Here's a statement of the obvious: The opinions expressed here are those of the participants, not those of the Mutual Fund Observer. We cannot vouch for the accuracy or appropriateness of any of it, though we do encourage civility and good humor.
Is there a way to back-test a funds positions? Let's say I wanted to look at the top 5 positions of a fund and track that against the S&P 500. Any sites that do that? Thanks!
If your question is how to input data into any backtesting tool, I've worked through an example, using FCNTX. According to M*, its top five positions are: AMZN (7.05%), FB (7.01%), BRK.A (5.28%), MSFT (4.62%), and CRM(3.56%). That totals 27.52%.
To construct a portfolio of these five securities, you would divide their percentages by 27.52%, thusly:
AMZN (25.62%), FB (25.47), BRK.A (19.19%), MSFT (16.79%), and CRM (12.94%). Due to rounding, that comes to 100.01%. Round one of these down. BRK.A was rounded up the most (19.1860% was rounded up to 19.19%), so round this one down to 19.18%.
Depending on the backtesting tool you use, you might need to represent Berkshire Hathaway A as BRK.A or BRK-A.
Some tools (e.g. Portfolio Asset Simulator, http://capitalpas.com/Tools/BacktestPortfolioAssetAllocation) offer you a choice of benchmarks (such as the S&P 500 total return) to compare against. Others require you to pick an actual fund to compare with. For S&P 500 I might use VFIAX, or for longer periods, VFINX.
FB went public in May 2012, so one can't backtest this portfolio of five stocks further back than that. Backtesting from end of 2012 to end of 2018 shows that these five stocks did much better than FCNTX, which in turn beat the S&P 500 by a tad. Try running these tools and you can see this yourself.
msf, that's helpful but I was hoping for something that would look at it in real time, for lack of a better term.
For instance, it would look at the top 5 for 2014. Then if the position went out of the top 5 it would assume it's sold and the new stock would be added for performance tracking.
Just do the same thing I described, but point-to-point. Identify each segment of time where the top five don't change. For each segment, compute the total return using Portfolio Visualizer (PV), setting the start and end dates ot the segment. Then combine the total returns by multiplying, as illustrated below.
PV gives you the growth for each segment. It shows what the final value would be starting with $10K. For example, if it shows that you start with $10K and wind up with $12K at the end of a segment, you've got 120% of your starting value (and your total growth is 20%).
Suppose your top five stocks remain the same from Jan 2014 to June 2015, and then you've got a different top five from July 2015 to the present. If PV says that $10K grew to $12K in the first period, and that $10K grew to $180K in the second period, you combine them as:
$12K/$10K x $180K/$10K = 1.2 x 1.8 = 2.16. That is, $10K at the beginning grew to $21.6K at the end; equivalently, your total return was 216% - 100$ = 116%.
---
I am beginning to wonder what you're trying to do here. Even if the data you want existed (it doesn't because funds don't publish their holdings daily; they're only required to do so quarterly), it would be highly unstable. One stock might fall out of the top five merely because it grew slightly more slowly than the number six stock. Index funds have "buffers" to mute this sort of thrashing; here you're amplifying the effect.
I produced a list of top five equities for my example fund, FCNTX on half-yearly basis using the fund's (semi) annual statements. You can see the problem I described as well as another one - multiple classes of a company.
Alphabet (formerly Google) split into two share classes in early 2014. So what was a top holding of FCNTX was halved. In some periods, both share classes remained in the top five, pushing out another top five stock for no other reason than you were effectively including only the top four stocks (since Google took up two slots). In some other periods, it looked like the exposure to Google got cut in half, for no other reason that one of the two share classes missed the top five cut. Sometimes it looked like Google was "sold off" completely (e.g. June 2015), even though the fund held a lot of Google stock.
This thrashing didn't represent a change in exposure to Google. Such thrashing happened simply as a result of what you asked for.
Notice also that the fraction of the fund that the top five stocks represents ranges from 18% to 26.5%. Viewed this way, "top five" seems rather arbitrary. I might look at the top 20% of holdings, or come up with a more elegant metric for "high conviction" holdings. I'm not sure what "top five" signifies from an analytic perspective.
I'm basically trying to see how much performance is attributable to the top 5 holdings. I read a while ago that FPA ran an analysis and said that their top 5 holdings beat the index by a large amount. I'm mainly curious what FLPSX would look like with the top 5. It's not been a great performer over the past 10 years but part of that is due to its international exposure. I wonder if the fund would have outperformed the S&P 500 if one looks only at the top US holdings.
It looks like FCNTX is up about 80% over the past 5 years so it seems like the larger holding have done quite well.
@Paul, coming late and unread to this, but I can see the top 5 for FLPSX easily at M* Top Holdings, of course, and then could plot them the last six months or whatever period and see how they did compared with anything else. Including the fund overall, of which they were 17%. However, this is as of the end of April.
So even if I thought JT had bought them some time ago (I could go delve his reports) and I could determine roughly when, the outperformance, or whatever, would be only approximate, right?, dependent on timespan and exact periods. Also not up to date.
So I am not sure it is possible to do what you wish, and I wonder what FPA was talking about specifically.
Hi @Paul. What value would this top 5 fund information give you? Would it help you determine the managers value more-so than the funds overall return. Seems like a lot of work for... what purpose.
Yeah, I'm looking to run this on different time periods to play around with it. I'm kind of curious if FPA is right or if it's just an anomaly. I'd also like to know how much a manager's top 5 (or whatever number) drives overall performance.
