Post subsequently edited to correct the numbers because of a bug in the script to calculate percentages. Factor of 100 was missing.
Thanks to
@lljb for pointing me at Yahoo Finance site to get raw NAV data for mutual funds in the thread on Smead Value Fund. I quickly hacked up a script to compute some metrics listed below. I must stress that these are very preliminary results and must not be taken seriously until they have been further validated. But they are encouraging enough to suggest my intuition may have been right and that some new metrics might be able to better separate funds that deliver consistently from one hit wonders or streaky funds that do not deliver unless you are invested in some specific good times for the fund. All of them may have good standard calendar year performance.
In that previous thread I had expressed my intuition that calendar returns of mutual funds and computed values on those might mislead investors on what they may experience in practice where they don't care when they start to invest in a fund. The performance they experience may be disappointing even for highly rated funds either because they weren't in the fund when it realized most of its gains or the fund had a streaky performance characteristic which if you didn't catch would disappoint.
As a way to study the ability of a fund to deliver regardless of when an investor invested in it, I suggested computing returns from all possible intervals in the fund's history and aggregating that data.
In this calculation, I am not looking at the total returns but rather the difference in return from the index (all large cap core compared to S&P 500) in such periods, or the alpha the managers generate over the index (equivalent of the +/- S&P 500 TR USD row in M* performance tables)
I have done the first step to calculate the expected alpha return from a fund if you had invested in a randomly picked date rather than on a calendar date. The expected return is not the return you would have actually got but is the average over all possible dates of investment for a given length of holding (1yr, 3yr, 5yr). This is just one number and the next step is to find the distribution of returns to answer questions like "what is the probability that if I pick a fund to invest in a particular fund on any date, that the fund would provide a positive gain over the index". All back testing caveats apply just like all other metrics on this site and M*.
As a bench mark I used the mutual fund VFINX and the ETF SPY.
First I did the rolling computation between them. This is to establish the extent of the tracking variance for a fund that is just trying to mimic the index. This can be used as a first approximation to establish the "margin of error" for the computation on active funds. In other words any difference less than the tracking variance is meaningless and so same as index.
The reference tracking variance for the index funds were:
1yr 3yr 5yr
0.03% 0.15% 0.33%
Fairly small numbers.
Here are the alpha expected returns for each period for SMVLX that started this whole thing along with a number of well rated popular funds arbitrarily selected. All of them for the period from 2008 to current to compare with the youngest.
1yr 3yr 5yr
SMVLX 2.95% 15.38% 31.99%
PRBLX 0.70% 0.91% 1.36%
GLDLX -0.27% -0.03% -1.43%
PRDGX -0.38% -2.18% -4.82%
POSKX 1.23% 1.96% 4.11%
YAFFX 3.51% 5.34% 12.15%
Again these are deltas from index returns and they are average over all possible investment dates in that period. So this provides only half the story.
Observations:
1. For some the alphas are negative (not expected to beat index) or within margin of error (they are index huggers). But this is not to make passive vs active statements, just to see if there is a better probability of realizing alphas with some funds than others regardless of when you invested in it.
2. SMVLX is an overachiever in this lot followed by YAFFX. YAFFX had relative outsized returns in 2008 and 2009 to index which helped it similar to the good streak SMVLX had mid 2012 to mid 2013. POSKX is the only other one ahowing respectable numbers in that short list. This is why the next step is necessary, looking at the alpha distribution to see how likely you may see positive alphas for these funds so that a short streaky period does not unduly influence the above numbers. In other words how likely you will see positive alpha even if you missed those streaks. Funds that generate steady alpha (not necessarily consistent from year to year in beating the index) should do better than the streaky ones or one-hit wonders.
I will post additional results as I get to them. Suggestions on what other funds to include,welcome. It is still a manual process to grab the data from Yahoo so may be slow. Will make the script scrape automatically from Yahoo. Only considering Large Core equity funds that can be benchmarked with S&P 500.
Please take these results in the spirit with which they are provided. An experiment to satisfy curiosity and a hunch. Constructive suggestions or comments welcome.