P&F - System

Discussion in 'Strategy Development' started by Guitar-Fire, Oct 27, 2003.

  1. Hi,

    I'm currently thinking about a simple point and figure system for currency-trading via forex. In the beginning I was expecting it might to be easy to find a simple trend-following logic working on most pairss, but currently I'm struggling a bit...

    Anybody who likes to share his experiences ?

    Regards, Ralf
  2. agrau


    You've mentioned you are struggling a bit. That's fine and you are not alone. But it may help if you could describe at which part you are struggling:

    - scaling methodology (Traditional, Chartcraft, Log/Percent, volatility based ...)
    - reversal method (fixed number, vola based)
    - pattern definition
    - pattern detection
    - additional things you consider (relative strength, volume ...)
    - pattern based position sizing
    - back-testing the bunch
    - etc. etc.

    As ever so often, standards do not exist in TA. This is true for P&F as well. I would be willing to throw in my P&F knowledge, but need to know first where you are stuck.

    If you can describe your ideas, I could run them thru my P&F backtester and Monte Carlo simulator.


  3. Andreas,

    thanks for your reply.

    I'm testing different "frames" in EUR/USD, JPY/USD and GBP/USD like:

    5 bips/reversal 3 units,
    10 bips/ reversal 3 units,
    12bips/ reversal 3 units,
    24 bips/ reversal 3 units.

    I started with very simple strategies like:

    if marketposition<>1 then buy at highest(high,2)+trigger stop;
    if marketposition<>-1 then sell at highest(high,2)-trigger stop;

    I'm using 1 Minute Forex-Data,

    I'm aware off severall TS2000i-bugs and I believe to be able to avoid those.

    The Results are not stable over different "frames" or markets. I wonder if it's a essential part of P&F-strategies to work with profit targets.

    do you have any advice, or questions....

    Regards, Ralf
  4. agrau



    I am a complete TS illiterate, thus I don't understand your snippet
    but it leaves me wondering if you are testing p&f. Although p&f is most useful in detecting breakouts, it goes beyond this. A frequently used term in p&f is that it shows "the battle between supply and demand" which is why p&f-trading is mostly based on some patterns that exhibit this struggle of interests. I doubt a single line of TS code could catch this.

    Another thing I have noticed is that there are very few software packages that compute p&f charts as intended. In fact, I am only aware of Bull's Eye Broker and PNFscan who do so correctly. Both won't help you, as they are (to my knowledge) both just EOD.

    The two masters of p&f application, Chartcraft and Dorsey, will leave you helpless as well, because their work is (afaik) also only EOD, plus they add sector centric and sentiment views to their work.

    I doubt any of the commercially available software packages will give you the options at hand to really test p&f without major coding work. These packages are, to my knowledge, all based on bars or ticks, and this is what you have access to in their respective programming language.

    In p&f, a bar or tick is somewhat converted into columns of X and O - if a bar/tick doesn't lead to a new X or O it gets lost, it's considered "noise". So a truly p&f enabled software leaves you with the X and O information (most recent and past) and fire trigger events as new patterns evolve. For your testing, you would need to capture these events, which, in the simplest case, is a breakout. Which leaves "only" the definition open as to what a breakout is.

    What probably comes more intuitive are Renko charts, as they don't drop the time information, leaving the bar/tick concept somewhat intact, still using the idea of "noise reduction" as in p&f.

    My personal opinion on "standard" p&f charting is that it doesn't work anymore. Although, using proper money management, you could make some money using standard patterns, it won't make you rich. At least this is what my backtesting shows. Chartcraft and Dorsey add some twists in terms of risk management, but again this is not the holy grail.

    Best thing to do is probably to ask some of the many TS user groups and the Software forum on ET for some TS guru's to help you out with a p&f-starter-kit, then follow your own route. I promise p&f can be profitable, but it definitely is not, when used as a black-box system.

    I now may have created more new questions than answered. But what the hell :D

  5. Andreas,

    thank you for your effort.

    please allow me an example:

    - I used 10 bips as one X or O (a unit)
    - I used 3 units to create a change between X rows and O rows

    an upwards "breakout" is when you're current X row goes over the last two highest X-rows plus one unit (trigger) and vice versa

    the result is a pure reversal system

    I tested this over different parameters to get an indication of stability and the results were just OK

    I'm not sure if I should proceed with stops and profit targets ?
    Any ideas ?

    Regards, Ralf

    PS: from my experiences to "eliminate" the time factor might be big advantage of P&F
  6. agrau



    personally, I don't use price targets altogether. The formulae available for p&f price estimations have proved more often wrong than right during my tests. I see the logic behind (the longer it took to breakout, the stronger the move - or something like this) but there's more to consider.

    To me, the question comes back to whether a pattern or pattern combination shows some predictable outcome, you've called this stability. There are three sources of analysis regarding this question I know of: 1) the often-quoted study of Prof. Davis, 2) a re-make of this study done by a student in Dresden (Germany) and 3) a study to apply Genetic Programming on p&f patterns conducted at the University of Karlsruhe (Germany).

    Overall results showed that the patterns alone give you too small an edge to trade it alone. Having said so, it must be kept in mind that they all used the same standard chart settings - to a child with a hammer, everything looks like a nail.

    Coming back and trying to answer your question, I can say that a) using stops based on the patterns proved useful to me and b) using profit targets did not add value.

  7. Andreas,

    I like your "child-example". :)

    I'll try some stop strategies and will stay away of the profit targets.

    Thank you for your advice,

    Best to you, Ralf
  8. rognvald


    If you ar serious about P+F possibly the besttraditional P&F software is by Updata - you could also look at pfscan