Position Sizing

[Oadmani]

I have been trading with a really small account and learning how to trade.

I am slightly confused. When I read books, they give example of a single time series on which position sizing is applied (say trade 1 contract at each buy signal). How do we back test and use Monte Carlo Simulation with multiple stocks? The way I have been doing it is, I have the same rules that apply to about 80 stocks I am trading in. So, suppose, I get a buy signal for a particular stock. I divide my total capital by the number of expected trades I would get at a particular time. In other words, on average, I want 100 percent of the capital invested (though this is obviously not the case all the time).So, suppose I have capital X, and I have six signals on average, then if I get a signal, I trade with X/6 on that particular stock. The issue is how to apply monte carlo simulation in this context, as the examples I read in the book are for a single time series. Basically, you trade one contract on every signal so how do you allocate capital between multiple stocks?

 

[jnbadger]

Good question. From what I know about Monte Carlo, wouldn't you have to do an MC simulation on every stock (while using the exact same strategy on each one), take the worst possible outcome for each one, combine them, and calculate your risk using that result?

Just throwing my 2 cents out there, and I'm looking forward to answers from those who are much more knowledgeable than me.

 

[xandman]

Dividing capital by no. of expected trades is impractical...almost ludicrous. There is going to be a very large variation in the number of signals generated across each issue and in different market regimes. Your thinking like a generic data analyst/programmer by freely changing the data axis of a given table of values. If you divide the total capital by number of expected trades and each stock will have a different number of expected trades, then you have your capital allocation.

Since you are trading a single model, then using a market sector weighting scheme to mirror the SP500 would make benchmarks comparisons more valid and useful. This is better. But, there are other risk based schemes.

Frankly, I think insights from performance attribution analysis is where the money is. I would do just as well having a monkey trade my account.

 

[xandman]

Good question. From what I know about Monte Carlo, wouldn't you have to do an MC simulation on every stock (while using the exact same strategy on each one), take the worst possible outcome for each one, combine them, and calculate your risk using that result?

Just throwing my 2 cents out there, and I'm looking forward to answers from those who are much more knowledgeable than me.

@jnbadger is on point. We summon the power of math!! ..to aggregate each data point point result thru each time step or just the ending value.

Are you really hand-rolling this or mouse clicking on an expensive piece of software?

I just bet on red or black. Life is too short.

 

[Oadmani]

Good question. From what I know about Monte Carlo, wouldn't you have to do an MC simulation on every stock (while using the exact same strategy on each one), take the worst possible outcome for each one, combine them, and calculate your risk using that result?

Just throwing my 2 cents out there, and I'm looking forward to answers from those who are much more knowledgeable than me.

I thought about this approach. However, it distorts the results, as inevitably there are periods in which there are no trades for a particular stock (the capital is not invested), whereas in my case, if it is not in stock X, it is very likely in some other stock Y.

 

[M4-1]

I have been trading with a really small account and learning how to trade.

I am slightly confused. When I read books, they give example of a single time series on which position sizing is applied (say trade 1 contract at each buy signal). How do we back test and use Monte Carlo Simulation with multiple stocks? The way I have been doing it is, I have the same rules that apply to about 80 stocks I am trading in. So, suppose, I get a buy signal for a particular stock. I divide my total capital by the number of expected trades I would get at a particular time. In other words, on average, I want 100 percent of the capital invested (though this is obviously not the case all the time).So, suppose I have capital X, and I have six signals on average, then if I get a signal, I trade with X/6 on that particular stock. The issue is how to apply monte carlo simulation in this context, as the examples I read in the book are for a single time series. Basically, you trade one contract on every signal so how do you allocate capital between multiple stocks?

you are wasting your time..and money..

the best way to make some money trading is to trade the markets that offer the best return for the lowest risk..in other words..you don't have to risk too much to make an acceptable amount..

forget aboit BIG money..that is childish dreaming stuff..think about consistent small profits..as..

100 pennies make a Pound