In this email(s) to Pipcop (James) I arrived at some interesting conclusions during Invisible_FX_Landscape EA development that not only will help future EA designers (quite probably?) but also help to unload pipcop's forum weight of needing do server operation of all the EA's in the world for first post showing. However, sir_Pipcoip is extremely busy of late (so I understand and don't blame him) so he couldn't really examine the basis of the letter, so as recommended by him I paste the letter to my fellows at pipcop.com in this new thread for some "back 'n forth" yammer about if you all wish. OK? great.
jb
================================================== =
hi James,
RE: my hopeful EA development, it's been going relatively scientific this time.
Check-out last post of thread concerning (
http://www.PipCop.com/forums/general/135-possible-future-new-ea-invisible-fx-landscape.html), would like to know your thoughts re: this point brought up in last post there re: of way to eliminate 3 month demos in lieu of dirtectly to small live$$, just am interested in what you could possibly think. No matter for this EA though as I'll be doing a $100 micro account asap anyway and post the publication on this thread, just thinking for other EAs to arrive on the scene sooner this is something to contemplate?
Best,
jerry
--------------------
followed-up immediately with this letter:
------------------------
Plus it forces authors to be more scientific "from the get go".
Exceptions to rule rule are ANY (!!) hidden Martingale techniques (!!).
Even RoboMiner if done with dynamic TP's and SL's at bar-closing and microscoped live demo against back-test (MT4 has logs too that can be used to show if similar results occur from test to live!!!!) would be able to be promptly be brought to the forefront in a live test! But they use hard values which will NOT be replicated by back-tests due to tick movement! (you even elucidated this to me when I first came around with my first flimsy EA attempt, well, I'm employing my past engineering principles of late now! LOL). And even now Robo has no excuse to not know what the worst case scenarios WILL be because I have one MT4 that strangely in unknowing-as-to-why-manner has downloaded 1Min back-data from 1999 to now! So that 1-1/2 year ago range movement, Jerry of forex-goldmine or someone could run it, but Jerry would need to comply with above "at-bars-end" too in order to be candidate. Yea, they said they have run it from different starting points for a good amount of time, but this is not having correct-captured-at-end-of-bar, as that is more important than worrying about what different beginning DD there could be from such different begin back-test points. So overall, EA's could be affirmed and built correctly in my opinion "still" (my opinion still, to be seen of course). As for optimization over 3-months or 1-year, as long as it's giving parameter values that show most of the time EA performing "in-the-black", it has same percentage chance I think in live therefore, given wise stop-loss to avoid any major dives in market.
Even David.Jukis' P.SAR if done with some right-brain at first as to what additional indicator to tell when best to enter primary indicator trade monents and optimized would probably be a good EA, if done with all the above.
So I think in the end, not enough engineering thought has gone into EA's for BOTH vendors at websites and individuals availing EA's...... yes, like myself in the past, we're too sold on the beginning looks of it all (greed), and we're all throwing things up at you to see what sticks! Well, that's not the way to go and I "feel" (oops! LOL) that the route is being discovered as to the engineering/scientific correct route (I was electronic engineer for 11 years when younger, am 59 now), and this is what I'm finding after digging into all this more.
Perhaps these two letters to you should be copied and pasted on that programming thread? Well I post ONLY if this at least makes partial sense to you, for if no pipcop.com this all would not have progressed this far me thinks for sure!
jerry