These are results produced by my Playful AI simulation program. The first
line in each set specifies the settings for the run. For example,
"0 1 2 / 100000" means that the program was told to set aside zero dice when
a 1 is rolled, to set aside one die when a 2 is rolled, to set aside two dice
when a 3 is rolled, and to run the simulation for 100,000 trials.
   The win percentage is the percentage of trials when you would gain at least
one bit. Most Dice Rolled indicates the most dice rolled during any one of the
trials, and Most Bits Gained indicates the max number of bits gained by any
of the trials. Everything else is fairly self-explanatory, feel free to e-mail
me with any questions/suggestions: hedden@cs.utk.edu
   Much thanks to Steve Kertes for running these simulations for 10 million
trials.

0 1 2 / 10000000
Average Bit Gain/Trial: 0.999086
Win Percentage: 49%
Average Bit Gain/Win: 1.9992
Most Dice Rolled: 63
Most Bits Gained: 46

1 2 2 / 10000000
Average Bit Gain/Trial: 0.998453
Win Percentage: 30%
Average Bit Gain/Win: 3.29906
Most Dice Rolled: 537
Most Bits Gained: 111

0 0 0 / 10000000
Average Bit Gain/Trial: 0.999646
Win Percentage: 49%
Average Bit Gain/Win: 1.99989
Most Dice Rolled: 1
Most Bits Gained: 3

1 1 1 / 10000000
Average Bit Gain/Trial: 0.999086
Win Percentage: 39%
Average Bit Gain/Win: 2.49968
Most Dice Rolled: 27
Most Bits Gained: 28

1 2 0 / 10000000
Average Bit Gain/Trial: 0.999084
Win Percentage: 30%
Average Bit Gain/Win: 3.30232
Most Dice Rolled: 57
Most Bits Gained: 39

1 1 2 / 10000000
Average Bit Gain/Trial: 0.998571
Win Percentage: 39%
Average Bit Gain/Win: 2.49863
Most Dice Rolled: 131
Most Bits Gained: 65