Had SQ for couple years, super good news

Start below in next post. Thx.

A Mark from Germany was discussing with me, if it can help anyone. I paste the email thread here, start from the bottom and read going up. (Neither of our trials means thats the way to go)
———
2015-02-18 17:14 GMT+01:00 Jerry B :
Yes on random, then genetic.
Yes on Improver, “add” for entries, “”add or replace” for exits.
I do RT in retests after 3rd step of Improver-runs, usually find its ok for good equity results, but my settings to see ok RT graph is probably not as varied as most would use, just a few not so wide, I suppose I do a subjective look to see that 95% has some respectable amount and the different colored lines don’t spread apart too much.
—
Truly, Jerry B

On February 18, 2015 6:37:47 AM PST, Mark wrote:
Hi Jerry…
thank you for the quick response.
Let me get this straight… you generate first many random strategies, which you are using afterwards as bese for genetic generation? After that you enhance exits and/or entries with the improver tool?

Are you doing some special robustness test checks to avoid curve fitting?

Mark

2015-02-18 15:04 GMT+01:00 Jerry B :

Hi Mark,
I use an i7-4770 and like 30minTF bar-open-only (its fastest and most stable strategy) and ecn/Dukascopy data (tickdownloader). Do random for days, followed by genetic, followed by Improver to often get better dd and maybe even profit better. Repeat Improver as needed. This should get you somewhere. Doing one side only (longs) on the above if you wish concentrated effort in such first, especially attempts with other than eurusd, then when satisfied begin adding the other side, but may need edit SOMETHING in on short side first (like no-Sunday) so that “add” in Improver makes sense to add to something.
Good luck, let me know how it goes! 🙂

—
Truly, Jerry B

On February 18, 2015 5:28:53 AM PST, Mark wrote:
Hi Batch…
I just read that your note in the SQ forum:
https://strategyquant.com/forum/topic/2837-had-sq-for-couple-years-super-good-news/

I am also a struggeling owner for years, but I just can’t get reliable EAs that manages to live up in the future.
Would mind to share some thoughts or settings about your apparently successful approach?
Maybe I just need some fresh input… 😉

Thanks from Germany,
Mark

Hello Batch,

how is it going with your strategy? Are you having a success?

Just running forward test couple times a day, so far so good. 🙂

Here is one “created” between Leonardo in Italy & myself run on live acct for half a year:
http://www.myfxbook.com/strategies/eurusd-1h-steadygrowth-2008-2014/75092
Click on “forward test” to see live real cents acct.

Hello, Batch,

Your strategy looks very good.

How did you get that? only do random for some days? then genetic, improver?

Would you like to share your experience about using SQ in more details?

Thanks.

Hi Batch,

what are you doing to prevent curvefitting?

do you use the Robustnesstest?

Without Robustnesstest the strategies are not robust and I expect curvefitting.

Remark:

The Robustnesstest filter not all curvefitted strategies, but a part of it.

thomas

Of course I use RT section, who wouldn’t. But obvious thing is, its quite subjective as to settings and interpretation. Also having it test good on another broker’s backdata is another sign its robust/dynamic. The less I have to do optrun on mt4 (this other broker backdata) is another positive sign.

Since I’m getting queries as to how this was found, I wanted to point out this good result might not be findable w/o a higher end computer (i7-4770) with 1min resolution (its amazing how fast an i7-4770 could search at this resolution) and selecting all exit means to let SQ search for best ways (something on a usual computer type I didn’t know could be done!) and let it run for days ‘n days on random then genetic.
Hope that helps.

THanks for sharing.  Encouraging to see your success!