To paraphrase the old adage; "a monkey throwing darts will outperform most fund managers". I have seen this concept explored several times in relation to the SP500, but I was interested to see if it had any relevance to the ASX200.
Our monkey with darts will be a random number generator, selecting 10 stocks to buy from the XJO in equal weight. We test with $100,000 of capital.
In order to test this thoroughly, we need to set our benchmark as well as a testing window. The benchmark is simply the STW - the SPDR S&P/ASX 200 Fund. We will explore the period from 2011 to 2016, which should be just long enough to see if there's any substance to the old saying.
- Sharpe: 0.06
- Max Drawdown: $23,555
- Max Drawdown %: 22.27%
The almost-0 sharpe implies that day-to-day returns are very close to random. We end the 5 years with about 10% total return on equity (my dataset does not include dividends).
Now that our benchmark is defined, let's determine how a random basket of 10 stocks from the XJO compares. Very vanilla, we put $10,000 into 10 stocks and hold them for the 5 year test.
- Sharpe: 0.43
- Max Drawdown: $12,744
- Max Drawdown %: 20.26%
Ignoring total returns, our sharpe ratio has improved, which we can visualise by a smoother equity curve than the benchmark test. The highly volatile areas around the end of 2011 and 2015 with the Euro and China scares are also visibly smoother. For the average investor, this is a much more favourable equity curve.
Our total returns are roughly 50% with this strategy too, largely outperforming our benchmark.
But of course, this was the first of a random test -- how do we know it's not just lucky? The following chart illustrates how the strategy would have performed 5 times. The large spike on run #4 is likely due to a messy stock consolidation within my dataset (which is unfortunately, at this stage, only free end-of-day data).
The STW Buy-and-hold is represented by #0 on the chart, clearly the lowest performer over all of the random tests.
While the returns overall are fairly decent, how would the shape of the equity curve be affected by rotating the portfolio every month? That is, on the first of each month we liquidate our current holdings and allocate 10% of our cash to 10 random positions again.
About 100% return on equity. A much better return than the 55% we achieved on the first try with our 10 buy-and-holds. How does this look over a few more tests?
Surprisingly good, with a pretty an average return eyeballed at around 60%. The equity curves do look smoother than when testing without rotating stocks. I suggest that this could be to do with us only being exposed to a particular stock for a month at a time; meaning we will never be "stuck hoping" for a position to turn around while it continues to slide south.
The ASX200 is an example of a heavily skewed market-capitalisation weighted index, with financials and materials representing nearly two thirds of the index. The index is disproportionately represented by the top few stocks; a bad day for CBA will normally mean a bad day for the entire market.
ASX200 Sector Breakdown as of 31 August 2015
Image credit to wise-owl
The implication here is that "beating the market" really only means "beating CBA".
I believe the key takeaway point, and what the results seem to be hinting at, is that a random asset allocation forces the investor to take on a more diversified risk profile. By limiting the available pool of stocks to still being within the ASX200, we limit the volatility often associated with small to mid cap stocks. The random nature of the selection also ensures that we remain completely without bias in deciding which sectors we want exposure to.
While the above results clearly indicate something promising, there are several flaws to consider in the tests performed so far -- which if not addressed in a future post, could invalidate these results.
Historical XJO Constituents
This backtest was only performed using stocks that are in the XJO currently. The obvious drawback of this is that stocks which are no longer in the XJO due to drastic falls in share price cannot be among those randomly selected (the perfect example of survivorship bias), and stocks that are in the XJO today, but weren't there previously, could have been selected.
Rotating the portfolio monthly will slightly alleviate the inaccuracy of selecting stocks that have grown into the XJO, however a historical constituents list would be needed for a perfect test. In future I aim to address this by using the market cap of a stock as a proxy for being in the XJO -- so we would filter out any stock with a market cap below, say, $1bn.
The market cap filter hasn't been included in the above tests as my database only includes price information for stocks -- no fundamental data such as the number of shares available, which is needed for market cap calculations.
Slippage & Commissions
The monthly rotation does not take into account standard commissions, which are about 0.1% for a retail trader using a Australian online broker. Slippage may impact results slightly, but haven't been modelled in the above tests.
As mentioned above, my database does not currently include any dividend information. The strong dividends typically paid by ASX blue-chips go a long way to explaining the lacklustre looking equity curve when compared to, say, the SP500 or DJIA since 2008.
Despite the above shortcomings in this initial test, the results deserve further investigation. I aim to address the above drawbacks once I have a more comprehensive dataset to backtest upon.