Index futures automation

[ironchef]

Well, the strategy finally crapped out. Everything was working as per the trading plan. This strategy will have a big loss and will take months to recover. In the backtests, I've had max +900 days of recovery time. Even though it looked like a steady strategy, mentally that would probably drive me crazy to sit through that. So I'm going to tweak the risk a little bit to figure out how to balance income with the risk:


Loss happened the week of 3/18. So the year down about -25%. Good thing for that max stop loss though..sheesh

View attachment 337163

View attachment 337164

Your cum profit curve is very unusual with lots of vertical jumps. Is it because of the way you accounted for profits/losses?

Mine had bigger drawdown than 25% on occasions mainly because my strategy is mostly leveraged long but the cum profit curve did not have vertical jumps even during 2020 and 2022.

 

[hilmy83]

Your cum profit curve is very unusual with lots of vertical jumps. Is it because of the way you accounted for profits/losses?

Mine had bigger drawdown than 25% on occasions mainly because my strategy is mostly leveraged long but the cum profit curve did not have vertical jumps even during 2020 and 2022.

I documented some errors in the code during the early days where the bot left profitable trades opened. But if the bot did perform as expected, it would've been similar performance.

20-30% dd is normal from backtests, but I'm not sure if I can sit through it. My purpose for trading is to generate income, if i wanted to sit through stagnation, might as well just invest long term. Save money on taxes and less headache.


Live

vs Backtest

 

[ironchef]

I documented some errors in the code during the early days where the bot left profitable trades opened. But if the bot did perform as expected, it would've been similar performance.

20-30% dd is normal from backtests, but I'm not sure if I can sit through it. My purpose for trading is to generate income, if i wanted to sit through stagnation, might as well just invest long term. Save money on taxes and less headache.


Live
View attachment 337183



vs Backtest
View attachment 337182

I get it but my question is if you own any financial instruments, the market values don't all of a sudden go down, or jump 20%, unless it is an accounting artifact?

 

[NoahA]

I get it but my question is if you own any financial instruments, the market values don't all of a sudden go down, or jump 20%, unless it is an accounting artifact?

I'm guessing that its simply one averaged-in trade that is closed at a loss. From what I gather about his algo, he takes trades on breakouts, or fades them, and then there is a process for taking the opposite side of the trade, or averaging into a trade. I think we all know that many times, if we average in, we eventually can get out for a tiny profit. But everyone now and then, the dip we keep buying just keeps on dipping... LOL

 

[schizo]

I think there is a sweet spot between trading the anticipated breakout and when trading the anticipated range move.

We would all be rich if we knew that in advance. But it's worth considering the market will likely be rangebound until there's a catalyst that comes around to mess up the status quo. Also trending and range markets are binary (range=0, trend=1). So if price is DOWN in a range, it's UP in a trending market. Not an easy feat to program, but feasible with right calibration IMO.

 

[NoahA]

We would all be rich if we knew that in advance. But it's worth considering the market will likely be rangebound until there's a catalyst that comes around to mess up the status quo. Also trending and range markets are binary (range=0, trend=1). So if price is DOWN in a range, it's UP in a trending market. Not an easy feat to program, but feasible with right calibration IMO.

I'm guessing that in hilmy's case, trading on a small time frame, there are lots of ranges but also lots of trends. If the NQ is stuck in a 200 point ranger, on a fast chart, that will still lead to a nice rally or consistent drop that will kill you if you're not in the right direction and scaling in.

 

[schizo]

I'm guessing that in hilmy's case, trading on a small time frame, there are lots of ranges but also lots of trends. If the NQ is stuck in a 200 point ranger, on a fast chart, that will still lead to a nice rally or consistent drop that will kill you if you're not in the right direction and scaling in.

That is why you need to combine more than just 1 TF (especially if @hilmy flips back and forth like I do). But I'm getting ahead myself.

Let me just say in passing that short term players are usually concentrated in the lower TF whereas the longer term players on the higher TF. And they usually go the opposite direction. The "sweet spot", however, is when both parties are in cahoots with one another.

 

[NoahA]

That is why you need to combine more than just 1 TF (especially if @hilmy flips back and forth like I do). But I'm getting ahead myself.

Let me just say in passing that short term players are usually concentrated in the lower TF whereas the longer term players on the higher TF. And they usually go the opposite direction. The "sweet spot", however, is when both parties are in cahoots with one another.

His algo would be 10 times more complicated I'm sure if it had to reference other time frames.

 

[ironchef]

I'm guessing that its simply one averaged-in trade that is closed at a loss. From what I gather about his algo, he takes trades on breakouts, or fades them, and then there is a process for taking the opposite side of the trade, or averaging into a trade. I think we all know that many times, if we average in, we eventually can get out for a tiny profit. But everyone now and then, the dip we keep buying just keeps on dipping... LOL

Thanks.

I know he doesn't like a 25% drawdown but I am very impressed with his overall results.

 

[hilmy83]

Back in the day, what I would used to do was fade the NQs (average in) until it broke the ranges, and just flipped the whole position for the breakout. It was consistent and scary at the same time. But it crapped out when the flip doesn't work and it just mean revert again.

This time, I am thinking to basically play the ORB/reverse until failure to a certain point then averaging in.

Studying my losses in the backtests/live, I notice that when price breakout out nicely from OR, it does it in 1-2 setups. When price zig zags around OR, there's likelihood it would just range going forward until end of session.

So I'm going back to the drawing board to reprogram the bot to do that. Run some tests and let it run live.