Tick Database

Is anyone familiar with a quality tick database? Free if possible.

I've been searching a lot lately, but can't find a good one. I'm thinking of creating my own.

well i've heard that gain capital offers the largest database however it's not the best as it has a lot of gaps.

you can also get some data from Oanda.

on both cases however you will need to open an account.

The one at Dukascopy is free for personal use. https://www.dukascopy.com/swiss/english/data_feed/historical/ 😉

Yonex posted:
    The one at Dukascopy is free for personal use. https://www.dukascopy.com/swiss/english/data_feed/historical/ 😉

Thanks!

I've haven't had the time to test it yet. Have you worked with it? Is it reliable?

biz0101 posted:
    well i've heard that gain capital offers the largest database however it's not the best as it has a lot of gaps.

you can also get some data from Oanda.

on both cases however you will need to open an account.

Yeah tried both. Problem with Oanda is that you can get only one free pair data at a time and it goes only back to 2004.

I have worked with it, and I believe the data from Dukascopy is the most reliable you can get for free. 😎

Yonex posted:
    I have worked with it, and I believe the data from Dukascopy is the most reliable you can get for free. 😎

Great! I will try it later.

Thanks!

PipCollector posted:
    Is anyone familiar with a quality tick database? Free if possible.

I've been searching a lot lately, but can't find a good one. I'm thinking of creating my own.

https://www.forexfactory.com/showthread.php?t=211652

Why are ticks worth looking for?

With good quality tick data, you can achieve 99% backtest modelling quality. At present MT4 can only achieve 90% because you cannot get below the M1 bar level......MT4 mathematically interpolates the numbers below this level which leads to inaccurate testing, especially with scalpers.

Yes, although past perfomrmance is non indicative of future results, I think there is only so many scenarios you can test your system on.

So when it comes to testing, you should have the closest prices to the real ones to test your system on, or else the backrest will be useless.

So I do think the past performance is somehow correlated to future performance.

Dunno hey,

Have you guys ever tried running the same data, with the same code, on a different processor?

No matter how good your data, a different processor will give a different result. And I'm not talking 1% or 2% here. On one pc the system will margin call the next it will do 1000%.

I have never, ever, been able to match a back test for a period to a real result. Back testing is la la land....

Before you put any energy into it, test back testing. My humble suggestion.

Here's an exercise,

Run the same data, same processor once online then offline on a period you have live date for. You'll end up with three different results as metatrader will decide to make up stuff or not depending on how online and offline you are.

Then run the tests on a different machine, online offline. See what happens. 6 results.

There's only one way to test. Run it live....

Uhmmmm......Yeah, I think that is the point of having good data to begin with.

Elkart posted:
Before you put any energy into it, test back testing. My humble suggestion.

Elkart, despite the cynism of your post, you do make a good point albeit unintentionally, I suspect.

There is far more to backtesting than hitting the start button on the tester.......and this is one of the reasons that most people fail in Automated Trading. They think cos they can run a backtest, they are suddenly an expert🙄

If you want to get ahead, you cannot say that backtesting is useless......first you need to know what you are doing, understand what the limitations of the backtester are and then find ways to improve the testing(ie tick data, MT5, etc). Even then backtesting is only the start of a 3-stage process of getting an EA onto an account.

There are many good systems in the dustbin because of back testing. Metatrader takes guesses at data. I got so frustrated once I spent a month just running test then comparing to live results. Using different machines. It was chaos. I didn't get one coherent result.

Any point I make is intentional.

And it's not like we don't know what we're doing. The same code that fails in back testing is now running live for us, perfectly profitable.

Data has a lot to do with it, but before we even get to data, the processors have differences. Back testing hardly works for anything, all it's good for is to see of the code does what it's supposed to. We use it to test code, so lets say we want to test tp software we'll check the graph, if it goes up we'll put down a few buys and see if we tp correctly, other than that it's like learning to drive a car on a bicycle. Has no value at all, very misleading...

And another thing,

I'm sitting here scratching my head, we're running on IBFX and Varengold bank, one bucket shop, one ECN, both live, same code. Same signal, 190 pips difference in entry. How the F$%#k is that possible?

And now we want to throw back testing into it...?!

It's not that the rates are different at point of entry, but the data feed we calculate on is so different that over a 24 hour period it makes that big difference in entry...so what chance does back testing stand?

Elkart posted:
    Here's an exercise,

Run the same data, same processor once online then offline on a period you have live date for. You'll end up with three different results as metatrader will decide to make up stuff or not depending on how online and offline you are.

Then run the tests on a different machine, online offline. See what happens. 6 results.

There's only one way to test. Run it live....

I think the difference is because that when you are online, the spread for your currency pair changes, and thus the backtested result changes. When I do my backtest, with 99% quality, I can adjust the spread I want to use, and get identical results online/offline. However, I haven't tried backtesting the data on another computer, but I will. Thanks for your suggestion! 😉

It's because Mt4 goes and fetches data if it can. Quite a bit on that on the mt4 forum. But it stands to reason that a week-end result will be different to a week result then.

Basically it's just such a #$%$ mess that I don't bother. No one should take it seriously.