So, if any of you are on the DLB group in facebook, you may have seen Jamie lose his cool at a customer earlier in the week who claimed that the dice roller was bias. One interesting tidbit that came out of this is that Jamie shared a page on DLB online that he’s been using since mid-May to keep track of ALL the dice rolls in the online game. That page is updated every 10 minutes, and linked here:

http://www.dynastyleaguebaseball.com/DiceStats.aspx

Lots of interesting stuff can be done with the data, but simply, I wanted to know which “ranges” of dice rolls have been “overachieving” (ie being rolled more often than expected) or underachieving (ie not being rolled enough). Before I get to the results, I should note that I did a binomial distribution test (which google tells me is appropriate for determining if dice are bias; it’s been too long since my first year university stats class!) and even with the variances that show up on the page above, the dice roller is NOT bias (ie the variances are explained by random chance given the number of times the dice has been rolled and the distribution so far).

All of that said, here’s a pretty graph:

So on the x-axis is arbitrary groupings that I created for dice ranges. On the y-axis is the actual frequency that a roll has taken place within that grouping. The vertical scale is exaggerated so the differences look stark, but you can see the variances around the theoretical expectation that each band of 100 would be rolled 10% of the time.

So, what does this mean? Well, any hitters that have a lot of numbers from 300-499 could be overperforming their cards. Numbers in the typical HR/BB ranges (100-300) are slightly down. Same with numbers usually associated with K’s on some pitcher’s cards (800-899 is the biggest underperforming range by far). Hey! Maybe that’s why Strasburg sucked for me (he’s got K’s in that range, at least against RHB from my memory…).

Obviously, it’s misleading to draw those kinds of conclusions from the above data since the data is for ALL rolls in online, not just the BLB, and the data only goes back a month, but it’s still interesting to see.

Mike