I agree that it's well-specified.
Markowitz portfolio theory to balance option portfolio?
- Thread starter Gustaf
- Start date
Steven.Davis:
One flaw with the Markowitz optimization for stocks that Ive found out is that it focuses on including stocks that already have gone up alot through the sampling period. That is abit counterproductive.
In this model the "reward" part is risk premium, so I hope I have eliminated that flaw.
Whats your go on that, that it picks stocks that have already increased in value. Ideally we wanna by at a low average price.
I also found mixed integer programs to be quite finicky, sometimes it just doesnt solve. It was very hard to find a quadric integer solver btw...
Br Gustaf
One flaw with the Markowitz optimization for stocks that Ive found out is that it focuses on including stocks that already have gone up alot through the sampling period. That is abit counterproductive.
In this model the "reward" part is risk premium, so I hope I have eliminated that flaw.
Whats your go on that, that it picks stocks that have already increased in value. Ideally we wanna by at a low average price.
I also found mixed integer programs to be quite finicky, sometimes it just doesnt solve. It was very hard to find a quadric integer solver btw...
Br Gustaf
The round peg comment was not meant to be insulting, but more to emphasize that you're solving for your desired outcome in too complicated a manner. Trying to use an approach to portfolio construction with the derivative is the round peg issue.Quote from Gustaf:
I wanna sell OTM puts and if assigned sell OTM calls (as in the other recent thread).
While doing so, I like to optimize for the lowest possibility of stuff going wrong at the same time.
What is so roung peg about that?
Br Gustaf
Your overall aim is fairly straightforward. I'd suggest you generate your universe of long equities (however you like to do this) and then proceed to alternately short puts and then calls as you describe. Your p/l will depend almost entirely on how correct you are with your equity choice, not how clever you are with the choice of the short put or call strike. Just remember that you are are more or less long delta with your approach as the underlying moves around. It's a bit contrarian-more cautious after a big move up and subsequently more aggressive after assignment on a big move down. If that's what you want, then go for it.
My previous post was meant to encourage you to look at options as a way by which you can take a long equity portfolio and modify risk and even enhance return if lady luck is on your side. The aim with this suggestion is to try to sneak out with the decay or make a calculated bet on the mean reversion of vol. This type of option play is not typically the way that big girls trade options because their capital would be tied up in the underlying, but it is an approach that a traditional long equity trader can use. Gearing and decay are the attributes you want to optimize. You're still going to get smacked in the nose in a big downdraft, but less so than with an exclusively long delta portfolio, whether that portfolio is a collection of short puts, synthetic or otherwise, or the underlyings. In a big downdraft, an equity trader's correlations will all go to unity anyway, so you can't win on that score. The much more common scenario of modest swings around some mean price is where you'll make a few extra dollars.
One idea I had off the top of my head might be to do this exercise:
Instead of making a typical covar matrix based on component price changes, instead prepare the data going into that matrix as if each component had underlying options. So in the married put example, the return of the put is factored in, and you are looking at aggregate position returns. That way, you generate a covar matrix that factors in each option's impact already, and can do a vanilla optimization.
On the other hand, you'll have to adjust your E(R)'s of components to reflect the option premium.
If you want to get fancy, rather than modeling constant vol / black scholes valuation of the offsetting option, you could either get actual option prices to factor in the changing vol situation. But my thought is this -if you prepped a covar matrix using data through the fin. crisis, a lot of aggregate positions that were deep in the money wouldn't have been too affected by changing vol profile, and in aggregate, you'd see blown out correlations (super low) between assets that were DITM versus those far OTM.
Now that I think of this, maybe junk idea. But would be fascinating to model just to see the impact on the covar matrices given the presence of offsetting married puts.
Likely the covar matrix output would just be a function of the in-the-moneyness of the respective assets... An entire DITM portfolio would become cash. A half DITM and half far OTM would become a portfolio of effectively fewer assets. etc... What would be instructional is to view how the covar matrix changed and see how the optimization itself changed as assets changed from OTM to ITM, etc.
Instead of making a typical covar matrix based on component price changes, instead prepare the data going into that matrix as if each component had underlying options. So in the married put example, the return of the put is factored in, and you are looking at aggregate position returns. That way, you generate a covar matrix that factors in each option's impact already, and can do a vanilla optimization.
On the other hand, you'll have to adjust your E(R)'s of components to reflect the option premium.
If you want to get fancy, rather than modeling constant vol / black scholes valuation of the offsetting option, you could either get actual option prices to factor in the changing vol situation. But my thought is this -if you prepped a covar matrix using data through the fin. crisis, a lot of aggregate positions that were deep in the money wouldn't have been too affected by changing vol profile, and in aggregate, you'd see blown out correlations (super low) between assets that were DITM versus those far OTM.
Now that I think of this, maybe junk idea. But would be fascinating to model just to see the impact on the covar matrices given the presence of offsetting married puts.
Likely the covar matrix output would just be a function of the in-the-moneyness of the respective assets... An entire DITM portfolio would become cash. A half DITM and half far OTM would become a portfolio of effectively fewer assets. etc... What would be instructional is to view how the covar matrix changed and see how the optimization itself changed as assets changed from OTM to ITM, etc.
Quote from Gustaf:
"Oh Yes they might!"
I cannot imagine a worse timing trying this the last two weeks, that correlation stuff is just bullshit, everyting dropped.
Wasn't that mentioned earlier in this thread?