Methodology: Coin Flip

Discussion in 'Strategy Development' started by Pension_Admin, Jul 13, 2007.

  1. My experience with coin flip hasn't been so great and I also come to a conclusion that the market is not random.

    However, I still would like to know if anybody is using this strategy and that if they are profitable from using it.

  2. wave


    Use loaded coin and the resulting binomial distributions to your favor. You now have an edge that can be measured and traded with confidence.
  3. Perhaps you are flipping the coin the wrong way.

    <img src= \img>

    Following is a trading model for Microsoft Corporation stock symbol MSFT. The system buys if closing price is greater than the 200 day exponential moving average value. The system sells if closing price is less than the 200 day exponential moving average value. Position size = (2 % of account equity) / (10 times the 20 day average true range). Trading long positions only. Initial account equity $ 100000. Simulation uses 21.33 years of daily historic price data.

    Number of trades 59
    Total profit $ 487450
    Profit after subtracting $ 10.00 commission per transaction: $ 486270
    Greatest draw down is 0.0792 (7.92 per cent).
    Cumulative Annual Growth Rate (CAGR) is 22.80 per cent.

    The simulation shows lots of losing trades. Losing trades are part of trading. I know of no trading system that wins every time.
  4. Thanks for the reply, but I still think the coin flipping method may not work.

    In the case of testing on MSFT, the test itself may be survivor biased. Also, it would be interesting to see how it compare to buy and hold.

    Also, what risk/reward do you use in your system and where do you place your stop loss?


  5. What's the logic behind division by 10 times 20 day ATR?
  6. A guy posted some test results of an unusual random-entry, random-exit system. Its similar to flipping coins. You randomly choose a trade entry day and you randomly choose a trade exit day. But you don't randomly choose a trade direction.

    1. Use a pair of moving averages to determine the main trend: if fastMA > slowMA the main trend is UP, if fastMA < slowMA the main trend is DOWN.

    2. (The day after you exit a trade): generate a new random number R. Wait R days (you have no position either long or short) and then enter a new trade at market on the open. Here's the trick: if the fastMA is above the slowMA, the main trend is UP and you enter Long. If the fasMA is below the slowMA, the main trend is DOWN and you enter Short. You don't care when the MA's crossed over each other, you just look to see which one is higher on the randomly chosen entry day.

    3. (The day after you enter a trade): generate a new random number S. Hang on to your existing position for exactly S days and then exit at market on the open. Go to flat. Notice that you have no Stop at all. Or some might say, you have a "time stop" but not a "price stop". Also notice that once you enter a trade, you totally disregard the moving averages. All you do is stay in the trade and count to S.

    4. Repeat. After each exit, generate a new random number R and after each entry generate a new random number S.

    In essence this is "Trade with the trend" but with random timing of entries and exits.