How far back is ideal for backtesting? | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Features of ideal Forex Broker? 11 replies

Ideal number of pairs 4 beginner 30 replies

IDEAL PRO is driving me nuts! Please help. 29 replies

[ask]ideal target pips around news 0 replies

Ideal S/L and T/P ? 1 reply

How far back is ideal for backtesting?

  • Poll

  • How long should a system be tested (constant profit yield) to let you have confidence
  •  
  •  
  •  
  •  
  •  
  •  
  • #1
  • First Post: Feb 12, 2011 5:24am Feb 12, 2011 5:24am
Just interested in doing a survey, There are certainly many systems out there and i am one of the many that actually tested many changed many systems and wasted a lot of time in system hopping until i finally found a system that is suitable for me which i can build consistent profits with.

i wasted i think too much time finding systems testing them and learning them. So i am just curious if there is a manual backtest of 1 or 2 year to see the potential of the system first before trying the system will it be better for new traders?
  • #2
  • Feb 13, 2011 1:54pm Feb 13, 2011 1:54pm
I went back for 8 years on EURUSD with 800 deals and still not sure, someone said needed much more.

I think it depends upon the type of strategy, the more trades per day and shorter time scale approach would probably need less time. For long term trending strategies for instance would expect need minimum of 100 in a range of market conditions to ensure less chance of surprises.
  • #3
  • Feb 13, 2011 2:47pm Feb 13, 2011 2:47pm
As far back as possible and make sure you backtest your trades in different trade environment to compare, i.e. test the trades in both active markets and slow markets (summer and holiday months) and make sure you backtest in the market that matches your everyday trading environment.

And also keep in mind Forex market has changed a great deal from a few years ago to now. Before, the range and possible moves were a lot bigger; 100+pip candles were very common. Now, you would be lucky to see 10+pip candles.

Good luck!
Make your losses in demo. Earn your profits live.
  • #4
  • Feb 14, 2011 1:50am Feb 14, 2011 1:50am
As above mentioned, test in all types of market conditions. Then you will see how the system does in the period of drawdown and whether it is acceptable to your uncle point. the biggest DD is always in the future
  • #5
  • Feb 14, 2011 5:44am Feb 14, 2011 5:44am
Quoting ash234
As above mentioned, test in all types of market conditions. Then you will see how the system does in the period of drawdown and whether it is acceptable to your uncle point. the biggest DD is always in the future
6 month is enough for me. Real forward trading, Not backtesting, your demo trading. MONEY! This is all about money. If somebody would put in the table half of the year confirmed results that would be enough for me, to make a decision.
  • #6
  • Feb 17, 2011 7:10am Feb 17, 2011 7:10am
But for me be careful of too much backtesting the strategy, it will make you confuse, because you will realise that your strategy not really work in certain market condition, market only move in 3 ways which is downtrend, uptrend and sideway, take 3 examples of each movement and tested your strategy on it, then you also need the forward test, especially know your own trading emotion, better take a longer time to built and test your strategy money management , because i think no matter how good your system, bad risk management and bad trading emotion will lead to bad result also
  • #7
  • Feb 19, 2011 3:48am Feb 19, 2011 3:48am
May I know how do you guys backtest your system manually? For I have my own system but I don't know how to make it into an EA. And I don't really trust on EA for backtesting my system. So, I would like to know if there is any system available that can be used to backtest my system manually? Preferably its free.

Thank you in advance.

Good trading
  • #8
  • Feb 20, 2011 1:54pm Feb 20, 2011 1:54pm
Quoting jokes_erm
May I know how do you guys backtest your system manually? For I have my own system but I don't know how to make it into an EA. And I don't really trust on EA for backtesting my system. So, I would like to know if there is any system available that can be used to backtest my system manually? Preferably its free.

Thank you in advance.

Good trading
yes, it's called microsoft office excel, you can buy it in any software store
  • #9
  • Feb 20, 2011 9:18pm Feb 20, 2011 9:18pm
Ideally you should be looking at 10,000+ trades. That's the number of times a system needs to be tested in order for it to be statistically significant. However that doesn't mean you have to do it that many times to ensure its stability. I would suggest however that you look at it in terms of number of trades instead of number of years. Makes much more sense that way.
  • #10
  • Feb 21, 2011 5:06am Feb 21, 2011 5:06am
Quoting Custos
yes, it's called microsoft office excel, you can buy it in any software store
Or use Open Office Calc, it is free, although I am sure they will welcome a contribution from your eventual winnings to help keep the project going.
  • #11
  • Feb 21, 2011 5:13am Feb 21, 2011 5:13am
Quoting karma1
Ideally you should be looking at 10,000+ trades. That's the number of times a system needs to be tested in order for it to be statistically significant. However that doesn't mean you have to do it that many times to ensure its stability. I would suggest however that you look at it in terms of number of trades instead of number of years. Makes much more sense that way.
Interested where the 10000 number comes from. As far as I know the exact number depends upon the type of information being collected and how variable it is. Often there a lot of underlying assumptions in the calculations such as an acceptable error rate.

