Similarities and Differences

In the last post, we compared the ratings produced by my system and those produced by the ICC's , and found that, although there’s a current disagreement, the two systems have been largely in concordance when it comes to the identification of the best team in the world.  Moreover, many of the differences could probably be eliminated by reworking my system to predict the results of whole series (which is the unit on which the ICC ratings are calculated) and then update, instead of doing so after each match.  Which begs two questions: is there any point to my system, if it’s only mirroring the ICC system?  And why should the two systems produce similar results?

To address the second question first. There are, as I mentioned in my last post, certain Elo-like concepts in the ICC system – it’s different, but not completely alien.  But perhaps this is a secondary issue. At times in cricket history, one team has been clearly dominant.  Almost every cricket lover, asked to make a call on the world’s best team in the mid to late 1980s, would have mentioned the West Indians; twenty years later, the Australians would have gained unquestioned support.  And in those eras, almost any sane system for calculating the best team objectively from past results, even a very naïve one (for example, one simply counting the ratio of recent wins to losses), would have given the same result.  It’s only when the pattern isn’t clear cut that a different system will produce different outputs.  

To address the first question, one reason for performing my own calculations is to give me easy access to a great wealth of historical data, for which ICC data is either non-existent (for example, there is no ICC data before 1952) or difficult to access.  But more than that, the ICC system is really very complex, and depends on a host of arbitrary choices that are almost impossible to justify from theory, even if the overall output seems sane.  Although the system seems empirically to do a reasonably good job of identifying the world’s best team, it’s not immediately obvious why it should do so.  It feels like a hack compared with the beauty of a straightforward Elo-derived system like my own.  And with this preamble out the way, it’s now time to delve into the system’s rather horrid guts.