Great video and explanation. I do have a question. I n the CVaroptimization, I'm not sure what are the alphas for. Could someone please explain htat to me? thank you so much in advance.
Yes, even I was wondering that because one of the main features of CVaR optimization is that the possibility of such reduction to linear programming does not depend on those alphas having a special distribution, such as a normal distribution; it works for non-normal distributions just as well.
As explained by Scott at 58:03 and further, alpha here is just an arbitrary variable that captures some degree of desirability of individual assets in the future. It could be historical avg. return as a proxy of expected return for example, as he did in the Markowitz experimentation.
alpha in this case would represent expected return of a strategy on some instrument k. (In most cases) a strategy return may be affected by instrument's negative returns, he adds in bottom n percent > max loss constraint But i think he could get rid of generated alphas in the code, derive alphas from the returns directly. overall he is a good lecturer 🎉
very helpful video!
You did sensitivity analysis at 42!!
Great video and explanation. I do have a question. I n the CVaroptimization, I'm not sure what are the alphas for. Could someone please explain htat to me? thank you so much in advance.
Yes, even I was wondering that because one of the main features of CVaR optimization is that the possibility of such reduction to linear programming does not depend on those alphas having a special distribution, such as a normal distribution; it works for non-normal distributions just as well.
As explained by Scott at 58:03 and further, alpha here is just an arbitrary variable that captures some degree of desirability of individual assets in the future. It could be historical avg. return as a proxy of expected return for example, as he did in the Markowitz experimentation.
alpha in this case would represent expected return of a strategy on some instrument k. (In most cases) a strategy return may be affected by instrument's negative returns, he adds in bottom n percent > max loss constraint
But i think he could get rid of generated alphas in the code, derive alphas from the returns directly.
overall he is a good lecturer 🎉
I just want to get good at online strategy games.