My guess is that you need enough to ensure that all market conditions have been adequately tested and all major eventualities examined. For instance if your system has any martingale element then you need to have enough deals to ensure that the worst case situation has happened several times.
  • #12
  • Mar 3, 2011 11:16am Mar 3, 2011 11:16am
isn't it depended on the time frame you're trading?

like min 6 months on 1H
or 2 months on 5 min TF?

I was also told 200 trades minimum to backtest before you proceed

just an opinion from a beginner...
  • #13
  • Mar 4, 2011 6:07am Mar 4, 2011 6:07am
Dont forget that the market is dynamic.........if you back test it, you assume it's constant for that amount of time.......

I rather put more weight-age on what works these months than finding fossil systems that stopped working.........
  • #14
  • Mar 4, 2011 6:26am Mar 4, 2011 6:26am
Quoting newborn1000
Dont forget that the market is dynamic.........if you back test it, you assume it's constant for that amount of time.......

I rather put more weight-age on what works these months than finding fossil systems that stopped working.........
I would be intrigued to see your research or academic articles on why the market is dynamic and in which instances. Or better yet, if you have some research on when the market is not constant, I would be glad if you could show me.

It's easy to just state something, but you need to back it up by thorough research, like in any other science.

Similar thing was when people believed the earth is flat, when in fact this has never been proven, until it was finally proven that the earth is roundish.
  • #15
  • Mar 4, 2011 6:37am Mar 4, 2011 6:37am
Quoting Custos
I would be intrigued to see your research or academic articles on why the market is dynamic and in which instances. Or better yet, if you have some research on when the market is not constant, I would be glad if you could show me.

It's easy to just state something, but you need to back it up by thorough research, like in any other science.

Similar thing was when people believed the earth is flat, when in fact this has never been proven, until it was finally proven that the earth is roundish.
I say it's dynamic because when enough ppl realize that a certain system works, it stops working, it changes...........If you really need prove, there is none..........

But experienced traders would know what I am referring to, pg 193 Market Wizards (The Ultimate trading system)..........

I guess either you dont do systems, or yours havent get obsolete yet.........
  • #16
  • Mar 4, 2011 7:47am Mar 4, 2011 7:47am
Quoting newborn1000
I say it's dynamic because when enough ppl realize that a certain system works, it stops working, it changes...........If you really need prove, there is none..........

But experienced traders would know what I am referring to, pg 193 Market Wizards (The Ultimate trading system)..........

I guess either you dont do systems, or yours havent get obsolete yet.........
there certainly is proof, if you do enough statistical analyses and base them on a good fundament.

Well, and yes, my systems haven't got obsolete yet, cause I tested them now on over 100 instruments and decades worth of data.

Certain aspects of the market never change.

Anyways, my point was and stays: first do enough quantitative research and base that research on proven scientific statistical methods, and only then you should state something.
  • #17
  • Mar 4, 2011 10:13am Mar 4, 2011 10:13am
Quoting newborn1000
Dont forget that the market is dynamic.........if you back test it, you assume it's constant for that amount of time.......
Testing surely does not necessarily imply that you are assuming the market dynamics are constant. If you get consistent returns then that is a demonstration that the strategy is invariant over the testing period. If not then it is a demonstration that it is suited to some conditions and not others. The latter is not a bad thing if you can anticipate the conditions and trade or not trade as appropriate.
  • #18
  • Mar 4, 2011 10:25am Mar 4, 2011 10:25am
Quoting Custos
Certain aspects of the market never change.
Nothing personal, but that was the point I was trying to make in this post http://www.forexfactory.com/showthre...46#post4436846 , but your reply at that time seem to imply that the market can change enough to invalidate all such behaviour. Slightly different issue I know as that discussion was about time frames so that may have clouded the issue.

The real question is what are the aspects that never change and can thus hopefully be used to produce consistent profits.
  • #19
  • Mar 4, 2011 10:33am Mar 4, 2011 10:33am
Quoting RandomOne
Nothing personal, but that was the point I was trying to make in this post http://www.forexfactory.com/showthre...46#post4436846 , but your reply at that time seem to imply that the market can change enough to invalidate all such behaviour. Slightly different issue I know as that discussion was about time frames so that may have clouded the issue.

The real question is what are the aspects that never change and can thus hopefully be used to produce consistent profits.
Most systems are just working temporarily. Finding out which systems work for decades is what I am researching on a constant basis. And for that, 600 trades on a 1-minute time-frame is far from enough. It won't show how this system performs over the long haul. You still need major economic shifts in your trading data, and that only comes from years or decades of data.
Otherwise you will never know if it is just a temporary inefficiency/luck or a real market inefficiency that has worked over the last years and is likely to continue working for years to come.
  • #20
  • Mar 4, 2011 10:34am Mar 4, 2011 10:34am
Quoting RandomOne
Testing surely does not necessarily imply that you are assuming the market dynamics are constant. If you get consistent returns then that is a demonstration that the strategy is invariant over the testing period. If not then it is a demonstration that it is suited to some conditions and not others. The latter is not a bad thing if you can anticipate the conditions and trade or not trade as appropriate.
Or you can also diversify systems, which are more than offsetting each other's losses and thus leading you to stable returns over the long haul.
0 traders viewing now