Can a slot machine be aproximately mapped by sampling the decisions of the reels?

Can the aproximate odds of a machine be determined by observing the outcome of the reels and apoximating the reel mapping?

In many cases yes. It just takes a long time.

I think that you can get a good idea of the symbol weights if you were to track a few hundred spins. Most video slots don't have more than about 50 or 60 symbols per reel and you can see three symbols per spin. Once you have an approximate symbol weighting, you can calculate the return percentage.

There is one casino downtown that puts a return placard on the side of each machine.
Have you ever thought to test your calculations against a known value?

As I recall, the Wiz did just that with a real reel machine.

I think it would be interesting to see real tests. I was thinking that it could be done with the Wizard's Atkins game, for which we know the return. As I recall, that game has rather short reel strips and I have the math in Excel.

How many decisions do you think you would have to observe to get a rough estimate of the reel weights on say a 64 stop machine?

Quote: FleaStiff

There is one casino downtown that puts a return placard on the side of each machine.

Which one?

Quote: Jim123

How many decisions do you think you would have to observe to get a rough estimate of the reel weights on say a 64 stop machine?

It would take 303.4 spins on average to see every possible stopping point per reel. I've actually thought about doing this. However, casino security usually gets nervous about people taking a lot of notes at the machine for an extended period of time. What might work better would be to record about 500 spins on a cell phone and then go over them carefully when you get home.

Quote: Jim123

How many decisions do you think you would have to observe to get a rough estimate of the reel weights on say a 64 stop machine?

The Wizard did a sampling of 4000 samples in Slot Machine Appendix 1 for a 128 stop machine. Using what he knew about design, he was able to make a reasonable guess at the settings to calculate the probabilities. It's all explained in detail in this appendix.

When he did his survey of the slot machines in Vegas, he has the PARS for four different machines. He was able to write a computer program to determine that some unique combinations would show up on the display depending on which house edge the slot was set at. He could find one of these unique combinations, with as few as one coin, but usually after 5 or 6. That was how he was able to sample all the machines of a particular type in all the different casinos. He recorded the settings, but he did not betray the manufacturer by telling how you would determine the setting for an individual machine.

He also observed that casino operators had a strong tendency to set all the machines of a particular brand to the same house edge. His finding disputed the commonly held belief that hot machines were placed near the entrances to the casinos or near the buffet.

Without the PARS sheets it would have been an impossible task to sample all the machines in multiple casinos. It would simply take to long if you had to play 4000 coins in each machine.

There seems to be confusion whether we're taking about video slots, a weighted single-line slot, or both. Since video slots are more popular and would be more easily reverse engineered, I think it makes more sense to focus on those, for the sake of discussion.