Does the Magic Formula work on the ASX?
Introduction
I recently read Joel Greenblatt’s The Little Book that Still Beats the Market and was intrigued by the Magic Formula. It seemed far too simple to generate the returns it claimed yet there seemed to be an abundance of studies backing its success. After delving into some of the blogs/articles/papers on the formula, I was surprised by the lack of research on the formula's performance within Australia, especially given the maturity of the ASX. So, I decided to give it a fair crack and share the results.
The Magic Formula
Simply put, Greenblatt's formula recommends the following: Screen for companies with a market cap over $50 million and rank each company's earnings yield and return on capital. Combine the ranks to form a "cheapness to value” score (rank 1 = company with lowest yield rank + ROC rank). Pick the top 30 companies with the lowest score and hold them for a year. At the end of each year, sell any depreciated stocks before the 12-month mark to harvest tax losses and sell appreciated stocks after 12 months to minimize capital gains tax. After, rerun the screen, pick another 30 companies and repeat annually.
Backtesting sample
I used a screening tool to run the Magic Formula on all ASX-listed companies excluding any with market cap under $50 million AUD. Following Greenblatt’s instructions, I substituted (ROA) for Return on Capital and the P/E ratio for Earnings Yield. I then ranked the companies and added the top 30 to a portfolio.
This screening process was run iteratively from August 31st, 2025, back to January 31st, 2002. Unfortunately, data before 2002 was wonky so that is all the history I got to work with. I also wanted to see if the rebalancing date mattered so I ran the screen quarterly, saving the top 30 value stocks for January, April, July, and October of each year.
To avoid look ahead bias, all ROA and P/E values were lagged by 45 days. This is because an investor on, for example, June 30th would not have had access to the financial results from that same date until the annual report was released. In this case, on August 15th. Therefore, they would have had to rely on past data in their screen.
Top 30 companies for July 31 2025
Performance setup
The universe was then divided into four portfolios based on their rebalancing month: January, April, July, and October. Each portfolio rebalanced on the same day of its designated month each year, for example, the January portfolio rebalanced every year on January 31st. This was done to determine if the rebalancing date significantly impacted returns. All portfolios were equally weighted.
Daily returns for each portfolio were then generated for the entire period. All returns were calculated using end-of-day closing prices and included dividends. I then saved daily returns and compared them against the relevant benchmarks.
Data transformation
The initial results were promising. Each monthly portfolio's performance was generally in line with one another and with the broader market. However, the July portfolio (yellow) stood out significantly, outperforming all others by a considerable margin. Most notably, there was a large spike in 2015.
Growth of $10,000 chart
A deeper look into the July portfolio's large spike reveals it was entirely due to Nido Petroleum Limited (NDO-ASX). The price surged from $0.02 to $0.82 on June 9, 2015, resulting in a massive 3,685.71% return. This one event contributed to a 123% one day return for the entire portfolio and is the sole reason for its outperformance.
Unfortunately, Nido has since delisted so I can’t verify if the return is accurate. To control for this outlier, I zeroed out the one day return and reran the analysis. Surprisingly the July portfolio's performance completely reversed and now significantly underperforms all other months. This really shows how a one day return could make or break an entire strategy.
To standardize returns for analysis, I'll use the average return of all four portfolios. For the July portfolio, I'll use the returns with the outlier zeroed out.
July portfolio (yellow) vs outlier-removed July portfolio (red)
Results & Analysis
After charting the returns, the results look quite positive. I used the ASX All Ordinaries as the main benchmark for analysis since it holds the largest number of companies. I also included the ASX 300 and ASX 200 as they are more commonly used in the industry. To no ones surprise, the magic formula outperformed all three benchmarks.
Growth of 10,000
Breaking down the returns into annualized figures, we can see that the magic formula portfolio's best year was 2009 while its worst was 2008. Interestingly, the last 13 years have been quite rocky as the magic formula underperformed the ASX except for 2017, 2020, and 2021. If it were not for the outperformance in 2020 and 2021, there is a good chance the magic formula would have underperformed the index.
Annual returns
I also calculated a few ex-post risk statistics. I used monthly returns for this bit and the 90-day Aus Bond Bankbill as the risk-free rate. The magic formula had a higher return and standard deviation compared to the market.
However, what surprised me was the beta, which was lower than the market's. I would have thought the beta would have been higher, as most of these companies would have price movements that are much more volatile compared to the ASX.
Another thing to note is that the Sharpe ratio is higher than the All Ords but lower than the ASX 200/300. This seems relatively in line, as our portfolio is relatively diversified (in the loosest way possible), so it makes sense that our Sharpe ratio is about the same as the market's.
Ex-post risk statistics
Risks & Caveats
Since this is my first backtest I wanted to mention a few caveats to the sample:
- Transaction costs and taxes are ignored.
- This is a significant caveat as the software I used doesn't account for taxes and transaction costs. Since the magic formula requires an almost 100% turnover every year, this would add up to a significant expense and would most definitely reduce our alpha. In spite of this, I still believe the Magic Formula would still have outperformed the market for the horizon I tested
- The screen fails to exclude delisted companies between fiscal year end and rebalance date.
- I noticed the screening tool returning delisted companies on rebalance dates. For example, when I run a screen on January 31st to find the top 30 Magic Formula companies for the January portfolio, the screen uses fundamental data from the most recent report date, being June 30th of the previous year. Due to the mismatch in when fundamentals are reported and when I run the screen, it's possible for a company to pass the screen based while being delisted from the ASX before the January 31st rebalance date. I've included a timeline below to clarify this point.
Fortunately, this only seems to affect a tiny portion of the universe so the impact should not be too significant. However, it does mean that the Magic Formula portfolio sometimes had fewer than 30 stocks as the analysis software would exclude the delisted company. In these situations, the analysis platform would rebalance the weights so that the remaining companies were equally weighted
Summary & Conclusion
Does the Magic Formula work in Australia? The answer seems to be abundantly clear. After backtesting the formula over a 23-year horizon across four rebalance dates, the magic formula does outperform the ASX. In fact, the formula produced a CAGR of 11.9% while the market produced around 8.5% (which is still not bad). That being said, does this mean the magic formula will continue to outperform in the near future? Given its recent rocky performance, I'm not so sure.
CAGR table
What do I think? The strategy makes intuitive sense and likely works over a long enough horizon (at least 20+ years). However, the reality is quite different, as, like all active investing, it requires consistency, patience, and a strong will to push through.
From what I've seen with people attempting to replicate the Magic Formula, most experience a string of bad years and quit early. Others might be more persistent but aren't strict about adhering to the rules of the formula (e.g., holding a lot fewer than 30 companies or inconsistent rebalancing). While I don't think people need to follow Greenblatt's rules to a tee, any deviation opens them up to other risks that might detract from the strategy's performance.
Perhaps the Magic Formula is best left to a robo advisor, since its implementation is extremely formulaic and requires little interpretation.
Sources/ References
- https://www.fool.com.au/2017/10/23/is-this-the-magic-formula-to-investing-in-asx-shares/
- https://www.oldschoolvalue.com/investing-strategy/the-magic-formula-investing/
- https://reasonabledeviations.com/2020/06/08/greenblatt-magic-formula/
Comments
Post a Comment