# Using alpha to evaluate trading strategy

I have a trading strategy that generates returns $R_{t}$. I want to test the strategy by looking at the alpha:

$R_t - R_{f,t} = \alpha + \beta (R_{m,t} - R_{f,t}) + e_t$

I compare my alpha against the strategy of just doing an equally weighted portfolio over my assets universe. I also use an analogous method for factor models with more regressors.

Does it matter that my $\hat{\alpha}$ is always insignificant at the $10\%$ level? I've tested many strategies using this type of factor model and $\hat{\alpha}$ is never significant at $10\%$. However I usually find that my $\hat{\alpha}$ is much larger for the strategies than for an unweighted portfolio.

-
Whether is matters or not depends on what question are you interested in answering using this model. – Ryogi Oct 29 '12 at 18:58

## 1 Answer

I'm not sure what you mean exactly by "does it matter...", but generally speaking it should not surprise you that your alpha is not significant, as many trading strategies are more or less "transformations" of beta.

In the purest sense, alpha is not easy to accomplish, and various forms of the EMH would say that it is nearly impossible to achieve it for a sustainable time period, at least without some non-public information advantage.

Many trading strategies show positive alpha by a combination of tricks and mis-representations, such as: using an irrelevant benchmark, calculating alpha gross of fees/expenses/trading costs, etc., simply calculating it as

$\alpha=R_t-\beta R_i$ where $R_i$ is the index return.

I have seen so many bastardizations of what alpha is truly meant to be that you should not read anything into the results that you have "insignificant" alpha.

-