How to get 99% back testing

[raiderspaz]

I’m using tickstory but still can’t seam to get a higher back test, I there a good walkthrough on backtesting and importing tick data

[TipsyWisdom]

@raiderspaz find a broker that prides themselves on their historical data keeping is my suggestion.

[TipsyWisdom]

also, use MT5 only.

[l'andorrà]

@raiderspaz If you can afford it, I recommend Tick Data Suite 2. Their data is never below 99% quality. Not too expensive, but it isn't cheap either.

[q8carpenter]

@raiderspaz @TipsyWisdom @l-andorrà Hello,

This may be a bit off topic, but when you have a strategy to trade on MT5 and try backtesting it on different broker servers with their own data, you will find that your results will vary even though the differences are small between different broker prices. What that showed me throughout the years is that:

1. MT5 is the simplest way to get the data of my broker.
2. Best to optimize your EA for your broker and MT5 is the way to go.
3. If you see that your strategy loses or fails based on data accuracy, then keep in mind that different brokers already have differences in live prices and that shows me that accurate data is not the important factor in your backtesting or knowing that a strategy is a winner.

I believe that a good strategy should care that much on how accurate the data is. I mean if a pip here and there affects it that much, that's not a good strategy to me. Trying different brokers should just shows how solid the strategy is or how overoptimized it is.

That's my opinion and I wonder guy what do you think?

[TipsyWisdom]

@q8carpenter

I agree, but to a point. If its a strategy that is truly winning, it will at least have an up trend no matter where you run it is my experience. When data varies between brokers, its normally in the form of wicks being longer or shorter. A member of my group one day will have 40% quality running an EA, and then 2 days later, a different EA but the same pair and time frame will produce 99% data quality. Kinda just ones of those things you dont really know or think about until you test it across many platforms.

I'll go a step further with a suggestion and example:

Darwinex as a broker platform, at least just for testing and building purposes. https://www.darwinex.com/algorithmic-trading/tick-data

They are a platform that invites traders to come use their platform because they themselves are also algo people. They also have an incentive to give you quality data, because people can build bots, become a "darwin," and make money for ourselves by Darwinex investing in us.

As I said before, a broker that prides themselves on their data. These guys have down to the minute data for most pairs back to Sep 2017. When I read that I was like.."wait a minute," immediately I stopped trying to build bots further in history than Jan 1 2018 as its largely a waste of time. I wont go into the reasoning and theory of it all, but its advice that should be taken.

After I build bots in Darwinex, as long as its a winning strat I have no problems testing it on my broker AAFX, or any of the other brokers of people in my group. I think recently I saw the backtest results of one on IC Markets and while it didn't match mine specifically and slightly more DD, it was still a winning strat while his data was 60%? if I remember right.

In the end though, its all up to the user, personal preference. I dont mess with custom symbols and importing data. If that tickles your fancy, theres a free option called Tick Story.

[TipsyWisdom]

1 final food for thought, if you are testing and building on all ticks, all ticks isn't reality. You will never get all ticks trading on a live platform.

[raiderspaz]

@tipsywisdom i didnt that fxdreema worked on MT5

[tcanuto]

I also recommend optimization through MT5. I've been using it for several years and it's the best way to optimize faster using the computing power of the cloud. Thus, an optimization that would last months in MT5 can last a few hours in MT5 through MQL5 Cloud Network, not forgetting also Forward Testing already included in the Strategy Tester.

[bk7]

@tipsywisdom
darwinex really ? since when do they customers/clients allow to cash out on profits? did they change the policy, if yes please let me know

[TipsyWisdom]

@bk7 https://help.darwinex.com/how-to-create-a-darwin

[l'andorrà]

@q8carpenter Just curious. Does ANY broker offer good quality data for MT5 for the last, I don't know, 10 years? Because if they do it only for the last 2 years that is worthless when backtesting as strategy in the long run.

[TipsyWisdom]

@l-andorrà

thats purely personal preference. The undisputed great, Larry Williams is quoted as saying that he doesn't use any history over a year old.

More and more the market becomes algorithmically driven and it doesn't move in the same way that it used to is the ultimate basis for why you don't need to conquer the previous 10 years.

Just to further attest to that, I created an EA purely based on the statistical data over the last 5 years. I tried to run that as an EA....using hours of the day, days of the week, and months of the year to tell me when to long or short for that day and when to do it. It looks great in backtest. total garbage in forward test.

The objectives of the markets are the same, but it doesn't gyrate like it used to, the people who play the game aren't the same. Hell, JP Chase Morgan bank was barred from trading Forex for the last year or 3? One of the largest banks in the USA....probably the world....that bar was lifted a few days ago. They will significantly change the playing field adding a new dealer to the scene.

[q8carpenter]

@l-andorrà don't take my word as I'm not sure, but it seems to me and it would make sense that all brokers historical data is the actual data that they were running live, but stored up for us to use. It makes sense this way because you would see big differences between their results. But then again, brokers have data from before they were established. So I'm not sure how they really work. In any case, I ran the same EAs on FXCM and Swissquote once to see the difference. These EAs were optimized on MT4s historical data. These EAs ran very well on FXCM but were worse on swissquote. This showed me that in reality strategies should have some slack to the difference in pricing between brokers.

[l'andorrà]

Thank you for both replies. Appreciated.