Backtest problem with tickdata

I have a big problem and I can’t find a solution for it.
I designed a robot that works based on chart structure in 5 seconds,
The problem is that I use dukascopy data in the backtest through Tick Data software, and the accuracy of the backtest is 99.9%, but the results are different in live and backtest.
I tested the 5-second candles in three different brokers, I make them very closely with dukascopy and fxblue candles, actually the same candle, but the difference between backtest and live in different brokers is still very large!
Does anyone have such experience to help me?

How much will you pay for this info?

1 Like

oh :))
I think this platform is a place for sharing our information and experiences freely!
if I want to pay for it, I would consult with an expert in a professional way not in a forum.

it is, but you don’t even introduce yourself. First post and you wrote about problem, what a rude.

BTW, you have no idea how many experts are on this forum, want support in this case? Time to looking for professional in other place. No regards Greg

dear greg
First of all, I am not very familiar with this site and its space, and I found it through Google search, and when I saw that people were comfortable asking their questions here, I did the same. If you pay attention, I asked my question as soon as I created an account.
Definitely, the person who answers my question is a professional who shares his experiences with others.
Of course, rude and arrogant people can be found in any context :slight_smile:
But forgive my rudeness, I did not intend to offend and I am not very familiar with this space
With respect, Reza

Of course there are large differences, firstly backtest does not work under 1min timeframes, and under 1min all datasets from brokers are unstable because their fintech is not world class but ‘just good enough’. I know this because the architecture I have access to is HFT quality, but actually is primarily for wealth management and investing, processing algos every 50ms down to 100ms bars on Forex.

These consultants charge $1,000s/hr (I will be the only one on public forums that come close to this knowledge base, the architecture I use has Quantum/AI quality algos so my workload is very low), sometimes those consultants charge more, what you want to do is not something anyone except professionals with access to professional tools could achieve.

When it comes to sub minute datasets you need to stick with one broker/dataset and tune for that, if you try to compare or move it you are going to find inconsistencies everywhere which one day will be fixed and the next day not, for example you get data spikes at certain points that completely throw out the algos unless they are designed in a very specific way.

I don’t have to get access to hft , quantum, ai and so on, to know that backtest even on D1 chart will be different because of data, also two backtest software programs can give you different results. Did you create this AI or you have only access to it?

Thank you very much for the unexpected information you gave me :heart_eyes:
Is there a tool where I can use HFT or Quantum/Ai in it?
I recently implemented this advice about the broker, and my robot is running on a broker with a low spread, and at the same time I’m collecting all the price ticks during the day live to compare later with the tick that the broker gives me.