*NEW* LONG TERM PIOTROSKI BACKTEST ON THE JSE RELEASED!PowerStocks Analytics is extremely proud to present the first ever public Strategy Index Back-test Performance (SIBP) report. We have chosen a relatively complex strategy as our first history-making exercise, namely the Piotroski strategy, because our model portfolio is performing so well, and the market timing for a value based strategy such as this is perfect.
We analysed over 2,500 financial reports from the 14 year period since 1994 to enable us to create this SIBP. It has taken about 3 months of serious after hours work (we have day jobs) and analysis, cleaning and checking but we are very happy with the results, and this first exercise helped us establish our SIBP methodology.
The Piotroski SIBP will examine risk adjusted performance of Piotroski-scored shares on the JSE since 1994, against 1, 2, 3 and 5 year rolling periods. We will examine the effectiveness of Piotroski on the whole JSE universe, stocks with Price/Book less than 1 and stocks with Price/Book in the lowest 20% (this final universe is the one preferred by Joseph Piotroski)
We created 10 sets of data, one for each year starting in 1994. For each set of data, we computed the Piotroski "F" score of each share listed on the JSE at that time, from the annual financial reports (Finals) published for that company in that year. We then created portfolios grouped by Piotroski "F" scores (10 portfolios in total) and calculated each shares' subsequent 1,2,3 and 5 year appreciation in share prices from date of the publication of the results.
We assumed equal weighted portfolios to derive an assumed growth for each portfolio based on individual constituent share growth. We then averaged the figures from the 10 sets of portfolio data to arrive at a long term trend. We also examine standard deviation of the data to compute Sharpe Ratios to arrive at risk adjusted portfolio performance. We will also inspect base rates (number of times a portfolio beats the performance of the entire group of shares for that year). and total return (to measure which portfolios made the most out of R1.00 since 1994 to 2008.)
For the 5-year growth rates we use the 10 periods 1994-1999, 1995-2000,....,2003-2008. For the 1 year growth rates we use 14 periods of 1994-1995, 1995-1996,...,2007-2008. One year portfolios are rebalanced every year, five year portfolios are only rebalanced every fifth anniversary.
TESTS ON THE WHOLE JSE UNIVERSE
We see from the below graphic that Piotroski scores are closely correlated to overall growth performance regardless of Price/Book value and holding period. The below graphic represents the average growth of various sets of portfolios for 10 measurement periods. The name of each portfolio is indicative of the shares that it consists of. For example "9+8" consists of shares with Piotroski "F" scores of 9 and 8 respectively. "F" scores are from 0 (very weak) to 9 (strong).
We see that on average, portfolios of shares with Piotroski scores of 9 that are bought and held for 5 years can be expected to grow 208%, or 25.3% CAGR. Portfolios bought and held for 3 years can be expected to grow 111% or 28.3% CAGR. Portfolios rebalanced annually deliver 40.7% per annum. In all the tests we have conducted, annual rebalancing delivers the most powerful results, so lets inspect this strategy closer in the figure below.
Here we compare portfolio volatility (standard deviation) and risk-adjusted return (the Sharpe ratio, a simplified version of which we use which is just average portfolio growth divided by its standard deviation). We see that portfolios with "F" scores of 9, 8 plus 9, 8 and 7 have high Sharpe ratios (the best risk adjusted return), whereas portfolios with "F" < 7 demonstrate a standard deviation greater than the average appreciation (an uncomfortable place to be for any investor.)
TESTS ON THE PRICE/BOOK < 1 UNIVERSE
Piotroski's 20 years of research on the US markets shows that his financial strength rating system works best on "distressed stocks" with low price/book ratios, making it an ideal companion to any value investor seeking to minimise risk. We duplicated the above tests, but this time only built portfolios from shares that were trading at less than book value (a classic value investor metric). The results are shown below:
We see that growths achieved with high scoring Piotroski shares are significantly more when chosen from the low price/book universe, confirming this is the universe we should focus on when building Piotroski Indexing Strategies. We see that on average, over the last 14 years, portfolios of shares with price/book ratios less than 1 with Piotroski scores of 9, that are bought and held for 5 years can be expected to grow 267%, or 29.7% CAGR (vs previous sections' 25.3%) Portfolios bought and held for 3 years can be expected to grow 149% or 35.6% CAGR (vs 28.3%). Portfolios rebalanced annually deliver on average a stunning 61.5% per annum, almost double that of the previous section. But average growth is not the full story, we need to examine volatility and risk adjusted return:
We see that the higher returns of the price/book<1 universe is accompanied by higher volatility (risk) as evident in the lower Sharpe ratios. Whilst the "9" portfolios delivered the highest average annual growth of 61.5% over the 14 year period, it only has the 3rd highest risk adjusted return (0.93), with 1st place going to "7" (1.14) and 2nd place to "4" (1.04). We must however point out that shares with "F" scores of 9 are rare and the "7" and "4" portfolios had the benefits of holdings triple the size of "9" (average "9" portfolio size was 3 shares) which aided them in reducing their standard deviations (and boosting their Sharpe ratios). We will show however that "9" is a far superior index strategy based on base-rates and total return.
