Testing the Ranking System Across Multiple Universes

One test for ranking system robustness is to test it on several different universes.  Today I tested the ranking system Olikea's Brainchild on the three main marketcap universes as well as the S&P500 and my custom baseline_universe.

In my last post I provided a brief description of the EXCEL spreadsheet I created for ranking system evaluation.   Today I would like to summarize the test results for several different universes.  As a sanity check and for performance reasons I did a comparison against another popular ranking system "Filip's Supervalue 92".

Here are the results:

SMS_Sharpe
Olikea's Brainchild
Universe	5 Stocks	10 Stocks	20 Stocks	40 Stocks	80 Stocks	160 Stks
Baseline	3.67	4.34	4.31	3.98	3.82	3.69
S&P500	2.52	2.22	2.41	2.07	1.95	1.99
LargeCap	1.50	1.36	1.14	1.18	1.10	1.42
MidCap	1.50	1.64	1.80	1.56	1.67	1.84
SmallCap	1.48	2.15	2.29	2.57	2.79	2.72

As I stated in the last post the ratios are calculated using the baseline universe as the Minimum Acceptable Return (MAR).  Normally Sharpe and Sortino Ratios are calculated using a Risk-Free Return (RFR) which would normally be T-BILL interest rate.  The SMS_Sharpe is quite solid for Olikea'S Brainchild 19Dec08.  Worst case is for the LargeCap universe where the figure drops to 1.10, still a relatively good number.

SMS_Sortino
Olikea's Brainchild
Universe	5 Stocks	10 Stocks	20 Stocks	40 Stocks	80 Stocks	160 Stks
Baseline	7.18	8.34	7.77	6.90	6.35	6.15
S&P500	4.26	3.72	4.11	3.41	3.17	3.14
LargeCap	2.50	2.28	1.89	1.89	1.76	2.27
MidCap	2.55	2.78	3.05	2.53	2.75	3.17
SmallCap	2.44	3.51	3.79	4.16	4.73	4.79

Likewise, the SMS_Sortino is quite good across the board, with a low of 1.76 for trading 80 top-ranked stocks within the LargeCap universe.  So I have to conclude that the ranking system Olikea's Brainchild 19Dec08 is solid for a wide range of universes.

I decided to graph the results and compare against Filip's Supervalue 92, another popular ranking system.

Click to enlarge

As can be seen from the graphs, Olikea's Brainchild performs better for most of the universes.  But I would probably choose Filip's Supervalue 92 if I was strictly designing a SmallCap system, simply because of the consistency for small and large number of stock holdings.

An interesting effect that I have observed from this exercise is that Olikea's Brainchild performs better with small number of stocks for LargeCap universe and S&P500.  Performance declines for larger number of stock holdings. On the other hand the ranking system performs better on SmallCaps with a large number of stock holdings.

For Filip's Supervalue 92 ranking system, performance for LargeCap and S&P500 is best holding a large number of stocks and relatively independent of number of stocks for the remainder of the universes.  It seems that this tool allows the user to select a ranking system based on the number of stocks that he/she wants to hold.

Steve

Sponsored by Hot Stocks