...thenks for your replies. Compared to the wisdom in Grob109's post above, anything I say clearly is beginnerish. However, I have been backtesting with ESignal for a couple of years and think I know the pitfalls (and pratfalls) of that. I think it is axiomatic that if you are backtesting you plan to trade the system mechanically, even automate it. After all, if you were an intuitive trader and made money, why would you need to backtest? The often unspoken benefit of mechanical trading is that it is deeply therapeutic. Forcing yourself to take the indicated trades, often against "logic" and intuition, forces you to confront your demons and make an accomodation with them. Re ESignal, once you get a core trading code knocked out (their documentation is marginal), it is a snap to code up new ideas (optimizing them is another matter). You can easily test simple strategies such as is there really a Friday effect? should you buy a gap down? does a first 30 minute breakout work? While I am bored to death between the mechanical entries and exits, I try to devine meaning in the chicken bones and jot down (mostly worthless) ideas to test. I find that a failed strategy occasionally lead to something which has merit. I start out with the simplest possible rule set for the idea. I find that if that isn't minimum wage tradeable with commish and slippage but without more rules and optimization, it isn't worth taking the time to optimize. Unfortunately I find that the best stuff has five rules, which is time consuming to optimize since it has to be done iteratively and ESignal doesn't automate that. I check the optimization of my favorite system and two backup systems each weekend, which doesn't leave much time to delve deeply into the performance of my long backlog of ideas. I find that at best I can look seriously at one new promising idea a week. Best regards. - Mike
...I have found that stops and targets (and sometimes reverses) will improve a system 10-20%. Without getting into specifics, are your systems which are "good to begin with" already yielding >=60% accuracy? How many rules are already in the system before you try to optimize it? Is there a possible distinction here between predictive and reactive systems? Thanks. - Mike
i must not be understanding you well, because to me that sounds as self-evident as saying "can't do better than selling at the top". which is generally true. from your other post you seemed to imply finding the "failure point" was more discretionary than rule-based. if that's the case, how do you backtest it? genuinely confused...
Damir, I hear you... These were the same questions I posed to Girlpower a few pages back in this thread... I didn't get an answer. Maybe I wasn't clear in my post. Hopefully she'll respond to yours. To reiterate (1) how do you define a failure point , (2) how do you mechanically or discretionarily determine failure points, and (3) if you (Girlpower) say anything short of a failure point is mediocre as an exit, how do you know, that if you say that you can't back-test failure points. I find this while a great theme for research/discussion, a bit contradictory in the way they have been presented. Thanks, all CPTrader
Hi Damir, There are indeed rules based upon shapes, the problem is trying to define those shapes in a way that is loose enough for a computer to understand involves concepts that are sadly beyond me. shapes all based around lines, triangles, boxes and momentum. While they are easily identified visually, it is much more of a problem when it comes to all the variations that look pretty much the same, but actually aren't identical. Hi CPTrader Actually you did get an answer if you look back - maybe I didn't direct it specifically to you... As to the question of mediocrity. What I said was. BTW, It is possible to backtest compeletely by hand with a few pieces of paper and a slide-rule but it is very time consuming. And as I said before So, If I have both backtested the failure points by hand, and forward tested them for real, and then take a lowest demoninator for the more discretionary exits and further reduced that, then that would pretty much constitute realistic testing with pretty decent data wouldn't it? Kind regards Natalie
i think i understand now, thanks for the detailed reply. in the end forward-testing is more important than backtesting anyway.
Thanks Girlpower. It seems your implied definition of a failure point (as you have not precisely defined a failure point) is some break of a reisistance/support level that bounds some geometric shape on a price chart?? Please correct me or educate me..
