
But I let it resolve to optimal sharpe ratio and simply display the beta’s as derived from MDYG (SP1500).
So based on CAPM Expected Return (Average Risk Premium for past 5 years is .0142 (1.42%), the CAPM return is 4.33% + 1.42% * Portfolio Beta of 1.00116592, which comes out to be 5.75% for next quarter.
A different forecast, one based on Markowitz simulations has 9% for next quarter.
Another forecast based on an expected return factor model forecasted results using a model that has 13% MAPE, the weighted forecasted return is 13% for next quarter (i.e. 13% +/- (13%^2) (i.e. 13% +/- 0.0169%)
What’s frustrating is knowing I hit the ball out of the park when it comes to CAPM portfolio’s and Markowitz, but to know that those in academia that actively trade are not fans of the material they are hamstrung to teach. So I get various strong opinions about what works. Very cult of personality about methodologies, but not me. I’m open to trying as much as I can just for the opportunity to learn.
The Inefficient Stock Market is a gold mine in terms of what factors to look for. I’ve been doing my own research (FRED data, commodities, foreign exchanges, indexes, sectors, SP1500 prices, fundamentals, financial statements, Critiques of Piotroski, French Fama 3 and 5 Factor Models, Arbitrate Pricing Theory). The book suggests improved/revised factor models using a mix of financials and fundamentals offering 30 to look out for.
If it works and proves to match the projected expected returns within the risks shown. Then this could be used to borrow money on margin call knowing your returns are modeled/controlled for and you can make money on the spread, but it’s risky. Borrowed money is usually at the Risk Free Rate, so you aim for a risk premium return by controlling for risk.
The philosophy behind the filters is, “this vs that. Bifurcation.” Split everything somewhat subjectively to a simple filter no matter how complex the calculation is on the back end, aka a 1 or 0 is coded for every value with default being 0 (such as na’s), and add these filters together across ETF’s and sift the top results. Which allows me to focus on revising and expanding individual logic in factors encapsulated in sql and/or python files. For example modifying thresholds which affect proportion of occurrence for a given factor(field). If query logic is based on median’s, it’s easy to get 50% of the values every time for each factor.