Independent events allows easier analysis. Consider baseball. There are very good metrics for baseball...Cricket is more complicated to analyze l because of the great variation in pitches and much smaller sample size. If teams played 25-30 tests a year we could come closer to an adequate sample size for analysis.
Dispute a couple of things there. Decent, methodical sampling of available data will give you a representative sample even if a team didn't play that many matches in a year. Also, you could just take the sample over a longer period. The same problem confronts serial violent crime/killer researchers where there are often years between murders/violent crimes.
Also, if the events in cricket were truly discrete events, that makes comparison far more difficult. As it is, like baseball, the events (e.g. a ball being bowled) aren't discrete so they are able to be compared but not without significant caveats. The caveats often do render comparisons and subsequent conclusions pretty dodgy too.
This is true of averages and strike rates. Those stats are far from perfect. Detailed play by play data would allow better analysis. Wagon wheel, pitch map, spider data contain a wealth of info what is currently under-utilized by analysts (except by international coaches).
You're still measuring and calculating from the same macro-level variables, though. They are still just descriptive statistics and speak to fairly broad trends without being terribly useful for prediction, judgement or research no matter how much data you have. You'd have to find a way to deal with the uncontrollable factors and, even then, making it defensible/credible would be incredibly difficult at such a macro level.
Once better play by play data is available to the public (cricinfo is detailed but can't be parsed by a computer for analysis), I think you will see more sabermetric work done on cricket.
Doubt it. Baseball doesn't have anywhere near the same degree of variation by inning/day/match/series in, for example, pitch conditions and their impact on the game is significant. It's all, essentially, based in descriptive stats using the same imperfect measures of what went on (runs/wickets). For deciding who won and who did what in the game, they do the job.
For research, serious data analysis, ranking of players, etc. they're pretty awful measures. Broad-based trends, sure. Definitive guides as to who's better/contributed more to wins/etc., not so much. Too many uncontrollable/unmeasurable factors. I know some people have tried to apply the same logic as sabermetricians do to cricket but they mostly end up being based on some very questionable assumptions and raising more questions than they actually answer because of it.
One example; I'd say being able to compare, say, a year's worth of pitchmap/wagon wheel data on Steve Harmison won't give you a statistically defendable answer to whether his short balls are more likely to contribute to him taking a wicket for himself, let alone his team. Got no way of accounting for pitch/weather/shot variation in a given single game, let alone being able to extrapolate that data over an entire series/season/year. You could absolutely conclude some very broad trends but that's about it.
Complete bollox of course. But worth it nonetheless for the comedy value of the one-eyed Australian accusing anyone else of "bias."
