The best way to build a long/short portfolio

Question

For several mutual funds, I have their performance (in %) as well as their TNA (in bn) on a monthly basis (months 1 to 193).

Thanks to a previous step, I know for each fund, each month, whether I should take a long or a short position (or do nothing). What do you think is the best way to build my long/short portfolio? Should I have 2 portfolios (one long and one short)? How do I calculate its return?

Thank you very much for your help,

Vanie

EDIT QUESTION:

I'm evaluating a strategy ex-post. In a few words, I split funds into different clusters using a dynamic clustering method. In fact, I'm interested in disconnections. The principle: if a fund goes out of its reference cluster (I set 2 periods/month in a row) then I can take a position: long or short, depending on whether it outperforms or underperforms its reference cluster (I compare the fund's return with the average return of its cluster). This until he is back in his cluster. For example, for long positions, I hope that a found that underperforms its cluster, will "go up" to return to it.

From there, I create a portfolio with the returns (weighted with TNA) of the long and short positions.

I've done all these steps but I get results that are not very logical (extreme). Do you think there is a problem somewhere?

Answer 1

I have their performance (in %) as well as their TNA (in bn) on a monthly basis (months 1 to 193).

Thanks to a previous step, I know for each fund, each month, whether I should take a long or a short position (or do nothing).

If you are using information that was not available at the time of the back-test's transactions, the strategy is not reproducible going forward. No one can use next month's return to build their portfolio today. They do not know next month's return.

In practice, you do not know whether your should take a long or short position.

What do you think is the best way to build my long/short portfolio?

If you could see the future, you would buy whatever will have the highest percentage gain.

How do I calculate its return?

Return = (gain or loss) / capital.

Answer 2

Thanks to a previous step, I know for each fund, each month, whether I should take a long or a short position (or do nothing). What do you think is the best way to build my long/short portfolio?

You know this only for the past, not for the future.

If you do not want to expose yourself to potentially unlimited losses, I will argue that the "short" part of your portfolio should be zero.

Any nonzero short-selled position has potentially unlimited losses.

In particular, no stop-loss order is going to limit your losses in every single case. It is possible there is a discontinuous jump on the price of an asset, and thus, if the price skyrockets in a short amount of time, the stop-loss order will not act quickly enough, and you need to buy back the short-selled security at a very great price.

Based on this, the "long" part of your portfolio should be equivalent to your investable assets.

Answer 3

In my world, there are a few reasons for shorting.

(1) I think that the security is going to drop in price and therefore, I short it.

(2) It's highly correlated to another security, the spread between the two has widened and I am trying to capture some reversion to the mean. This long/short pairs strategy is far more effective in a volatile market (see 2008, 2009, when the China tariffs were implemented, the December 2018 rising rates inverted yield curve correction, and March of this year).

A subset of (2) is that I'm shorting a smaller position of a correlated issue as a hedge going into ex-div.

I'm not big on calculating statistical returns on complex individual positions. The objective is to make a profit from trading and the stats won't increase the yield on the next trade. Find good trades and do your best to limit their losses when wrong.