I didn't computerize until 2002. For video poker I did everything with a calculator and scratch pad, except for the few games you could buy strategy cards for.
I remember that Lenny Frome article on Flush Attack in Card Player Magazine.
I did my first empirical study on a slot machine in 96/97. It was on my first slot play, the quarter Piggy Bankin' game. I started out playing when there was at least 50 coins in the bank. Since I was making a healthy profit I kept lowering the number. I wondered just how low I could go so I collected stats on 20K spins.
The main game had a 69% return.
The bank averaged breaking every 90 spins.
The Piggy Bank started at 10 coins so was worth 10/90 = 11.1%
1 unit was added to the bank every 11 spins, 1/11 = 9.1%
That totals to a payback of 89.2%
So a bank with 21 coins in it put the game at 100% plus 9% meter movement. I stayed well above 21 coins in the bank.
A few years later par sheet information showed the quarter pigs returned 89%, which was the number I came up with. I've been using the same methodology on slots ever since. When you don't have any other information an empirical study is called for.