It seems more problematic for SQ4 to build/find strategies at least when comparing with SQ3

To put it into context I would just like to pin point that when both SQ3 and SQ4 are running in parallel, under the approximate conditions I mentioned above, SQ3 found approx 500 strategies where SQ4 found around 100. 

But even if SQ4 was as fast as SQ3, then again, the idea is not to have SQ4 running configured like a kind of a SQ3 (in that case I prefer SQ3). The idea behind SQ4 - I think - is to have a more powerful SQ4 that still can find a good amount of better strategies than SQ3 would ever find, in a reasonable (not very large) amount of time.

Just another quite relevant comment:

The strategies found in SQ4 (around 1500 now) are all, without exception, highly biased to SELL orders, so for instance a strategy with 1480 trades or so can have 1430 sell trades and only 50 buy trades. This is highly unusual and strange. The process running in SQ3 with similar settings has much more evenly balanced results between Sell and Buy orders (as it would be expected).

Settings in SQ4 in this respect are "Both Long and Short" and "Symmetry in both Entry and Exit", so there is no bias in settings that could justify this kind of results.

could you please attach your builder config when you tried building with the configuraiton you described above?

It is true that SQ4 is slower than SQ3, but it is partly because it has more features - it has "true" trading engine, compared to a simplified one in SQ3.

For that reason strategies in SQ3 were bound to a fixed format, and in SQ4 you can create your own strategy template to use in generation.

Many things including the problems you describe could be caused either by configuration, ot also by some fine tuning on our side.

Just settig up some weights internally could make a difference in the generated strategies, and it will take some testing to fine tune everything perfectly.

Apologies for the late answer.

Thanks for your comments and clarifications, Mark.

I don't totally grasp what you mentioned: "in SQ4 you can create your own strategy template to use in generation". I think I need access to the future training materials to fully understand what you mean.

Regarding the SQ4 engine (as compared to SQ3) being the cause of a big speed disadvantage, would that still be valid if one configures SQ4 as close as a SQ3 (including SQ3 "old architecture") ? You see I've been running a test for more than 5 days now in my I7 8 core PC , and with SQ4 configured as much as likely as it would be a SQ3 and sure enough in this time I got more than 2600 strategies, but still far away from the 5000 strategies I got with SQ3 with same data, same pair and in little more than half of the time.

But, then again, the idea is not for SQ4 to be configured as a SQ3, I've just configured it like a "kind of" SQ3 because I wanted to be sure the system would be able to generate strategies since when tried to use SQ4 like a "SQ4" and after several days I couldn't get anything interesting out of the tool.

As requested I attach the following config files:

"Config_M15_1H_11.xml" - this is the config file I used trying to exploit SQ4 advantages like looking to two charts (M15+1H) to decide on taking a trade: I couldn't get any strategies out of this, after several days of "building process". Another SQ team member told me in another ticker where I added this config file that it contained errors, I tried to understand what errors were those but got no reply (maybe because ticker was already closed)

"AUDNZD_M15_SQ3_II.xml" - Attempt to configure SQ4 like a SQ3: here, I get strategies but a much lower pace than SQ3 (about 30% of the pace) and, as expected, with no evident quality advantage to the strategies I got with SQ3. But then again, the idea is not to use SQ4 like a modified SQ3, in that concept, naturally I would prefer to use SQ3 instead.

If there are any errors in my config files I would appreciate it a lot of I could get some guidelines to understand how can I guide my workflow in the future.

Thank you!

just another comment: I understand a more complex and powerful tool can take more time to do the same as a more simpler tool. But it's also true that the total SQ4 workflow is extremely time consuming and if we are adding up a big deal of time in the generation phase (where the user is not sure if there will be anything with the faintest chance potential coming out or not) it might frustrate or remove faith from users about the potential of the tool, it's my opinion.

I don't think SQ4 workflow is any more time consuming than SQ3 workflow. Do you mean that because SQ4 is slower?

Yes, SQ3 is still faster, but SQ4 is still not final version. It took a lot of time to make all the speed optimizations in SQ3, SQ4 should also be getting faster in the with new updates. Also RC6 is faster than RC5.

I checked your configs, few suggestions:

- use Random generation first with any new setting, at least for some time - this way you'll be able to easily find out if your filters all work, if they aren't too strict, etc. Genetic evolution adds another layer of complexity, it takes a lot of time only to generate initial population in your settings.

- you have too many conditions - from 1 to 7. More conditions mean the tsrategy is curve fitted to data, I'd recommend using 1 to max. 3 conditions.

- you use Global Shift from 0 to 25, again not good. Especially Shift=0 is slow, because the indicator has to be checked with every tick. In SQ3 you cannot configure it, and the default used in SQ3 is from 1 to1.

I'd recommend Global Shift range 1-1, or 1-2.

- use Selected timeframe precision to generate strategies quickly., You can then use cross check Retest with higher precision to verify the results with better precision. It wil run only after your strategy  passes the filtering, so it will save you time. 

Just by changing the last 3 points the generation speed doubled.

SQ4 has more settings than SQ3, so it can be fine tuned better, but it also means it is more complex.

Thanks a lot for your inputs Mark, I will surely test them in a few days, as I am going on vacation today.

About the conditions I thought it was supposed to be the maximum number of indicators used for entering and exiting a trade. If I look to my strategies developed with SQ3 some of them have surely 5 or 6 indicators, all strategy rules considered.

I had followed already the approach to test with selected time frame and then retest with M1 only the strategies that obey some kind of filter but basically because of other problems you mentioned this didn't produce any results.

It's good to know that future SQ4 versions will be optimized for speed. I didn't you had the same problem with SQ3.

Clearly I misunderstood the Global Shift parameter.. I have to wait for more education materials.

I meant that SQ workflow to reach a point to put a strategy to a live test is extremely time consuming. I didn't mean SQ4 in particular. Apologies for the typo error.,

Thanks again.