I have a question about Monte Carlo. For example, we are testing a strategy like KnifeJugger, where transactions are made randomly due to lack of capital. In order to see different options, we use Monte Carlo, we get 1000+ equity options. Usually, they all turn out to be profitable, some with less profitability, others with more. But if we test the same strategy in WL6, where there is an option "Use Worst Trades in Portfolio Simulation", then we get a very unprofitable result. This option, if I understand correctly, should be one of the options that Monte Carlo should issue. But Monte Carlo does not give such a bad result as in WL6 with the "Use Worst Trades in Portfolio Simulation" option. Why?
And another question, do I understand correctly that for such cases (lack of capital) it is necessary to use the "Same Date Scramble" option?
And another question, do I understand correctly that for such cases (lack of capital) it is necessary to use the "Same Date Scramble" option?
Rename
Worst Trades is a peeking option that always uses the worst trades, every day. Being an extremely improbable outcome, you might see it once by running several trillion MC simulations.
Just to give you an idea, imagine one day there were 30 trades and the backtest could only put on 10 of them. All else being equal, what is the probability that you could select the 10 worst trades that day?
The statistics gurus out there could tell us, but I think it's 10/30 * 9/29 * .. 1/21 = 3.32833916042312E-08. If I'm right (I'm probably just close) you'd have to try about 30 million times just to get the 10 worst trades to come up in the same backtest for that one day.
Re: Same Date Scramble
I think it provides the most realistic probability distribution, especially for dip buyers, since it doesn't result in spreading trades across all dates in a backtest.
Just to give you an idea, imagine one day there were 30 trades and the backtest could only put on 10 of them. All else being equal, what is the probability that you could select the 10 worst trades that day?
The statistics gurus out there could tell us, but I think it's 10/30 * 9/29 * .. 1/21 = 3.32833916042312E-08. If I'm right (I'm probably just close) you'd have to try about 30 million times just to get the 10 worst trades to come up in the same backtest for that one day.
Re: Same Date Scramble
I think it provides the most realistic probability distribution, especially for dip buyers, since it doesn't result in spreading trades across all dates in a backtest.
Thanks for the answer. That is, if Monte Carlo is not turned off for a long time (30 million times), then there is a high probability of getting the most negative result :)
Your Response
Post
Edit Post
Login is required