Yep that's about right. They may also coincide with pivot points or significant levels from today or previous days etc. Charts are littered with geometric shapes but they are difficult to define mechanically. One day the software we use will be as developed as that inside our heads, but preferably without some of the bugs that exist inside our internal software... Kind regards Natalie
to reinforce the commentary made so far, their areways to define this arena of failure. Giving specific consideration to trend failure is a very important and far reaching set of considerations. It is like backing up in a trade from the condition of going out on stops. This is a bad way to exit; you give up so much. If you continue to back up you get to the trend boundary type exit. Channel breakouts and that genre. You preserve more profits by doing this. Backing up further you get to what girlpowers susgests: looking atformations and stuff within the trend profit movement. The two branches of this consideration focus on the channel characteristics. The primary characteristic is the the nature of the money velocity of the trade. Fast channel accumulate profits continually, medium channels are "on/off" channels; slow channels are "on/pullback (loss)" channels. This gets you far enough back to focus on the channel "traverses that cause the two parts of the profit characterization. It is hard to get people to become conscious of opportunities on this level of discussion. Most of it is foreign and almost always does not incite questions. The key "failure" that logic and software may be applied to is the "failure to traverse". What makes this such a deep and rich place to mine is that there are two directional traverse subsets in any money making channel. A person could superficially conclude that "fast" channels make the most money. If you continue to read and engage, you will find out thatthings are very different than quick looks and intuitive judgements provide. By knowing about the channel traverses in a "failure" context you get to some major stratigical break throughs. This stuff also puts nail after nail in the macro coffin. At some point in strategic financial industry practices, the pendulum is going to swing to a point of really pulling down all the potential profits the market proferrs. Fialure with regard to traversing channels takes you to a new level of analysis that is a level finer and more precise than the first two levels of analysis. The gross level which provides perspective keeps a person on point for dealling with the contorlling side of sentiment (the "buyer/seller market stuff); the middle level deals with trend analysis; channel land so to speak. What you and girlpower are bantering about, and it is significant, is the essential ingrediant for moving to continuous profit extraction form the market's potential. It is breath taking to evolve into this territory. It is the last part of the trilogy of trading to make money. I live in this space and you will get that in a few years. Failure to traverse, either way in the channel is where it is at. This consideration drives you as far away from stops and risk as you will ever find yourself. You read girlpower composing pictures of it. Magnificent. what builds in this space after it is first available to see. The usual. Optimizing the opportunity at all times. The surly unruly ogre that prevents one fomr getting focused is uncomplex but has everything to do with the clarity of thinking and the emotions of decision making. The fact is, that you cannot mix levels in analysis. Doing the Dr seus "biggering" often mixes levels into a stew of ingredients. I insist that you deal with the gross. confidently define it without question. Set it in place. Set your channel considerations in place as well by a complete characterization. Then , and only then, can you monitor and view the bar at hand, the NOW bar. You examine it in one context only: the context of the prior bar. Your view is a single measure of a single consideration. It is this: IS this bar continuing to keep me on the right side of the trade? YES or NO. Yes means there is no failure. You continue your strategy. There is no causal factor inducing change. None at all. The alternative answer leads you to a selection of possible actions. It places you at choice and as always the choices are defined and you go with the flow. wehere you are is in a plce where singular signals popping up from some half ass strategy to get an edge are not in the picture whatsoever. That stuff is how stew is made which continually mixes levels and fogs the possibility of making money. Until we get to where this thread got us, there is no opportunity to consider what I am typing. ET does not deal in ideas at all. It deals in responses when and only when an opportunity appears. Today's trading was in a context that ocurs once every two years. That is a rare thing. It is a context where a person states unequivacably that their construct is not functional. So here we get to NO. What NO means is that you are no longer in the trade in a correct orientation. You must change orientation immediately and repeat an assessment on the thrid level only. In effect, all you can do is reorient (reverse in trading terms) and test. You either get a YES or a NO. Viola! No one inET get to think like this, ever. Unless it is put before you for your benefit or for you to exercise Power and reject. The YES means that you have taken profits and completed an action to stop all risks of uncertainty and, then, continue to make another slug of profits. what I am saying is that I have taken you to a new place and you made all the money that was available to make by answering one question twice. Once from each side of the market phenomena present. No one could ask for anything more simple, comprehensive and 100% successful all the time. Let me say again what you did. you exited a trade, took profits and took on another trade. Lastly, in the retest you could get a NO after reversing. you exit and sideline. WHY? Because you have just proven their is not trend at present. NO money to be made and only risk to endure until something new happens. Et has not seen elegance often. By reading this you get to see elegance.. You get to see ONE MAJOR COCEPT that no one uses here at all. The concept is that to optimize making money you have to operate on several levels, and at no time can you mix the levels and perform analysis. The market is a synthesis of levels. I pry them apart correctly and deal completely and independantly with each levels. By doing this it is possible to be continuously successful. The neat news is that this kind of stuff can be done in many ways. and all of the resultinf approaches are mecahnical and can be put into software as well. I am providing wake up calls to a lot of people on a lot of levels. the pragmatic ones like provideing graphs in advance of the actions that do follow is more or less a wake up call using a blunt force insturment. This post is one where some people really get to enjoy a really positive "eureka" moment. If a person gets to the point of enquiry on how ot optimize taking profits on trend endings, then they are sitting in a place for a magnificent and major breakthrough. Enjoy. And have a nice look at something pervasive and elegant.