Kelly vs. Markowitz Portfolio Optimization

Подписаться 26 тыс.

Просмотров 14 тыс.

50% 1

Comparison of portfolio optimization using Markowitz (mean-variance) and the Kelly Criterion.
Portfolio Optimization and Monte Carlo Simulation
ssrn.com/abstract=2438121
Source-Code in R
github.com/Hvass-Labs/Finance...
MS Excel Spreadsheet
github.com/Hvass-Labs/Finance...

Наука

Опубликовано:

20 июл 2024

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 16

@jude3046 7 лет назад

Thanks this was very helpful for understanding how to optimize Kelly portfolios

@robertbond 3 года назад

A comment on slides #5 and #6: I think the statement that "variance (of returns) is not a suitable risk measure" (and as the numerical example impressively shows) is only true for discrete distributions. If you switch to continuous distributions (e.g. normal distribution), the claim that portfolio B is always better than the MVP can no longer be made.

@hvasslabs 3 года назад

It was a long time ago that I made this video, so I can't remember exactly what I said. I am currently working on research on new portfolio optimization methods. If I write another paper on this research in the future, then I will make a section about why Markowitz mean-variance portfolio optimization is wrong, where I assemble all the criticism from several of my videos. I will also show it for continuous distributions e.g. normal-distributions. The main problem is that variance measures the spread of the distribution and not whether a part of the return-distribution is actually a loss. You may want to watch these videos which also have criticism of variance as a risk-measure: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-CbkvOiQnpM4.html ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-DzTlH6ipx98.html

@nhoj277 7 лет назад

thanks that was a great lecture!

@whatitmeans 2 года назад

Really good video... I learned about Kelly Criterion 10yrs ago studying teleco in electrical engineering during signals processing course (John Larry Kelly Jr. was a student of Claude Shannon, the father of Information Theory), I found it amazing and couldn't believe it were not related to finance (I doing by luck my finance courses in the same term), even the Wiki page about Kelly Criterion were had only a few lines of text... now it have spread a lot but still Markowitz portfolio rules, even where is incompatible with modern theories like Black-Scholes Wienner Process models. I would like to know, if you know about a portfolio optimization model similar to Markowitz, but that instead of maximizing the expected value (for a limited variance), it maximized the mode of it probability distribution??

@hvasslabs 2 года назад

Thanks. I would suggest you look at my recent paper and video entitled "Simple Portfolio Optimization That Works!" and the shorter paper "Fast Portfolio Diversification" which only presents the algorithm without all the tests.

@k.alipardhan6957 3 года назад

Thank you!

@paperinvesting 4 года назад

I have been thinking of ways to supplement this.. regarding the returns, assuming you do not know the returns as a value investor i often average a company's DCF, Cash Yield and Earning Power Value instead and later compare it with the current share price to decide a company's rough potential returns. Can i use that method? I have been looking for a way to use the kelly formulae for my contrarian investment style to manage my portfolio. Any enlightenment will really help me. Can I also ask why you didn't use geometric mean and instead chose arithmetic given that we are working on compound rate to achieve the compound value of returns? Another question is todo with fractional kelly where you maximum return but also minimise losses. Does it improve returns?

@hvasslabs 4 года назад

The big problem is to make a roughly accurate prediction of the future returns. If you can predict the mean return somewhat accurately, and you can make a diversified portfolio, then it should do OK. When you are doing classic valuation like Discounted Cash Flow, then you are really just making a guess, and the problem with the Kelly formula, is that it will give a very large portfolio weight to the stock with highest estimated returns. So if your estimate is wrong, Kelly will invest heavily in the wrong stocks. The Kelly formula is probably more useful for more quantitative approaches than classic DCF estimates.

@catseng3949 3 года назад

so you mean in excel or calculating, I need to add constraints such as and ?

@OWAIS843 6 лет назад

How do you calculate the value of E

@hvasslabs 6 лет назад

It is the mean or average.

@OWAIS843 6 лет назад

mean or average of what? Do you mean average return of all the stocks i.e if you have 10 stocks in the portfolio you calculate the average of them? second question How does asset A has 3 returns e.g Google Stock can have 1 return not 3 for any given period.