Target Trading

Discussion in 'Strategy Development' started by profitseer, Sep 21, 2002.

  1. let me try this one more time. Forget the numbers for a minute, just tell me if this concept has any value.

    Start with $10,000 and the following targets
    Max loss=400
    Average loss=200
    Average win=300
    hit rate=40%

    1.enter trade using coin flip or crossover or whatever I don't care.
    trade 1 goes against you, at -200 you take your loss, you have now hit one target.

    2. Trade 2 goes ok, you take a profit at 300 and now you have hit 2 targets.

    3. Now on trade 3, you see you have hit 2 targets, but your hit rate is now 50%. You have already determined that you only need 40% so you go into this trade armed with one important fact. You don't need a win. And you still haven't hit max loss so the no brainer is to put on a 400 trailing stop. It never sees the light of day and you get stopped out for 400.

    4.Trade 4. one problem. average loss is now 300 and target is 200. Same thing, you now have one important goal. On this trade, no matter what, your loss must be less than 200.

    So etc. When your hit rate is higher than target, you take more risk. When your hit rate is low you scalp. When your average profit is below target, you look for trades with higher profit targets (or you simply refuse to take a small profit.)

    The value is, you know what you need and don't need on every trade. (and once figured out can be programmed into any existing trading system and many more targets could be added or modified)

    Any good?
  2. I must admit I don't quite understand everything you are saying, but what I find interesting is that you try to look at everything kind of backwards. Usually, you have some sort of indicator or buy and sell signals, and you test how well those signals perform, i.e. you determine a hit/miss ratio, avg profit, avg loss, etc. empirically.

    What you are proposing sounds like you try to determine a profit target / stop price independently of the market. You also seem to rely on the outcomes of previous trades to make decisions about new trades. While that is not necessarily a bad idea you might want to look at it this way: If you want to gain an edge in the markets you have to focus on the markets. If instead you ignore what everyone else is doing and make decisions based solely upon your "money management rules", you cannot be successful in the long run.
  3. lobster, I already know how to read the market. Yes, looking backwards is exactly what I am trying to do. And yes, the whole foundation of any crazy system I come up with is money/risk management. What I don't know is if my reading of the market is even neccessary. What I don't know is if reading the market and risk management are exactly the same.

    I don't trade this stuff, after all, I have bills to pay. I just think about it.

    By the way, your thread on roulette strategy was the greatest thing ever written on expectency.
  4. <b>A Control System: Regulating Risk Rased on Recent Performance:</b> It sounds like profitseer wants to create a control system for choosing between two (or more) risk levels of trading. If trading is proceeding well (meets or exceeds win targets and well within loss limits), then higher risk trades can be taken. If the system has hit or under-performed crucial loss targets, a lower risk trades must be taken. In profitseer's example, being worse than a target causes the trader to be more conservative, being better than a target causes the trader to try for more risky trades.

    This system is not unlike the rule of trading smaller during the bad times and increasing trade size during the good times (Antimartingale position sizing). Both approaches seek the same goal: attempting to regulate performance to reduce long-term risk. Regulating the max loss on each trade seems better than regulating trading size where it is not possible to reduce size (e.g., you cannot trade smaller than e-mini 1 contract).

    <b>Multiple Targets = Messy Control System</b> The example system has 4 targets (technically it has 3 targets and 1 constraint, but who's counting). If all of these targets are totally independent of each other, then controlling them is easy. All you need to do is put on trades that increase the average win AND decrease the average loss AND increase the hit rate (LOL! :))!! Unfortunately, the outcomes are coupled -- a given trade will always affect the hit rate and might affect either the average loss or the average win. This coupling means that the system is ill-specified -- it needs some additional rules or trade-offs. For example, what should one do if the Average loss is over $200 (so more conservative trades seem advisable), the hit rate is under 40%, and the average win is a miserable $100???

    With 3 true performance variables (average loss, average win, and hit rate), there are 8 combinations of being above or below these variables. At the very least, one would need a table of 8 rules to define which type of trade to put on in each of the 8 respective above/below target combinations. A fancier system might use fuzzy logic or analytic optimization to set the next trade's risk limit as a function of the numerical difference between actual and target performance.

    <b>Big Bad Assumption: What's the Hit Rate?</b> The coupling of the target variables to each other has some unfortunate side-effects. This method assumes that the more conservative trades (e.g., those with under a $200 risk limit) have the same or better 40% win rate as the "riskier" trades. Are we sure of this? If the trades with a very tight risk limits get wiggled out with high frequency, then this system will go into a death spiral. Although one may be able to keep the average loss at $200 or less, the risk-tightening control rule may cause the hit rate to plummet. Likewise, trying to increase the average win may have the side effect of reducing the hit rate.

