continuation of ticker sqr_5056.
On what it was already a problematic development strategy building process, I picked a strategy in which I was interested to backtest.
it produces quite a lot of errors in MT4 backtester. See doc 6, attached. These errors occur only massively in some periods of the tick data (in others also occur but in a lesser degree). It seems however that if we have the bad luck to catch one of these periods, trading is highly impacted, as backtesting for instance from July 2007-June2019 the strategy results in only 282 trades, whereas if I trade from July 2009-June2019 the strategy trades 452 times.
So, in a nutshell 2 years less of trading time, results in 170 trades more, which is really strange.
I would not place this ticker here only by itself, but as the process has stumbled a lot in the first phase (see sqr_5056) I wonder if problems are connected.
Thanks.
Attachment Strategy 121138.sqx added
Attachment USDCAD_H4_Strategy 121138.mq4 added
Meanwhile I have the following extra information.
- I made the building process in SQX using data scope from 072009 to 072019. No problems reported.
- When I retest from 2007, SQX sets this strategy as "failed" as it seems there was a trade in 042009 that was unable to be closed for more than 2000 days and only closed in 2015. This means that until 2015 there was no other trade possible (I don't know why, as there are plenty of market conditions after 042009 to open and close trades, so I wonder why this trade was stuck there for 7 years).
- This explains why, when BT after 072009 the strategy has more trades because it didn't catch that 042009 trade with a 2000 days duration.
Bottom line: even if this now is maybe not to be considered a SQX specific error, let me challenge you with the following:
a) I've tried to run this building process in USDCAD 4H for many days in SQX The process returned a very large number of strategies with this kind of problem and also many profitable strategies in SQX but that don't trade at all at the MT4 tester (I will open another case for this here)
b) In the end, the real number of strategies that one could use (even before robustness tests) is minimal compared with the number of generated strategies. To find a "testable" strategy in MT4 BT without issues is like finding a needle in a haystack.
c) I've made the same process in SQ3, and the system returns a significant number of testable strategies without any of the problems described above.
d) in the many of SQ3 building processes I've made before, I've never found this kind of problems before (e.g. like one strategy be so sensitive to one trade only).
its obvious, that this kind of strategy is fake and cant be traded on real account
But in any case your comment misses the point. The problem here is not if this strategy is usable or not. The problem is why SQX generated so much of these "Fake" (as you call it) strategies. After a multi-day building process the number of testable strategies (after 1st round of build stage and before any robustness tests elimination) is minimal and so many strategies are non.testable.
Paraphrasing you: it's maybe so obvious that these strategies are not testable/tradeable that even a simpler tool like SQ3 can easily avoid to block them before they are placed in the databank, and by this increasing the level of complexity in handling SQX and the time lost throughout the development process.
Status changed from Waiting for information to Refused
Attachment strategy.jpg added
If you'll look at the results you'll see that it sometimes doesn't trade for several years.
The reason is that the strategy uses only Trailing Stop + Profit Target, and no Stop Loss.
See example in my screenshot - strategy takes Buy trade at 28.6.2004 that lasts until 8.12.2015 !
And it is because market is going down all the time since start of the trade until 2015.
So Trailing stop is never set, and Profit target is also not reached until market gets back to the previous levels in 2015.
All this because strategy doens't use Stop Loss.
You have to use either normal stop loss or time stop (Exit After bars) in your strategies, otherwise it will produce strategies like this.
Since SQ3 I've been focusing in developing strategies that favor existing trades with indicator values, instead of SL. I've developed thousands of strategies with that focus without incurring in this kind of problem. Maybe somehow SQ3 would filter our this kind of strategies ?
I understand this is not a bug from SQX, rather a feature, but my point is that SQX seems to allow/generate much more unusable strategies than the experience we had with SQ3. Some strategies don't even generate any trades in MT4 tester, when they seem to trade in SQX with no problem or warning. This places more work and burden on the shoulder of the user, when using a tool,which is already more complex to use than SQ3. To me personally is very annoying to go through the strategies that I can really use, like finding a needle in a haystack and it's also not a reasonable solution for the user, at the early stages of development, to go through te strategies and understand the nature of each one of them to conclude if they would have problems trading or not.That was my point, but as I said, I understand this ticker specifically is not a bug but a feature of SQX.
On the other hand, it gives you new tools to filter the strategies.
I just commented ask https://roadmap.strategyquant.com/tasks/sq4_5078 with example how you can se open drawdown and use it in filter.
As for the "errors" you are referring to - these are not errors, but log messages so that you know what is going on.