I bet if you take the last 20 hands of a ring game and average the pot you will find that it is close to the published, but not exact. Keeping 20 numbers in a table, throwing out the oldest, adding in the newest, and averaging is correct, but a computing pain in the butt. The usual way it is done is take 85% of the old pot average and add 15% of the current pot to get a pseudo 20 hand pot average. Uses only 5% of the memory for storage and 1/7 of the FLOPs. Only programmers would give a rat's patootie about the difference in methods since the results shown are so close to the same.