Profitable Strategies?

Has anyone ever managed to create any profitable strategies with SQ? I’m using it for over a month now, constantly generating strategies with the most basic filters and nothing good has come out yet. It seems like it doesn’t have the capability to create something good.

HI Sloth

Check out my post: StrategyQuant success

So definitely.. yes I have a few successful strategies but you do need to put in the work

The biggest issue I see with SQ is once you just slightly tweak filters to look for a profitable strategy with low dd it becomes basically impossible to generate anything. And I’m not talking of using anything unusual with dozens of filters.

I followed the SQX’s courses, and also tweak nearly all parameters as same as the course used.

I tested for EURUSD & GBPUSD, there are some strategies that passed the basic robust tests. But not good for other currency pairs.

I think it need to develop a technic to tweak those parameters, but sadly the SQX’s community is too small to discuss those tweaking technics. You can only learn by try and error. I guess it take at least couple of months.

Getting a profitable strategy in a single-market build is straightforward, but its not easy to find a strategy with good OOS performance across >3 markets. Some tweaking of the default build settings is probably required.

Alternatively you can decrease the difficulty of your robustness tests, but compensate by diversifying your portfolio with more strategies.

huangwh88 wrote:

Getting a profitable strategy in a single-market build is straightforward, but its not easy to find a strategy with good OOS performance across >3 markets. Some tweaking of the default build settings is probably required. Alternatively you can decrease the difficulty of your robustness tests, but compensate by diversifying your portfolio with more strategies.

I’ve been trying for over a month now, using a single market and even decreased the robustness testing to a minimum but it seems like SQ is still unable to come up with something reasonable if filters are even just slightly advanced (eg. DD below 50%, higher win rate above 50% and profit factor above 1.3). The 1 or 2 strategies generated after a week which actually work with these parameters already fail with a higher precision backtest.

At this point I’m not really convinced of it, especially when it comes to strategies on lower TF

sloth011 wrote:

huangwh88 wrote:

Getting a profitable strategy in a single-market build is straightforward, but its not easy to find a strategy with good OOS performance across >3 markets. Some tweaking of the default build settings is probably required. Alternatively you can decrease the difficulty of your robustness tests, but compensate by diversifying your portfolio with more strategies.

I’ve been trying for over a month now, using a single market and even decreased the robustness testing to a minimum but it seems like SQ is still unable to come up with something reasonable if filters are even just slightly advanced (eg. DD below 50%, higher win rate above 50% and profit factor above 1.3). The 1 or 2 strategies generated after a week which actually work with these parameters already fail with a higher precision backtest. At this point I’m not really convinced of it, especially when it comes to strategies on lower TF

I generally lean towards the higher timeframes, M30 and up because 1) I think SQ is better at producing trend strategies and 2) higher timeframe strats take less time to build (using selected timeframe only precision).

Tick backtests usually pass for me, but the additional market tests kill almost everything. Whether this is due to curve fitting, or the markets being too different, is difficult to tell.

Again I want to add the efficiency was strongly depends on the market you choose. I run a test last night:

5000 builds come out with 50 strategies that passed robustness test pipeline (prev+last OOS, slippage, cross market+cross TF, MC series tests, WMF) for EURUSD

But for AUDUSD or EURGBP, I’d tested with same condition with high correlated cross market, for each pair I’d built at least 15000 builds, none of them come out with any result. Also, it is too slow comparing to EURUSD/GBPUSD/GBPJPY when finding accepted strategies at building stage.

I found that for AUDUSD and EURGBP they are almost eliminated by OOS, slippage tests.

Maybe one of the solution to speed up the efficiency is just directly build from whole time period instead of split them into OOSs, after all, they are mathematical equivalent when the filter is ‘Ret/DD+prfit factor+# trades’.

Another solution is increase SL+TP range and risked money to accept more variety strategies.

You might be using the wrong building blocks.

If you share your config i can take a look and try to help.

kasinath wrote:

You might be using the wrong building blocks. If you share your config i can take a look and try to help.

Hi Kasinath, what building blocks do you recommend?

What building blocks do you recommend?

Heheh, thats a bit like an artist asking another artist : “what paint colors do you recommend”.

