How Many Years of Positive Results?

[gaz0001]

Hey Guys,

I have created an algo, and have been working on for about a week. Testing largely, because it just showed great results right off the bat.

I largely used my trading experiences from 2023 to develop it.
I did not optimise it, i simply tested it on weekly, monthly then yearly data. It performed fantastic for 2023 and 2024 data.

I extended the backtest for multiple years, and found that it was not good 2012 - 2018.

So, i have to the question - how many years of data are relevant for a daytrading algo?


Below is annual net profit, and monthly net profit.
Essentially, if i had launched this since 2019, i would of not needed to switch it off. I could handle the drawdown of the bad months.

 

[hilmy83]

Hey Guys,

I have created an algo, and have been working on for about a week. Testing largely, because it just showed great results right off the bat.

I largely used my trading experiences from 2023 to develop it.
I did not optimise it, i simply tested it on weekly, monthly then yearly data. It performed fantastic for 2023 and 2024 data.

I extended the backtest for multiple years, and found that it was not good 2012 - 2018.

So, i have to the question - how many years of data are relevant for a daytrading algo?


Below is annual net profit, and monthly net profit.
Essentially, if i had launched this since 2019, i would of not needed to switch it off. I could handle the drawdown of the bad months.
View attachment 335537

View attachment 335538

So you just decide to brush over the elephant in the room?

"I extended the backtest for multiple years, and found that it was not good 2012 - 2018."

 

[Zwaen]

How many years would be satisfactory for you to trade it?

 

[gaz0001]

So you just decide to brush over the elephant in the room?

"I extended the backtest for multiple years, and found that it was not good 2012 - 2018."

I didnt notice the elephant so far. Would you point it out?

When looking at the data on a yearly perspective, i looked at the annual return for example.
19 - 24 were big years +50% to -33%.
But then 13 and 17 were +30% years too, but the perfromance of these 30% years do not reflect that of the later ones.

My strategy is Long/Short. In 21 & 22 it was a similar percentage gain/loss - my strat shows the same relative gains to the relative bias.

 

[SimpleMeLike]

Hey Guys,

I have created an algo, and have been working on for about a week. Testing largely, because it just showed great results right off the bat.

I largely used my trading experiences from 2023 to develop it.
I did not optimise it, i simply tested it on weekly, monthly then yearly data. It performed fantastic for 2023 and 2024 data.

I extended the backtest for multiple years, and found that it was not good 2012 - 2018.

So, i have to the question - how many years of data are relevant for a daytrading algo?


Below is annual net profit, and monthly net profit.
Essentially, if i had launched this since 2019, i would of not needed to switch it off. I could handle the drawdown of the bad months.
View attachment 335537

View attachment 335538

Good Morning gaz0001,

Looks good. Great work.

It is better than nothing. I would go for it if you are ok with the overall drawdown.

Also, go back to 2006 and back test from there as well. Atleast you know overall how bad or good the system may do.

 

[SimpleMeLike]

So you just decide to brush over the elephant in the room?

"I extended the backtest for multiple years, and found that it was not good 2012 - 2018."

Good Moring hilmy83,

Good question for you.

Do your systems back test show profitability year to year from 2006?

 

[gaz0001]

How many years would be satisfactory for you to trade it?

Thats kinda what i was hoping to get advice on from other automated traders.

I wanted to benchmark 3 years

18 = Flat
22 = Bear
23 = Bull

 

[tomkat22]

Yeah I'd be investigating why it performed poorly 2012-2017. What was different back then. Lower volatility? Choppier market?

 

[hilmy83]

Good Moring hilmy83,

Good question for you.

Do your systems back test show profitability year to year from 2006?

Yes. Tested on 1 NQ contract. Volatility was really low back in those years. Would've probably traded 5-10 contracts back then to make it worthwhile.

But the strategy was profitable.

I didnt notice the elephant so far. Would you point it out?

When looking at the data on a yearly perspective, i looked at the annual return for example.
19 - 24 were big years +50% to -33%.
But then 13 and 17 were +30% years too, but the perfromance of these 30% years do not reflect that of the later ones.

My strategy is Long/Short. In 21 & 22 it was a similar percentage gain/loss - my strat shows the same relative gains to the relative bias.

Am I slow or you don't make any sense here. Nothing you said makes sense to me based on what you presented bro. You lost money literally 6 years straight in a very bullish market.

 

[danielc1]

I wonder how many people do a backtest and then start trading with a positive outcome.