Which setups are more attractive?

Discussion in 'Risk Management' started by elitetradesman, Oct 15, 2012.

  1. We can agree I do not live in your neighborhood at any time of the year.
     
    #21     Nov 16, 2012
  2. rknas

    rknas

    Go with the one with the lower win%.

    When you do real execution, you are bound to miss trades/execute less than optimally/etc...The one with lower Win % will be less forgiving if a trade or 2 is messed up.
     
    #22     Nov 17, 2012
  3. This is an interesting one. I started analyzing it with the expectation that I would get an unambiguous result from it, but it came out not so clear. So, let's assume that each setup (let's call them setup A and setup B) had triggered 100 times. This is what would happen:

    Net Profit: $5500 for both A and B (draw)
    Expectancy: $55 for both A and B (draw)
    Profit Factor: 1.3 for A, 2.22 for B (B wins)
    Standard deviation of trades: 420 for A, 317 for B (B wins)
    Kelly Criterion: 12.7% for A, 5.5% for B (A wins)
    System Quality Number (aka Van Tharp): 1.31 for A, 1.74 for B (B wins)

    So, discarding the draws, setup B beats setup A on all metrics except Kelly. However, Kelly suggests that you can trade setup A with 2.3 times the size of B, so in fact, setup A may be preferable to B.

    As a side note, both A and B have highly skewed distributions, which are way far from normal distributions, so something like the omega functions (which is a relatively new way to evaluate the performance) is more appropriate.
     
    #23     Nov 27, 2012
  4. achilles28

    achilles28

    Take the higher win rate. The lower win rate implies a failing/faltering/deteriorating edge. Trade that, and it's likely to blow up.
     
    #24     Nov 27, 2012
  5. If the target is 100 times the stop loss. it is clear that the stop lose would be triggered more often. I would like to say that one in 100 trades would reach the target but I am a bit hesitant.

    The same way, a target 10 times or even 2 times that of the stop SHOULD hit the target 10 and 2 times more often then the stop.

    SHOULD, If this is consistently not true you can graduate from ET with honors.

    Also it is important to differentiate robot trading form discretionary trading.
    And to differentiate exits that based on the same idea of entry's vs a system that say enters on an indicator and exits on volume or time.
     
    #25     Apr 6, 2013
  6. Visaria

    Visaria

    Same number of trades with each set up? Then trade number 2, profit factor higher, total ev the same.
     
    #26     Apr 7, 2013
  7. Interesting to see how these various measures stack up. Thanks!

    Although both strategies (A & B) have the same expectation, B requires more capital to trade (as WideTaliz pointed out earlier ... "larger drawdown"). In my view, that would make B less attractive.

    This is perhaps easier to see if B is replaced by a new C (which is just an exaggerated version of B) as follows:
    - win rate 0.1%,
    - potential profit $100,000,
    - potential loss is $45.05.
    Expectancy is still $55.

    The problem with C is that on average a winning trade comes once every 1,000 trades, so on average you need to sit through a losing streak of 999 trades (on average). i.e. average drawdown will be almost $45,000.

    In the case of A, a winning trade comes on average once every 0.55 trades, and on average you need to sit through a drawdown of $183 (45% x $407) between each winner.

    So, on average, for A you need to sit through a drawdown of $183 between winners. For C you need to sit through a drawdown of $45,000 between winners.

    Both make the same average amount per trade (i.e. after 100,000 trades, they should have made approx the same amount, i.e. $55 x 100,000).

    But you need a lot more capital to trade C in order to sit through the drawdowns between winners.

    So better to trade A (you make the same as C but need less capital).

    C is just an exaggerated form of B. So the same argument applies when comparing A and B.
     
    #27     Apr 8, 2013
  8. IMHO, standard deviation is not a sensible way to compare these two strategies. As s.d. is calculated from variance, which is the square of returns, losses get treated equally to wins.

    B's s.d. is low mainly because 9 out of 10 trades have the same value ... but these 9 trades are all losers.

    Take a system that loses $50 100% of the time. This has an even lower s.d. than A or B. Is it s better system than A or B?
     
    #28     Apr 8, 2013
  9. kut2k2

    kut2k2

    SPS == k*W

    k(A) = .55/407 - .45/433 = .0003121

    SPS(A) = k(A)*433 = 0.135

    k(B) = .10/50 - .90/1000 = .0011

    SPS(B) = k(B)*1000 = 1.1

    System B is much more attractive.
     
    #29     Apr 9, 2013
  10. Visaria

    Visaria

    What does SPS stand for?
     
    #30     Apr 9, 2013