- Replace decennial graphs with loess
- add draws to the dependent variable for tests
- are toss decisions suboptimal?
- investigate if people use cheap heuristics with some auto-corr or regressing on short-term sequence of prior outcomes. here's a paper that has a small behavioral model of decision making fit to data: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3581750 (see also another previous paper). captain f.e. plausibly suggests the same though captain f.e./RIFLE and captain advantage, ala the Rahane estimator, are interesting in their own right.
- we have a plug-in adjustment for runs/wickets, etc. for toss advantage. why not make the case for it? as the analog is D/L. we can make the case for a data driven adjustment which leverages all the data and also show that D/L is a high variance, low bias adjustment when the optimal thing in matches plausibly should be low variance (which means the adjustment looks absurd on its face to most people), slightly higher bias.
- Does the advantage accrued from winning the toss vary by the captain? Can some people choose better?
- Over time: Till we reach the modern era when data science and coaching staffs exploded, I imagine there was a bunch of cross-captain variation