Thread: Automation and backtesting

Based on the length date 1 to date 2, I get a pips.


Here's the problem

Let us assume the period I'm using is 15 years, I attempt to test arbitrary intervals that is arbitrary beginning date and arbitrary end day over the 15 years (by way of instance, period one is 200 daysinterval 3 1000 days, etc)

Out of 100 runs with arbitrary start date and arbitrary end date about 20 intervals have negative outcome.

So 20% runs are negative, though the first run (15 years are positive).

What's going wrong? Why a egy would operate on a time but when attempting it on a arbitrary (sub) period it neglects 20 percent of the arbitrary runs?

I am no expert, but if you're running a 15 year test, that's inevitably going to comprise wins and losses... Have you calculated that the win/loss percentage of trades over the entire period?

Even when the overall test is profitable, it will be composed of profitable and loss-making periods.

It depends on your egy, but what I have to do is team results by month, week and day so that I can see very roughly the number of profitable days I could anticipate in a week, how many profitable weeks I can anticipate within a month and what number of profitable months I can expect annually. None of it's proof of anything but it gives a general idea of the system so that if it starts deviating in trading, I understand to pause and reassess to find out if anything has gone wrong to me.

What I am attempting to receive my head is when periods are sub portion outcomes fluctuates.

It's amazing the connection between cash and pips, you can have negative pips and almost twice the account or favorable pips but wind up with half account.

More interestingly,

Same pip count (or very close) for two periods: one with favorable cash and one with negative cash.

Well, based on my expertise, testing is important but maybe not important. For me, using a logical system is/was most crucial closely followed by if you think in your idea, simply fund it with a couple of hundred dollars and run it forward out there...Do something which just logically works: hopefully you have a system that uses logic (logical to you at least) and Isn't Only a pure black box to you because I think that makes this all work:IE something like trend after. The more it makes sense, the more likely you should follow along in bad and good times. Then dial the time to M1 or anything down IE start to place the trend and utilize another area.
Start small, utilize like 10:1 leverage.If you start to win, say making 1%, you get 10% wins each time Due to the leverage
Start small (part two), start running your version with a few hundred $.If you are not inclined to attempt your model out with a hundred dollars or two, why are you doing this? About the above: Getting actual execution of everything you are performing is essential since real vs. demo do have differences, particularly if you are working on quick/small timeframes. The experience seems to be worth a lot.
Basically wanted to mention this. My pet peeve is the people that simply never fund and examine and test and test and examine... You could give up something or 5 dinners out ($20 a pop) or some thing and fund yourself with $100 or anything... then as you gain confidence in your skills, sacrifice and add more.

ALTHOUGH, I am just getting more powerful so take with a grain of salt: I've been studying since 2008, trading with real cash since beginning of 2017 (breakeven) and demo trading since approximately 2014.

Thanks for input and the stepssignal, yes I agree with wanting the machine with account That is my next step.

I am not attempting to fit curve the system differently I wouldn't be performing 100 runs on random periods.

I am just hoping to see if I'm missing some thing before forwarding the machine.

Algo traders and backtesters know best but are all systems due to fail later on ?

It's hard to get definite answer at this point.

What I am trying now is to play with the money management and see if I will get superior results as results above are based on 2 percent risk.

Something is telling me that 2 percent rule are doomed to fail or it's not the best way at best

Alright 4 MM
1- MM only 2 percent (min run was 3698.65 maximum 37340.82)
2- MM 2 percent reduced when above 10000 (min 9461.97 maximum 10845.11)
3- MM 2 percent reduced if below 10000 (min 6623.54 maximum 16481.01)
4- MM 2 percent reduced when below and above 10000 (min 6868.08 maximum 16409.94)





Which would you choose?

Backtesting is tough but provides you an idea of how the egy will work in the market. Matters to contemplate slippage, swap rates, commission and effect news events. Finding a system which operates in background is great but running the machine for period in testing is really the only way.


Learning from history, you may but learning history is you are going to learn.

How can you specify X for testing that is forward? Can it be 20,100 or 1000?

Many would say 100 trades but in which the 100 came from?

Imagine if fails by X and succeeded by X-Y or X Y?

Even more interestingly suppose that if it succeeded by X and neglected by X Y?

Originally Posted by ;

How can you specify X for ahead testing? Is it 1000 or 20,100? Many would say where the 100 came from? What if neglects by X and succeeded by X-Y or X Y? Even more interestingly what when it neglected by X Y and succeeded by X?

I think that your answering your own question with respect to backtesting. Is it 15 decades vs 14 decades and 9 months vs 1 year.

The market changes, and new things are presented daily, can your EA adapt. When will you feel comfy with it, to exchange. This is the actual question.

For my backtesting my dataset is 15 years.

The system generated roughly 3000 pips from 7000 pips range.

Once completed, I conducted 100 random test (random beginning date and random end date from inside the 15 years). Out of the 100 runs, 20 runs were unfavorable (every run has its own positive and negative transactions but total pips where negative).

Now if I decided to forward test the system on a live account, just how long the forwards testing should be for?