How much backtesting is enough? | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

How much profit is too much for brokers? 4 replies

Trading Capital - How much is enough? 28 replies

Why does Backtesting differs so much ? 19 replies

Starting trouble: How much deposit is good enough? 21 replies

How much backtesting is enough?

  • #1
  • First Post: Jan 17, 2011 8:32am Jan 17, 2011 8:32am
Hi!
I would just like to ask if anyone has had any experience on how much backtesting is needed to prove that a system is constantly profitable? If we concentrate on forex then you can only get about 8 years worth of data. Is that enough? Or maybe it is impossible to say, one never knows what might happen in the future and everything could change dramatically. But has anyone else pondered on this subject. Because I atleast would like to learn a style of trading that always works. Ofcourse some changes to certain parameters might have to be updated but the principles should remain the same.

-Richard
  • #2
  • Jan 17, 2011 9:12am Jan 17, 2011 9:12am
depends on: average percentage of profitable trades, volatility of return ...

time is not the issue, number of trade is...
50 trades in one year proves nothing....
if you don't have 85%+ profitability per trade...
but it still proves nothing....
if your trading isn't robust enough....

I would say that after 500 trades on 50%, you can say you have something (and this is a minimum) but it really depends on the drawndown. Test on 10 year data for 20 trades per year isn't enough.
  • #3
  • Jan 18, 2011 9:12am Jan 18, 2011 9:12am
Thank you for your reply Rikers! Much appreciated!
I currently have about 550-570 trades on one pair and every year so far has been profitable even tough there have been swings in the profitability so I have some confidence in the system. Ofcourse I would like to go back something like 10000 trades but that is sadly not possible. But if I go through different pairs then maybe that could boost my confidence even further if they would also show some positive signs.

-Richard
  • #4
  • Jan 18, 2011 9:19am Jan 18, 2011 9:19am
Quoting ariku
Thank you for your reply Rikers! Much appreciated!
I currently have about 550-570 trades on one pair and every year so far has been profitable even tough there have been swings in the profitability so I have some confidence in the system. Ofcourse I would like to go back something like 10000 trades but that is sadly not possible. But if I go through different pairs then maybe that could boost my confidence even further if they would also show some positive signs.

-Richard
Hi Richard,
I am interested in backtesting my system. You mentioned that you have 7 years of data. Is it free, if not how much does it cost?
Thanks in advance.
  • #5
  • Jan 18, 2011 9:49am Jan 18, 2011 9:49am
i think is more good if you want to back test an strategy,to consolidate that strategy in one EA and get one vps,and let it run 3months should be enough if the strategy is good and have proper money management and risk management,of course if you don t want to test all the time just manualy.
And of course if it`s possible.just my point of view.
is more better to build more strategy and to back test them all on differently pairs in the same time.
that way on the final of the 3months,6months,1year you know what is more good to use,instead of just testing1 for one year.
  • #6
  • Jan 18, 2011 2:19pm Jan 18, 2011 2:19pm
Quoting ariku
But if I go through different pairs then maybe that could boost my confidence even further if they would also show some positive signs.

-Richard
You should know that this needs to be done on pairs that are not correlated. The problem is most of them are somewhat positively or negatively correlated. ex: I wouldn't suggest after testing EUR/USD to test the GBP/USD. since E/U and G/U behave the same way.

http://www.mataf.net/en/tools/01-01-correlation

- more closely to 0 the better... If you tested EUR/USD - USD/JPY would be a good choice
  • #7
  • Jan 18, 2011 11:44pm Jan 18, 2011 11:44pm
My experience is that, matching 10000 trades from eons ago does not matter........

The results from a few years till now matter most.......Markets are dynamic, looking too far back, you will only discover what work "then"
  • #8
  • Jan 19, 2011 4:18am Jan 19, 2011 4:18am
Quoting newborn1000
My experience is that, matching 10000 trades from eons ago does not matter........

The results from a few years till now matter most.......Markets are dynamic, looking too far back, you will only discover what work "then"
This might also be very very true but one can never be 100% sure on that. And I always believe that alot can be learned by studying history.



Thank you Rikers for that! That site looks great and I will do just as you said and continue to backtest on a pair that is not correlated!
  • #9
  • Jan 19, 2011 4:19am Jan 19, 2011 4:19am
Quoting enebro
Hi Richard,
I am interested in backtesting my system. You mentioned that you have 7 years of data. Is it free, if not how much does it cost?
Thanks in advance.

You would want EUR/USD data is that correct? And what kind of data? Tick, 5min, 1hour or daily?
  • #10
  • Jan 19, 2011 6:26am Jan 19, 2011 6:26am
Quoting ariku
Hi!
I would just like to ask if anyone has had any experience on how much backtesting is needed to prove that a system is constantly profitable? If we concentrate on forex then you can only get about 8 years worth of data. Is that enough? Or maybe it is impossible to say, one never knows what might happen in the future and everything could change dramatically. But has anyone else pondered on this subject. Because I atleast would like to learn a style of trading that always works. Ofcourse some changes to certain parameters might have to be updated...
As mentioned in other posts, it's about getting a larger sample as possible...

