I've recently tested quantopian.com platform for easy backtesting. I essentially plugged in the valuations I posted previously. Using quantopian forced me to come up with an algorithmic approach to sizing positions.
I ended up using a simple and natural approach. For each security I have a target price. Therefore I can calculate a percentage upside as (target_price - current_price) / current_price. Then I summarize all upsides into a total_upside and allocate to each security upside / total_upside percent of funds. This keeps me 100% in the market as long as any security has any upside (they usually did, but it varied greatly from single-digit percent to 4x).
I think this approach strikes a good balance between maximizing my upside, while keeping some diversification (if my valuations are correct, of course). Strictly speaking to maximize upside I should put 100% in the stock that has the highest upside at given point in time, but that will result in no diversification and no room for valuation error.
The full results are here. I would not pay too much attention to numbers, as the stated goal of mine (maximize returns) and risk metric (permanent capital loss due to misunderstanding of the underlying business) is quite different from what alpha/beta or volatility measure. I also used a very simple approach, with only 3-6 securities in the portfolio, annual updates to intrinsic values, etc. Having Google (which performed very well) in the portfolio skews dataset in two ways - it's of course a plus, but early on as Google price increases, the allocation decreases significantly and a good chunk of the gains is 'missed' that way.
The biggest benefit for me from this exercise is the new approach to portfolio sizing. I don't plan to use quantopian, but I will look into using Interactive Brokers API to automatically place my orders according to my intrinsic valuations and in line with portfolio sizing.
I ended up using a simple and natural approach. For each security I have a target price. Therefore I can calculate a percentage upside as (target_price - current_price) / current_price. Then I summarize all upsides into a total_upside and allocate to each security upside / total_upside percent of funds. This keeps me 100% in the market as long as any security has any upside (they usually did, but it varied greatly from single-digit percent to 4x).
I think this approach strikes a good balance between maximizing my upside, while keeping some diversification (if my valuations are correct, of course). Strictly speaking to maximize upside I should put 100% in the stock that has the highest upside at given point in time, but that will result in no diversification and no room for valuation error.
The full results are here. I would not pay too much attention to numbers, as the stated goal of mine (maximize returns) and risk metric (permanent capital loss due to misunderstanding of the underlying business) is quite different from what alpha/beta or volatility measure. I also used a very simple approach, with only 3-6 securities in the portfolio, annual updates to intrinsic values, etc. Having Google (which performed very well) in the portfolio skews dataset in two ways - it's of course a plus, but early on as Google price increases, the allocation decreases significantly and a good chunk of the gains is 'missed' that way.
The biggest benefit for me from this exercise is the new approach to portfolio sizing. I don't plan to use quantopian, but I will look into using Interactive Brokers API to automatically place my orders according to my intrinsic valuations and in line with portfolio sizing.
No comments:
Post a Comment