Oct

11

One wonders what that interesting personage would say about our inability to defend the Libyan compound and the amazing staged efforts of the militants to overcome 4 or 5 lines of our defense including the charted airline from Tripoli.

Chris Tucker writes:

Upon boarding the shuttle at LaGuardia bound for DCA yesterday I heard a very distinct, deep, German accent coming from the first row of seats, which could only be coming from one person. I looked down and saw Henry Kissinger chatting with his neighbor. For some reason I had forgotten that he was still living and was a bit taken aback. He looked quite old but in good spirits.

Vince Fulco writes: 

I am into the early chapters of Lone Survivor, a book written by a Seal Team member re: his early 2000s experiences in Afghanistan. I had no idea how hamstrung the boots on the ground were by rules of engagement (ROE) made thousands and thousands of miles away from the battle. He attributes these unrealistic rules to the direct loss of 3 of his squadmates. It parallels the stories of cruise missile operators with OBL in their sights. Regrettably they were unable to fire unless they received approval from then POTUS Clinton who was more worried about his own selfish pleasure with Ms. Lewinsky.


Comments

Name

Email

Website

Speak your mind

6 Comments so far

  1. Ed on October 11, 2012 9:21 pm

    This is off topic but I did not know where else to ask. I have a question related to back-testing methodology.

    My usual practice is to test for significance by benchmarking a holding period return that comes after a signal vs. all holding period returns.

    But what if the holding period is variable? I am looking at group of ideas that on average hold for less than two days, at times hold 1 day, and other times longer (very infrequently)over 4 days.
    I am not sure what to benchmark this against to test for significance. Would using the average holding period be a defective approach, or is it reasonable?

    Thanks in advance if anyone cares to respond.

  2. vic on October 13, 2012 10:49 pm

    take the opposite pattern and see how it compares. simulate the random distriution of subsequent moves compared to your event with similar duation and variability about that duration. vic

  3. nkonko on October 14, 2012 10:16 pm

    So the montecarlo solution would be like this:

    Step 0. Determine the length of the trades for the backtested system. For example if your system made 2 trades that lasted one day, 7 trades that lasted 2 days, 5 trades that lasted 3 days then the length of your trades is (2,7,5, ). This vector of trade lengths is then fed to the next step.

    Step 1. randomly draw days in consecutive blocks of such lengths. Add up the profit from such days.

    Step 2. Repeat Step 1 10,000 times and keep track of the 10,000 distinct P&L figures from this random trading and sort them in descending order.

    Step 3. Take the P&L from your backtest and see how it stands in the random trading distribution. For example if your backtest is ranked 905 within the 10000, this can be considered significant at the 10 percent level(i.e. it is in the top 1000 of the 10000 trades).

  4. Ed on October 15, 2012 10:51 am

    Vic and nkonko - Thanks a million

  5. david higgs on October 22, 2012 12:07 pm

    why don’t you ask Marcus>: marcus luttrell at facebook.com?

  6. dmi on November 5, 2012 1:37 pm

    Thanks for valuable info! I have the same question as Ed, just wanted to ask one more thing. Sampling is done without replacement from the full dataset including the days when the trades were done, right? Like in usual Monte Carlo simulation.

    Thanks!

Archives

Resources & Links

Search