If you test on 10 years of data while only generating 50 trades, that's not a decent sample at all...If your strategy generated 50 trades in 10 years, it means that it's long term by nature, which means that it needs more historical data to test than 10 years...

Generally speaking, 500+ trades would be a good start for a sample...You might not have a 100 years of data for currencies to test on, however, you can test on multiple currencies and aggregate the results....However, as mentioned you will have to watch for correlations closely...


Regards,

Nader
  • #11
  • Jan 19, 2011 6:36am Jan 19, 2011 6:36am
Quoting ariku
You would want EUR/USD data is that correct? And what kind of data? Tick, 5min, 1hour or daily?
Ideally I would love intra-day data so i could run my strategy manually. I trade off the 15 min chart but i look at hourly chart also and yes Eu/Us would be fantastic.
Many Thanks.
  • #12
  • Jan 19, 2011 6:52am Jan 19, 2011 6:52am
Quoting ariku
Or maybe it is impossible to say, one never knows what might happen in the future and everything could change dramatically



Rather than give you an essay on the topic I will only say..

Interpret backtest results with caution. Dont curve fit.Use common sense.

The market has a funny way of changing completely in to something it never was in any time during the last 10 years of back testing...the exact second you wager some real cash.

You will swear that the market is conspiring again't you.
  • #13
  • Jan 19, 2011 6:55am Jan 19, 2011 6:55am
Thanks narafa for you contribution!

As I mentioned before I have 500+ trades so I think that I am on the right track with my system. But I will make more tests on different uncorrelated pairs so that I can get an even further confirmation.
  • #14
  • Mar 9, 2012 3:03am Mar 9, 2012 3:03am
Quoting ariku
Thanks narafa for you contribution!

As I mentioned before I have 500+ trades so I think that I am on the right track with my system. But I will make more tests on different uncorrelated pairs so that I can get an even further confirmation.

I have to say that I agree with Narafa... No amount of back testing will tell you what the market will do. Also, I have yet to see a system that applies to all pairs on all time frames.
Kill the ego, or it will kill the mission.
  • #15
  • Mar 9, 2012 6:43pm Mar 9, 2012 6:43pm
"No amount of back testing will tell you what the market will do" absolutely TRUE!!


Quoting AaronWard
I have to say that I agree with Narafa... No amount of back testing will tell you what the market will do. Also, I have yet to see a system that applies to all pairs on all time frames.
  • #16
  • Mar 9, 2012 9:00pm Mar 9, 2012 9:00pm
Quoting ariku
Hi!
I would just like to ask if anyone has had any experience on how much backtesting is needed to prove that a system is constantly profitable? If we concentrate on forex then you can only get about 8 years worth of data. Is that enough? Or maybe it is impossible to say, one never knows what might happen in the future and everything could change dramatically. But has anyone else pondered on this subject. Because I atleast would like to learn a style of trading that always works. Ofcourse some changes to certain parameters might have to be updated...
in my opinion all the back testing in the world still can't foretell the future.
AVT INVENIAM VIAM AVT FACIAM
  • #17
  • Mar 10, 2012 1:31am Mar 10, 2012 1:31am
do some trading simulations on recent data.get at least 100 trades through and do some Monte Carlo simulations.you will save yourself a lot of time than trying to get 10000 trades to have 'statisticaly significant' numbers.If u have a real edge it will show up pretty fast.
wo-yoy! wo-yoy! wo-yoy! wo-yoi! wo-yoy-yoy-yoy!
  • #18
  • Jul 10, 2019 5:54pm Jul 10, 2019 5:54pm
IMO backtest on a regular basis, even when you are profitable, just to keep your strategy sharp. Has for how much backtesting is needed depends on the timeframe you trade how many pairs you trade ( Forex). What your risk to reward ratio is. What your win percent need to be and what precent of your account your going to risk. When you've decided that then you know what your personal goal should be. If you have a 1 to 1 risk reward then your win rate has to be 85% or higher, preferably higher. If your risk to reward is 1 to 3 then you can make good money at 50% win rate. Higher is always better. I backtest over all seasons of the market. Spring summer autumn and winter as the market changes. Then when your overall win ratio is at what you require to make a comfortable profit, go live. I'd also advise new traders not to demo so mush, but to go live as soon as possible with a micro account to get accustomed to trading real money. It takes some new traders time to become accustomed to trading real money.I'm speaking from experience when I say this
  • #19
  • Jul 11, 2019 2:39am Jul 11, 2019 2:39am
I backtested hundreds trading systems on Forex and Futures.
As mentioned above, more important than time is trade’s quantity.
Just to make sure that the system is one more shit, you should get backtest for 10 years and not less than 1000 trades without uncontrolled martingale.
  • #20
  • Jul 11, 2019 4:48am Jul 11, 2019 4:48am
the more the better
0 traders viewing now