Comments
If your question is how to input data into any backtesting tool, I've worked through an example, using FCNTX. According to M*, its top five positions are: AMZN (7.05%), FB (7.01%), BRK.A (5.28%), MSFT (4.62%), and CRM(3.56%). That totals 27.52%.
To construct a portfolio of these five securities, you would divide their percentages by 27.52%, thusly:
AMZN (25.62%), FB (25.47), BRK.A (19.19%), MSFT (16.79%), and CRM (12.94%). Due to rounding, that comes to 100.01%. Round one of these down. BRK.A was rounded up the most (19.1860% was rounded up to 19.19%), so round this one down to 19.18%.
Depending on the backtesting tool you use, you might need to represent Berkshire Hathaway A as BRK.A or BRK-A.
Some tools (e.g. Portfolio Asset Simulator, http://capitalpas.com/Tools/BacktestPortfolioAssetAllocation) offer you a choice of benchmarks (such as the S&P 500 total return) to compare against. Others require you to pick an actual fund to compare with. For S&P 500 I might use VFIAX, or for longer periods, VFINX.
FB went public in May 2012, so one can't backtest this portfolio of five stocks further back than that. Backtesting from end of 2012 to end of 2018 shows that these five stocks did much better than FCNTX, which in turn beat the S&P 500 by a tad. Try running these tools and you can see this yourself.
For instance, it would look at the top 5 for 2014. Then if the position went out of the top 5 it would assume it's sold and the new stock would be added for performance tracking.
PV gives you the growth for each segment. It shows what the final value would be starting with $10K. For example, if it shows that you start with $10K and wind up with $12K at the end of a segment, you've got 120% of your starting value (and your total growth is 20%).
Suppose your top five stocks remain the same from Jan 2014 to June 2015, and then you've got a different top five from July 2015 to the present. If PV says that $10K grew to $12K in the first period, and that $10K grew to $180K in the second period, you combine them as:
$12K/$10K x $180K/$10K = 1.2 x 1.8 = 2.16. That is, $10K at the beginning grew to $21.6K at the end; equivalently, your total return was 216% - 100$ = 116%.
---
I am beginning to wonder what you're trying to do here. Even if the data you want existed (it doesn't because funds don't publish their holdings daily; they're only required to do so quarterly), it would be highly unstable. One stock might fall out of the top five merely because it grew slightly more slowly than the number six stock. Index funds have "buffers" to mute this sort of thrashing; here you're amplifying the effect.
I produced a list of top five equities for my example fund, FCNTX on half-yearly basis using the fund's (semi) annual statements. You can see the problem I described as well as another one - multiple classes of a company.
Alphabet (formerly Google) split into two share classes in early 2014. So what was a top holding of FCNTX was halved. In some periods, both share classes remained in the top five, pushing out another top five stock for no other reason than you were effectively including only the top four stocks (since Google took up two slots). In some other periods, it looked like the exposure to Google got cut in half, for no other reason that one of the two share classes missed the top five cut. Sometimes it looked like Google was "sold off" completely (e.g. June 2015), even though the fund held a lot of Google stock.
This thrashing didn't represent a change in exposure to Google. Such thrashing happened simply as a result of what you asked for.
Notice also that the fraction of the fund that the top five stocks represents ranges from 18% to 26.5%. Viewed this way, "top five" seems rather arbitrary. I might look at the top 20% of holdings, or come up with a more elegant metric for "high conviction" holdings. I'm not sure what "top five" signifies from an analytic perspective.
Here are the (semi) annual reports from the SEC site that I used to create the table above:
12/31/18 https://www.sec.gov/Archives/edgar/data/24238/000137949119000865/filing706.htm
6/30/18 https://www.sec.gov/Archives/edgar/data/24238/000137949118004082/filing706.htm
12/31/17 https://www.sec.gov/Archives/edgar/data/24238/000137949118000849/filing706.htm
6/30/17 https://www.sec.gov/Archives/edgar/data/24238/000137949117005591/filing706.htm
12/31/16 https://www.sec.gov/Archives/edgar/data/24238/000137949117000981/filing706.htm
6/30/16 https://www.sec.gov/Archives/edgar/data/24238/000137949116005364/filing706.htm
12/31/15 https://www.sec.gov/Archives/edgar/data/24238/000002423816000025/Main.htm
6/30/15 https://www.sec.gov/Archives/edgar/data/24238/000072257415000288/main.htm
12/31/14 https://www.sec.gov/Archives/edgar/data/24238/000070420715000083/conmain.htm
6/30/14 https://www.sec.gov/Archives/edgar/data/24238/000002423814000041/Main.htm
12/31/13 https://www.sec.gov/Archives/edgar/data/24238/000087846714000111/contrafund.htm
I'm basically trying to see how much performance is attributable to the top 5 holdings. I read a while ago that FPA ran an analysis and said that their top 5 holdings beat the index by a large amount. I'm mainly curious what FLPSX would look like with the top 5. It's not been a great performer over the past 10 years but part of that is due to its international exposure. I wonder if the fund would have outperformed the S&P 500 if one looks only at the top US holdings.
It looks like FCNTX is up about 80% over the past 5 years so it seems like the larger holding have done quite well.
So even if I thought JT had bought them some time ago (I could go delve his reports) and I could determine roughly when, the outperformance, or whatever, would be only approximate, right?, dependent on timespan and exact periods. Also not up to date.
So I am not sure it is possible to do what you wish, and I wonder what FPA was talking about specifically.
Yeah, I'm looking to run this on different time periods to play around with it. I'm kind of curious if FPA is right or if it's just an anomaly. I'd also like to know how much a manager's top 5 (or whatever number) drives overall performance.