I have had a chance to play with the SH (Sharelock Holmes) database, which can give quarterly performance reviews on portfolios going back to 2002. That assumes there is sufficient data for the query in question. Note that I am totally dependent on the answers given by SH. I have no way of verifying the results myself.
SH scores work the opposite of Stockopedia. The lower the better.
What emerged is perhaps not surprising. Good scores gave better results, and bad score gave worse results.
It was important to avoid the worst decile. The worst decile for quality returned -9.5%; all figures that I state are average annual performance relative to the FTSE100. The worst decile for momentum returned -5.5%. For value it was -13.9%.
For Q<=50, the return was 5.6%, for V<=50 it was 5.6%, and for M<=50 it was 5.9%.
For thresholds of 10, Q’s return was 9.9%, V was 7.1%, and for M is was 9.5%.
Setting the threshold at 5, we obtain Q return 10.0%, V 8.0%, M10.8%.
For a threshold of 2, Q was 11.2%, V 7.9%, M 10.7%.
So returns increased as you lowered the threshold. The returns on value diminished rapidly, though, and were not as good as that for quality and momentum.
In other words, as single factors, Quality beat Momentum, beat Value.
The best results were from the Consistent Growth screen, however, which did not depend on any of the scores above. The screen is based on the criteria:
- years of eps growth > 5
- years of turnover growth > 5
- years of dividend growth > 5
- MScore < -1.78
- Piotroski score > 7
Outperformance was pretty impressive: 19.9%pa. The Slater Growth and Zulu screen also did exceptionally well. The Cornerstone Growth and Dremen screens were also good.