Tranche Investing: How to Reduce Transactions
Based on results from extensive back-testing, there is a degree of luck as to when a portfolio comes up for review and when securities are bought and sold. Did the review come when it was time to buy and the market dipped? Or were ETFs sold when the market was low. Was the trade delayed or missed due to setting a limit order when the market moved in the opposite direction of the order?
Tranche investing is a model designed to mitigate this “luck-of-the-transaction” problem by spreading out the transactions over what are called Portfolio Offsets or Offset Portfolios. The Tranche 1.6 spreadsheet is constructed to help users implement the Tranche Investing Model by allowing as many as 12 different offsets. We can also vary the number of trading days per offset. The default setting is six (6) trading days. A view of this worksheet is shown in the following video.
One of the negatives of the tranche model is the increase in the number to transactions per month. Click on this link to find the “Camtasia” where I put forth an idea that will reduce the number of transactions per month. The overall concept is to keep investors out of low performing ETFs and at the same time invest available cash in the better performing funds. The reduction in transactions comes by way of not paying attention to every buy and sell recommendation that emanates from the Tranche 1.6 spreadsheet.
This idea requires judgment rather than following a mechanical model. As a result, it is nearly impossible to back-test. Discussion is most welcome.
Lowell,
I think your suggestion shows a lot of investor common sense. Several comments:
– As you know, I am currently running a very large back test of the 12 X 6 set of adjustable tranche parameters in the tranche workbook. This test is going to shed some light on how to select these parameters, hopefully.
– I ran a back test in my original tranche article where I used a temporal weighting suggested by Herb Haynes where more recent tranche results are given a higher weighting and more distant (in time) results are given lower weight values. The results of this back test were positive, suggesting that temporal weighting of this type could be a useful adjunct to the tranche scheme by providing higher returns with a slight degradation of the tranche effect of reducing timing luck. I will include this twist in my current back test study.
– I also ran a back test on another twist of the tranche idea. I set a limit on the number of tranche assets I would allow in the final selection. For example, suppose I said I only wanted a maximum of 3 assets in the final tranche selection. If I run the tranche algorithm and I get 5 assets, I pick the 3 with the highest tranche weightings. I then re-normalize the weights of those 3 so that the sum of their weights is one. That reduced the number of trades at a slight reduction of the tranche effect of reducing the timing luck. I’ll also re-run this twist.
– Finally, one could wonder why you choose to use all 12 period if you are only going to pay attention to the last N of them. Why not just limit the number of tranche sub-periods to N to start with?
More to come.
Ernie
Ernie,
By using 12 offsets, it better illustrates my idea of not paying attention to every recommendation. As a default, I set the offsets to four (4). You will see that in the screenshot when I update the Einstein tomorrow.
Lowell