systems that work

Discussion in 'Educational Resources' started by toddmay, Apr 27, 2005.

  1. Just to be clear about what I am saying:

    I am not talking about edge as in casino gaming or market making. When I say that an 80% edge exists in a market, what I am saying is the following:

    A. I have identified a non-random behavior
    B. I have analyzed the target market to verify that this behavior repeats within a temporal window X times per week or month.
    C. When the behavior occurs, 80% (or "X" percent) of the occurences result in tradeable setups that would be profitable (net of expenses).

    Once again there are qualifications. First, these random behaviors exist when a market is exhibiting stationary behavior. Second, when markets exhibit random behavior that same setup may have no edge.

    Also, most of the random behavior that I have isolated is related to economic reports or to programatic trading. To the extent that economic reports may be delayed or postponed, the behavior may be affected. Program trades often excute at varying times within a temporal window. This affects the absolute number of opportunities and often the profitability of the setups as well. In other words there is an attrition rate as to setups and that is why you (I) need a high percent edge in order to make money.

    Good luck,
    Lefty
     
    #41     May 1, 2005
  2. Lefty,

    Perhaps you do not understand the definition of true edge, and it is you that should "think" prior to posting. The way you illustrated it, you are incinuating something mathematically impossible. An 80% true edge would indicate that you would quite literally be a billionaire in short order; which I very much doubt you are. There is a difference between accuracy and true edge my friend. An edge of 80% would suggest you are actually "right" 40% over random .5+.4 or 90% of the time. This is a load of crap unless you are selling cheap premium on OTM options. OTM premium can be accurate 90%+ with extreme risk and terrible r/r. I do not believe however that you are referencing selling premium, so I will assume the following: You would be insisting that you are making something at least 2:1 correct? Okay, 90% accuracy on 2:1 r/r would mean you could risk at least a third of your account at a time and still quite safely double your money every, umm, let's say hour.... Are you making 600-700% every day? Didn't think so. If only it were this easy... Did you mean something else Lefty? Now if you are a market maker/specialist, whatever, you make money every time, but this has more to do with your situation and less to do with any "edge."
     
    #42     May 1, 2005
  3. Well you make so many mistakes in your assumptions that it is stunning. First I am not your friend. Second I am the one doing the research. Third I have defined my terms clearly. Frankly I am not interested in your opinion of whether it is possible or not. I am interested in trading my setups and going home with profit.
    Please do not change a thing based on my comments.

    Thank you
    Lefty
     
    #43     May 1, 2005
  4. Ohhh - war!:(
     
    #44     May 1, 2005
  5. duard

    duard

    Lefty,

    I understand your points. I'm pretty sure I even know at least one of the "non-random events" you are referencing.

    BTW I did perform your "detrended analysis" and it does help with entries and exits.

    Thanks
     
    #45     May 2, 2005
  6. ptunic

    ptunic

    Just for humor's sake I will take a stab at identifying some of the potentially faulty assumptions:

    1. Just because the system is intraday doesn't mean it can trade once an hour. There are many "intraday" systems that trade only once a day or even once a month, but they are still intraday because they are out of the market before the close of each day.

    2. Trading with 33.3% risk per trade is rather high. Even with a win/loss ratio of over 80%, there are still reasons you may want to trade with lower risk (by lower I mean in the 1-5% risk range). The first problem is just because the strategy has worked for years doesn't mean it will work tomorrow for sure, thus you need some buffer so that even if your system stops working, you can stop trading your system with relatively small drawdown. The other potentially faulty assumption here is that the distribution of losses is evenly spread out. If it is common for the system to have 3 or more losses in a row, that alone would make trading 1/3 of equity size imprudent.

    3. There are liquidity concerns well; granted ES is highly liquid but I believe the word "billionaire" was mentioned and that amount of capital might be difficult to trade intraday.

    -Taric
     
    #46     May 2, 2005
  7. OK:
    Now people are starting to think a little bit. First about the "I am not your friend" comment. I do not mean that to be combative or insulting. We are not friends, we do not know each other. That is all I am saying.

    Next I have commented on the reason why you need a high "hit rate" in order to make money. Others are starting to think about it and make sense of this statement. You cannot trade all setups. Also a setup can simply stop occuring for periods of time. I have suggested why this occurs.

    As to the comment about detrending. It is the cornerstone of my work. Detrended data provides a lot of valuable information. You get to see how volatility changes over time. If you process the detrended data properly (Chi-Square for instance) you can see when and where price cycles from stationarity to random. Finally if you are skilled enough you can identify "temporal windows" where specific kinds of opportunities exist. To give a hypothetical example, lets say that between 11-11:30 EST, your research indicates that you should wait for a move of size "n" ticks. If you see that move expand to "n" + 2 ticks, you then have an 80% probability that the move will continue for more than 2 points.This is precisely the kind of data that you can recover from detrended intraday prices.

    The only qualification that I have offered (several times now) relates to stationary data sets. In other words, the challenge is for traders to find some way to identify when price is stationary and when it is random. In my opinion, if a trader is risking money during times when price is randomly distributed, they are gambling pure and simple, and that is why retail traders lose money. They keep risking the same money without knowing that their edge has disappeared. Because they don't have the skills to figure it out, the way the "learn" is by see their system fail (after the fact).

    I hope this helps someone.
    Lefty
     
    #47     May 2, 2005
  8. Your stab missed badly...

    1) Even if it's once a day, such a system would have no reason not to double an account at least twice per week, assuming 90% accuracy and at least a 1:1 r/r.

    2) His statement was 90% not 80% as edge is not the same thing as accuracy. If you study optimal position sizing techniques, one would be a fool to risk anything less than a third of their capital on a system that had a 90% hit rate with a 1:1 or better r/r. One third would actually be extremely conservative for such performance as the odds of ruin would be .000000001 (.1^9). Not very likely is it? 33% of your account each time it losses, is dynamic and would take 9-10 losses before your account becomes untradeable.

    3) If he is not scalping, liquidity would be of no concern on the ES.
     
    #48     May 2, 2005
  9. Allright:
    I have to thank you. I am really learning to have patience, and I have set that as my goal.

    You are missing the point. It is probably because I am clumsy in my expression.

    Here is why you cannot risk a third of your account on a system that has an 80% accuracy.

    Price does not always exhibit stationarity. Instead, it cycles in and out of that state, sometimes randomly distributed moves, sometimes stationary moves. I realize that the concept is hard to get your head around. Take your time, I am glad to respond to additional comments or questions as time permits.

    To give some idea of the opportunity rate, I trade about twice a day. My goal is to put on one trade and hold to end of session. In order to hit that target goal, you have to identify what type of market you are likely to have early in the session (within the first hour). Again detrending previous data long enough will give you a clue as to when this might occur.

    Lefty.
     
    #49     May 2, 2005
  10. Lefty, I am no stranger to statistics and yes one can risk 1/3 and relatively safely, despite binomial/guassian distribution. Even a fair coin will rarely see 10 losses in a row ("random" environment). 80-90% hit-rate will increase your account size a hundred-fold by the time you experience such an event. Statistically speaking, with an r/r of anything better than (Kelly Criterion and variations) 60%, one can risk 40% without much risk of ruin, that is if targets and losses are of the same size and sizing is dynamic. Now unless your system has some ungodly z-score dependency, ruin is statistically highly unlikely.
     
    #50     May 2, 2005