I see many systems and EAs described here, but it seems almost none has included an equity curve or performance sheet from a backtest. I wonder if people really trade untested systems? How can you know that it is profitable without a backtest over at least 4 years? An in-sample backtest is no guarantee for profit, but a missing backtest seems to me a certain recipe for loss.
1 Like
I dont think, that 4 years backtest is neccessary, but at least one year would be nice. But i do agree with your point. People are not posting backtest results probably because most of the system arent algos and are only manually traded.
1 Like
I think 1 year is not enough. This is a backtest of a system posted here, Hucks Trend Catcher 3:
It was only 5 minutes to write a script of this system and backtest it, and then you see that it is an unprofitable system. But when you backtested it only one year in 2011, it would look profitable. Only over 4 years you can see that it loses most time.
Problem is that even when you trade a system only manually, you can not know if itās profitable without a backtest. I would not manually trade something without backtest.
I believe the reason why we dont see a lot of backtested results on here is that people dont have the needed software skills to pull it off. If I am able to master the C properly, I will start a thread where I write algos for the popular mechanical systems on here and do the backtest. Atleast one system a week should be interesting
I ask the same question everyday.
True.
The other answers to the reason why backtesting is hardly done :
-Traders havenāt fully settled down on their trade logic as a trader. This is not a system issue, but a trader issue.
-Depends on your perception and belief system regarding the market. If you think the market is random, then you might think backtesting is unnecessary. If you think there are certain patterns in the market and will continue to do so in future, then you might take a little effort and time to do so.
-Insufficent research regarding the market (very impt point IMO). Backtesting is only a small part of the research.
-Even if the trader knows programming + has good trade logic that actually works, thinking of a correct and impartial model and pseudo-code can be a big hurdle. This is especially true with complex and deep trading systems. Without this, the traderās logic cannot be translated into actual working code. In this stage, potential flaws of the traderās logic can be exposed! (which is actually a good thing.
-Experienced traders who are so successful to the point, that they canāt be bothered to learn how to do backtesting at their stage of the game (either because computers and software are not available to them during their time or they feel programming is difficult even they did tried). So to them, a backtest can mean e.g. a weekās work trading on a smaller sized account, then manually testing further from 6 months to a year. Their vast experience of knowing vast market conditions makes up of the lack of backtesting (can both a good and bad thing).
-Even if one does backtesting, one should not mix money management into in. To mix money management into it would blur the definition of āsystemā. Heck, some traders even think MM is part of a trading system; it is not as :
1-the issue of MM is always personal
2-you can have a bad system, but good MM make up for the losses. This is dangerous as it gives the trader a misconception that the system is good.
3-you can have a good system, but with poor MM. Then this is no longer a system issue, but a trader issue. The other side of the coin will be questioned : is the system really that good? When the real problem is the trader!
Therefore backtests should be objectively measured by pips (objective), not equity curves (subjective).
Here you go. My side of things after doing a lot of my own work regarding backtesting / systems design / programming / trading and price action research.
Well, I backtest visually as far as my platform allows, then proceed to forward testing.
Any suggestion for a software?
Problem is that you can visually backtest only a few months, and even this with up to 30% error rate. But for any significant info about whether your system will lose money and how much, youāll need a backtest over at least 4 years, and with 0% error rate.
There are many platforms that can be used for backtesting, but you have to learn how to describe your system in code - thereās a course about that somewhere on this forum. Search for āAutomated trading courseā.
1 Like
You can download historical data from the History Center on MT4.
Thatās clear, the problem is not the lack of historical data, the problem is that you can visually test only so much data before getting mad.
Free backtesting software can test 4 years in a few seconds. For visually testing 4 years, you need to examine 24,000 candles with 1-hour bars. If you need 5 seconds per candle, you have 250 days work with 8 hours per day and no weekends for visually testing your system. I think itās a little more efficient to spend 10 days with learning programming, plus the 5 seconds for the backtest. 
I assumed it was the from your statement about only having a few months data to test. If you mean you canāt backtest enough because of time constraints thatās another thing altogether and no-one said it was going to be easy
I would never feel comfortable with only automated backtesting of a system. If you can write it good enough to work going backwards then it will be good enough to work going forwards and I believe in money making bots slightly less than the easter bunny. Of course thatās just my opinion.
I believe you can perform the backtest there as well?
I run an EA on historical data but only so that it hides āfutureā bars and allows me to progress it manually.
Sounds great. Mind sharing ?
I certainly donāt but I got it from somewhere that Iām probably am not allowed to post a link to on here so Iāll PM you the details.
Thank you in advance
No problem, if you have any trouble just let me know and Iāll try and help if I can.
Iāve never seen a system perform moving forward as it had going back. Iāve seen countless backtesting results show astronomical returns, only for future results to be unprofitable. Perhaps the inverse is where the Holy Grail lies.
I am working on learning how to backtest my EA. I have done it visually by looking at the indicators but it is obvious that this method has limitation - huge ones. I have also done some backtesting on MT4 but it was very slowā¦ and there are a few basic things I am having troubles withā¦ I have searched the internet and other forums. Maybe someone here could offer some suggestions?
- the historical data! I download from MT4 history center but there are only 5-6 months of 1M data for some currency pairs!
Suggestions on where to get a longer time series? - I followed some instructions from a link off āthe Trade Machineā - first of all the link to the heaps of historical data there is out of orderā¦ They suggest to set up another MT4 to be used for backtesting offline. Is this the way it is done?
slipp3ryWhippit: This other link for backtesting sounds interestingā¦ Does it backtest EAs? are you still willing to pass it on ?
Thanks!
Tickstory Lite - download and install. Its totally invaluable.