TOTAL RETURN TESTS
We now evaluate total return delivered by a Price/Book < 1 Piotroski strategy (portfolio). We start in 1994 and for each Piotroski portfolio, we take a hypothetical R1.00 and use it to buy equal rand amounts among the shares of the portfolio. We hold the portfolio for a year, after which we rebalance it (sell everything) and repeat the process. We exclude broker fees etc. which will be minimal due to only rebalancing annually.
Click the image above for an Excel table showing you how the R1.00 grew each year for each portfolio, together with each portfolios annual growth rates. The results are staggering. After 14 consecutive 12-month investments in portfolios of shares with Piotroski "full marks" scores of 9, our R1.00 is now worth R238.00. This equates to a growth of 23,700% over 14 years, or 47.8% CAGR. Note that although "5" (R99.00) and "6" (R74.00) delivered the next biggest returns, their Sharpe ratios are much less than "9" meaning their returns were isolated to a few "big hits" possibly preceded by protracted down periods.
The "9" portfolios' total returns started consistently outstripping the others from an early stage, as evidenced in the below graphic.
Note that since 2007 the Piotroski "9"ers have "stayed out" the market and hung onto cash due to lack of shares meeting the strict financial screening criteria. This is one of the reasons Piotroski is such a solid performer, as in the 1998/99 dotcom meltdown it also "opted out" due to shares' financial stresses during the market crash. In fact a lack of "9"ers thrown up during Piotroski screening serves as a useful "coal miners' canary" early warning for the market.
BASE RATE TESTS
Here we examine another important characteristic of a strategy, namely its historical tendency to outperform the market as a whole. This is very important as very few investors can stand by and be patient with a strategy whilst it is under-performing the market. For our base rates, we created two portfolios called JSE (or GRP in above tests) and P2BK which represented equal-capped indexes of the entire JSE universe and the price/book < 1 universe respectively. The below table shows that the Piotroski 9'ers beat the JSE peer group index 71% of the 14 year test period (10 years in 14).
Note the poor base rates of 40% for "6" and 35% for "5". Although these came the closest to "9" in the total return tests, these poor base rates meant the hapless investor had to spend 60%-65% of the time under-performing the market, sometimes up to 4 years in a row! We don't think ANY private investor has the psychological wherewithal to stick through with such a strategy without bailing out prematurely. One important point - the "9" and "7" portfolios had only one single negative result in the 14 years under review, a 93% positive rate.
PORTFOLIO SIZE & GROWTH CORRELATION
Piotroski "9"ers among the Price/Book < 1 universe are very rare. The minute a low price to book company announces results and you see that its a "9"er you are urged to acquire for your portfolio immediately. Our research and experience running the Piotroski model portfolio shows that the market cottons on very quickly (within a few weeks) and the share shoots up. You can run very small portfolios with Piotroski "9ers" as the figure below shows for the 14 year backtest we did. How a strategy that picks such small portfolio sizes manages to consistently outperform with such low volatility is an attestation to the use of Piotroski to acid test cheap shares for financial stability.
We can conclusively say that Piotroski is a valid and strong confirmation for use on the JSE. The best Piotroski strategy is one that is annually rebalanced or stocked as occasion permits (normally once a passing companies' results have been published) with issues scoring full-house marks of 9. In the long run, as our extensive backtests have shown, this strategy provides superior total returns coupled with excellent historical base rates of 71% and solid growth consistency. The mildly below parity Sharpe ratio of 0.93 is easily stomached by the investor. Whilst the annual standard deviation of 66% might raise eyebrows, this is more than compensated for by the average annual growth rate of 61.5% and the superior base rates and 93% positive periods.
Portfolio administration is low, since very few shares qualify for full marks, except in the midst of an economic bull market, and even then portfolios rarely exceed 8 shares. Piotroski has a nice fail-safe in that in tough economic periods, very few companies pass its strict financial screens and it has a tendancy to "sit out" bear markets, leaving you with the cash to pile in again at the bottom.
We finally note that portfolio "7" had very attractive characteristics throughout the test period, namely the highest Sharpe, base rates on par with "9" and a not-to-be-sniffed at total return over the 14 years of 4,500% or 31.6% CAGR. Our next research paper will look at some variations of Piotroski that will raise Sharpe and base rates and improve returns (such as combining "9"ers and "7"s into a single portfolio).
Below is a table listing the 52 transactions (26 buys and 26 sells) made over the last 14 years to manage the "Niners" portfolio. That's an average of 3.71 transactions per year, hardly an effort for multiplying your money 238 times! The bulk of the 2008 financial results are in as of 18 March, so now is a good time to start your "Niners" portfolio. Go to the Piotroski blog to see a published list of qualifying shares (some "Niners" have made an appearance again!)