EUR/USD, AUD/USD and it is taking on average 5-8 trades per day per currency pair.
After talking to a few people they are telling me that the robot is taking too many trades.
How many trades should we be aiming at?
Another question is Who is the best platform to run a robot on as we are not happy with the current platform we are using as occassionally it forgets tp place stoploss or take profit orders.
You may have to reconsider what you are attempting to do. Ultimately your chance of success is very small. Robots that trade mechanically (ie. no discretion) have very little chance of being better than 50% successful and of course the spread will make them unprofitable. Every system - even manual ones needs an edge and that usually means years and years of trading and finding and refining a strategy with an edge.
Plus you have challenges from back testing. Back testing is almost useless except to find code flaws and to gain familiarity with the strategy. The first problem is data quality. Unless you find tick data, testing is futile. Probably some of the problems you have mentioned can be explained by poor data quality, but that is only a guess. The other problem is that back testing is nothing like real time with slippage and execution delays.
.
I’m not sure what is meant by the comment too many trades. If the robot was making money, how could it take too many trades? Probably someone means that they as a trader would avoid some of them. But of course, a robot can only follow the rules in the program and that is not the way humans trade. Something you might consider is the time of day. I know from experience that the AUD / USD can be traded more successfully during the Asian session and the E/U should be avoided then.
If you are doing it for the learning experience, then keep on working on it. Otherwise, don’t waste much more time on it.
With the many different options for backtesting, you’ll find that one is able to import their own tick data, eliminating the issues with incomplete/choppy data. Slippage and sometimes even latency (although I have only heard about this, and never seen it in action myself) can be set as well prior to testing. Obviously, if you’re relying on a backtest from MetaTrader 4, well I wouldn’t take the results very seriously at all.
Clark, I have a few small quibbles with your statements. The Back Tester in MT4 is perfectly suited for finding code flaws even using broker data, but yes to use it to predict real time trading results is not possible. The other thing it is good for is for learning about coding and trading - kind of a set of “training wheels”.
The kind of testing you have proposed is not for the novice. It requires somebody with lots of trading experience to know how to set up test cases and to properly evaluate the results. I have the impression that the OP is still in the learning stage and what you are suggesting is a future target to shoot for.
The algorithm MT4 uses for backtesting as well as optimisation (exhaustive and genetic, although MT4 isn’t multithreading) will work just fine, it’s the data sample that one is using (assuming one is using the default data) that I find troubling. I would argue that if one were to attain and convert rich tick data into a MT4-compatible data, the Strategy Tester would be very suitable for most individuals (although there would still be no way of accounting for spreads, commissions, slippage, latency, etc). And yes, I would agree it is a great platform to learn on - hell, I still use it occasionally because of its simplicity.
I was always in the mentality of not doing any testing or any “work” before I understood how to do it properly right from the start. So I tended to avoid doing any strategy testing before I was able to fully understand my options and requirements. So I guess that is why I would recommend one to look at other options before jumping right in.
Maybe I’m so new at it, I don’t really know. Or maybe the tip is new to MT4?? So, I’m just passing on a suggestion that seems to work.
I download the 1 min data for a pair, then I download again. It asks for an evaluation of all timeframes or something. I accept that. And it recovers more ticks. I routinely do that 3 times. By then it has recovered enough ticks per minute to be useful.
Are you switching to MT5 already? Or what is more powerful, less simple, that you use more than MT4?
Ah, I see. I looked again. So, I guess what I’m doing with multiple downloads is getting a sufficient number of ticks in each minute. There is a column for, what I have to think of as (tick) Volume. The volume of ticks per/min.
I suppose it could mean the currency volume in 1 lot increments? Don’t know. It’s just that when I run a back test, I can only go in the 1 minute min. visual view. But, it sure looks like it’s ticking inside the bar.
The counter in the upper left corner of the visual view is quite granular. The 5th place digit is practically a blur it changes so much. Oh well, so much to learn. So easy to get side tracked in learning only.
The “ticks” that you are seeing are interpolated based on the given minute bar’s volume. So if a bar has a volume of say 254, MT4 will randomly generate 254 ticks that will meet the OHLC and V criteria of the bar. Those aren’t the actual ticks produced historically.
Yes I am new to MT4 but have been trading for a little while.
I have found by developing a robot (which I use to think was a waste of time in the past) it has helped me improve my own live trading.
The data has been collected by live demo trading on pepperstone and the last three weeks with a live account which is infront.
It has been making on good day’s 170-200 pips profit at the end of the day’s trading but sometimes be down 60 - 90 pips before it starts to come back and end up infront.
On bad day’s it might end up 20-30 down, been trying different things to eliminate the draw down but have been getting frustrated with execution of the system.
Anyway will keep looking at it and push on and thanks again for the feed back from everyone.
Good explanation on the trading simulation results. Backtesting is good reference but you have to run your strategy forward testing and on live trading to see the results. (Start demo trading first.)