ASM standards

Are there any standards for ASMs such as they must consistently produce X% of randomness? If so, how is this verified and how often do State Gambling regulators verify after they are in use? If not, why not? It seems to me that they do a very poor job of randomizing cards, at least at the 8 deck blackjack game that I have available. I repeatedly see cards of the same rank come out in the same deal all night. Please don’t tell me that I have recency bias or confirmation bias, I don’t, it is real over many sessions.

I think GLI/BMM verifies them

Quote: harris

I think GLI/BMM verifies them
link to original post

But what do they actually do?

Does every ASM device go through GLI? Do they verify its software is identical to the software for that particular type of ASM that they have previous verified? or do they perform randomness tests on actual shuffling? Do they verify there is no functionality in the ASM to alter the software?

I have never seen any technical standards organization that is so secretive about what they do. Their total lack of transparency doesn't inspire trust and confidence.

GLI does publish their standards regarding how they test shufflers.

It doesn't look like they test to see if the ASM can be altered from a cursory glance.

Quote: harris

GLI does publish their standards regarding how they test shufflers.

It doesn't look like they test to see if the ASM can be altered from a cursory glance.
link to original post

Thank you for providing this. I have never seen this before.

Very interesting, I’m glad to see that there are actual testing procedures and that the testing methods seem to be thorough and complete. However, the shufflers appear to be complex mechanical devices which would likely lead to frequent, maybe unnoticeable, malfunction. I would think that ASMs would need to be inspected and “tuned up” on a regular basis of at least a couple of times per year to ensure that they operate as designed, but I doubt that they are.

Your comment about altering is also interesting, but I’m sure there would be a stiff penalty if discovered, which would deter such actions.

This is where they fail their purpose:

“They can completely eradicate any pattern(s) introduced to the playing cards before being
placed into the shuffler that would affect the outcome of the next gam”.

technics your reading the wrong statement ... the interesting one ... Conform to the desired random distribution...

heres a sloppy answer for you to this suspicious wording...

Quote: chatgpt

Yes. **“Conform to the desired random distribution” is deliberately broad language.**

The important word is **desired** rather than **uniform**.

GLI is not saying:

> Every possible output must occur with equal probability.

It is essentially saying:

> The observed outcomes must match the probability distribution the system is designed and represented to produce.

That distinction can matter a lot.

For a conventional shuffler whose stated objective is an unbiased shuffle, the “desired random distribution” would ordinarily be the theoretical distribution of a fair shuffle. For a 52-card deck, conceptually that means the permutations should not exhibit systematic preference for particular cards, positions, adjacencies, runs, etc.

But the wording itself is broad enough to encompass a **non-uniform random distribution**, provided that distribution is actually the intended one and the device conforms to it.

For example, imagine an RNG produces integers 1–100 with this intended distribution:

* 1–50: combined probability 60%
* 51–100: combined probability 40%

That is still a random distribution. A chi-square test could test whether observed results conform to that specified 60/40 distribution. Randomness does **not inherently require uniformity**.

The surrounding GLI requirements constrain that flexibility, though. The same outcomes must also:

**a)** be statistically independent
**b)** conform to the desired distribution
**c)** pass recognized statistical tests
**d)** be unpredictable

So you cannot simply introduce a deterministic pattern or strong serial dependence and call it the “desired distribution.” Independence and unpredictability would still have to hold.

There is another interesting aspect of the language you quoted:

> “The RNG **and the physical mechanics of the shuffling device will mutually result** in the production of random card outcomes.”

That means GLI appears to be concerned primarily with the **final card outcomes**, not merely whether the underlying RNG generates uniformly distributed numbers.

So theoretically you could have:

**RNG distribution → mechanical transformation → final card distribution**

where the RNG itself isn't trivially mapped one-to-one onto cards, as long as the **combined system** produces outcomes satisfying the required statistical properties.

And that's probably why GLI uses **“desired random distribution”** instead of something much narrower like:

> “Every possible card permutation shall have equal probability.”

Those would be substantially different requirements.

If you're looking at this in connection with the **biased-card-deal / controlled-distribution concept we discussed before**, that particular wording is potentially significant. The next question would be **how GLI defines or constrains what counts as the “desired” distribution elsewhere in the shuffler standard**, because §2.1.1 by itself does leave considerable semantic room.