    <b>Multiple Targets = Infeasible System?</b> The discussion of hit rate raises a critical issue. Although we can always create a trading system that meets any ONE of the 4 targets in the example, it may be impossible to create a trading system that hits all of the targets of the system. With multiple targets and coupled outcomes, it may be impossible to meet all the target simultaneously. A cursory examination suggests we can always hit any 2 of the 3 main variables (average loss, average win, and hit rate). Although 2-out-of-3 ain't bad, its not sufficient to ensure profitability.

    <b>Bigger Issue: YOU NEED AN EDGE!</b> As Lobster points out, the key to any of this working is having some edge -- some means of knowing when to enter the market to create a good trade (it sounds like profitseer has such an edge). The trading setup must be intrinsically profitable at levels greater than those specified by the targets. Only where there is an edge is there the opportunity to control trading to meet a set of targets.

    Happy trading to all,

    P.S. The target numbers cited in the example are NOT profitable. In the long run, a $200 average loss, $300 average win, and 40% hit rate provide $0 average profit per trade because $300*40% - $200*60% = $0.
  5. traden4alpha, i think you're smarter than me.
  6. Traden4, It's the death spiral I was worried about.

    I'm well aware on any given trade I can't correct multiple problems, but the trader could correct the average which is the farthest out of whack.

    This system is not standalone per se, it is meant to be a risk management foundation.

    Once the trade is on, the trader uses the system to decide whether or not to hang on or take a profit or a loss.

    Thank You for discussing the concept and not the numbers. I just picked those numbers out of my head (because I know very well the breakeven numbers.)

    This is just the rough idea, let me work on that death spiral a little bit.
  7. Freeman


    Profitseer, in case you become interested in the numbers, I wrote a program that calculates expectancy based on probabilities you define. It's quick and dirty (I banged it out after reading a chapter in one of Tharp's books about his marble game) and lacks a couple of obvious features but you (and other forum members) may find it fun nonetheless.

    Since it is an EXE distribution will be a problem. Has anyone solved the problem of sharing untrusted files? ie, a trusted forum member with a file server that scans for viruses? I could mail it via hotmail (which has some kind of cursory virus scan), but as a general rule it's not a good idea to run untrusted EXE files.

    Maybe someone knows of a program like this already freely available on the net?

    More fun with numbers can be had with the random equity curve generator at hquotes. If I add any new features to my program it'll definitely be an equity graph with monte carlo and risk of ruin calculations.

  8. <b>Addressing Concepts, Not Details</b> I'm glad that you could tell that I was trying to address the concepts underlying your approach, rather than the specific numbers. Too many discussions devolve into simple arguments about the details -- missing the more interesting conceptual points that someone has raised.

    <b>Correcting the Averages</b> I like your idea of correcting the averages -- taking trading actions that have a maximum favorable impact on the performance measures that define profitability. Indeed, I suspected that your choice of numbers was intentionally breakeven -- that the point of the control system was to push each average to the profitable side of that example set of numbers. Interesting.

    To me, the problem comes down to understanding the following 3 issues:

    1) How changing one of the averages creates changes in the overall profitability. Or, to put it backwards, how sensitive profitability is to each of these trading averages (average win, average loss, and hit rate). This means calculating the first derivatives of the profit with respect to the averages.

    2) How changing the trading actions or rules creates changes in the averages. Or, to put it backwards, how sensitive each average is to the effects of a change in trading rules, trade management or trading parameters. This is an ugly problem requiring backtesting and/or analysis of trade trajectories.

    3) How 1 & 2 combine so we can deduce a favorable change in trading rules, trade management or trading parameters. (I won't say "best change in trading rules" because that is too hard. I would rather work steadily in the "better" direction that shoot for optimal) This is easy, if the data from item 2 are any good.

    <b>Item #2 is the real stickler:</b> It implies looking at how alternate trade parameters (e.g., profit targets, loss limits, trail stops, time limits, etc.) impact the averages and thus impact profitability. For example, a trailing stop helps preserve open profits, but too tight a trailing stop will cut profits short. Only by looking at enough trades (either in backtesting or in actual trading) will one gauge the boundary between a loose trailing stop that loses accumulated profits and a too tight trailing stop that prevents profits from accumulating in the first place.

    I do think that target trading has merits, the big challenge is getting a feel for how the trades behave when you change the exit rules. This lets you learn how to control the trading system to push the performance averages in the profitable direction. I'll be very interested in hearing what you come up with on coping with the threat of death spirals.

    Wishing everyone a high ( AvgWin*Pwin - AvgLoss*(1 - Pwin) )

    P.S. Thanx, Gordon for the compliment, but I don't know that it is true. Back when I was your age, I certainly did not know as much about trading as you do. We'll have to compare "smarts" when you get over 40.
  9. I was trying to post the html code for hquotes calculator, but was unable to.
  10. prox


    #10     Nov 1, 2002