There is an infinite number of combinations that can be suggested, so it wouldn’t be worthwhile to answer, without an idea of the goal (what are we trying to paint / generate).

If Sloth011 is able to share his Config file, i can give specific guidance.  Looking at it, I can answer the key questions:

1. What asset are we trying to trade?
2. What broker will we be going live with?
3. What is the initial deposit we will work with?
4. What frequency do we want trades to occur? (a few times a day? a few times a month?)
5. What kind of trading system? Trend following? Mean reverting?

kasinath wrote:

What building blocks do you recommend?

Heheh, thats a bit like an artist asking another artist : “what paint colors do you recommend”. There is an infinite number of combinations that can be suggested, so it wouldn’t be worthwhile to answer, without an idea of the goal (what are we trying to paint / generate). If Sloth011 is able to share his Config file, i can give specific guidance. Looking at it, I can answer the key questions:

1. What asset are we trying to trade?
2. What broker will we be going live with?
3. What is the initial deposit we will work with?
4. What frequency do we want trades to occur? (a few times a day? a few times a month?)
5. What kind of trading system? Trend following? Mean reverting?

Nice, spoken like a pro. Let’s hope Sloth011 shares his config and we can all learn from it.

Sorry for the late response. So far (as I am new to SQ), I have been following the course and used the default trend preset provided within SQ as I am not too familiar with building blocks yet. I confess that probably I have yet to learn everything more in detail and I’m working on that, so partly perhaps I’m also to blame for that, but so far I haven’t really managed to get anything really decent. If you tighten the filters a little bit, let’s say for having around 10 or more trades per month with a PF over 1 and low DD and certain  amounts of profitable trades, what you’re getting is basically very limited, especially once having more rigorous crosschecks.

I also tried working on a few scalpers on a lower TF (>M5) but didn’t manage to get anything either.

Well with that being said, perhaps I am to blame here for not setting it up correctly, but I am still not 100% convinced by SQ.

sloth011 wrote:

Sorry for the late response. So far (as I am new to SQ), I have been following the course and used the default trend preset provided within SQ as I am not too familiar with building blocks yet. I confess that probably I have yet to learn everything more in detail and I’m working on that, so partly perhaps I’m also to blame for that, but so far I haven’t really managed to get anything really decent. If you tighten the filters a little bit, let’s say for having around 10 or more trades per month with a PF over 1 and low DD and certain amounts of profitable trades, what you’re getting is basically very limited, especially once having more rigorous crosschecks. I also tried working on a few scalpers on a lower TF (>M5) but didn’t manage to get anything either. Well with that being said, perhaps I am to blame here for not setting it up correctly, but I am still not 100% convinced by SQ.

Personally I’ve had more success on the higher timeframes. Perhaps you could share your config and we could have a look.

In reviewing the 1000 test strategies provided created in strategy quant, there was nothing currently performing that I would be comfortable with on a live account. Just seems too random, rather than having a clear strategy that makes sense logically to test.

Using Algo Wizard, I have been able to create a strategy that I would be comfortable trading on 17 pairs / timeframes from 1 minute to 15 minutes, involving a combination of EUR, USD, GBP, CHF, AUD, NZD, CAD. Just hoping that it tests live as well as the backtests have been.

Hi,

I would like to share my personal setting

huangwh88 wrote:

kasinath wrote:

What building blocks do you recommend?

Heheh, thats a bit like an artist asking another artist : “what paint colors do you recommend”. There is an infinite number of combinations that can be suggested, so it wouldn’t be worthwhile to answer, without an idea of the goal (what are we trying to paint / generate). If Sloth011 is able to share his Config file, i can give specific guidance. Looking at it, I can answer the key questions:

1. What asset are we trying to trade?
2. What broker will we be going live with?
3. What is the initial deposit we will work with?
4. What frequency do we want trades to occur? (a few times a day? a few times a month?)
5. What kind of trading system? Trend following? Mean reverting?

Nice, spoken like a pro. Let’s hope Sloth011 shares his config and we can all learn from it.

Hi, I would share my personal setting.

I build from 2013 to 2018 wthit 50% is and 50% oos. In the 2010-2011 I run the first oos and from 2019 to now the second oos. Than i test the Slippage, other markets, other time frame and i run some Monte carlo.

Do you think it’s a good process?