Possible Categorical RNG

sayswhoto · May 26

Background
I recently read a forum post in General that explained that there is a high likelihood the RNG system is being sub-divided into categories for opening certain item containers. As an example, what this means is that the game first does an RNG roll for the type of weapon (example: one hand weapon) and then proceeds to roll for the specific weapon (example: sword, mace, dagger, etc.). This could also explain the higher chance of getting something like a shield since there is only 1 type (shield) with one specific item in that category (shield). I will use this later as a potential explanation of the relatively more common occurrence of Hlaalu Kwama Egg/Ebony Mine showing up in repeats since Hlaalu only has one type of contract card.

I want to emphasize this is a guess at what the RNG is like. A categorical RNG system could not be what most players expect and might help explain why certain events like chaining contract cards occur more frequently.

How This Applies to ToT
I suspect ToT uses the same categorical RNG system and will outline some qualitative reasons later. An example of what a categorical RNG system could be for ToT is shown here:

The interpretation of this table is that a spot in the tavern is first determined using an RNG roll for Patron. There are 4 patrons and 1 Tavern (which I count as a Patron). That is the first Sub-Probability column, each of the five has a 20% chance. Next, an RNG roll is done for what type of card it will be. I just have them listed as Card, Contract, or Agent, each with a 33% chance. Finally, among either Card, Contract, or Agent, the specific card is rolled for, with X%, Y%, Z% chance depending on how many are currently available. You'll notice the Tavern is always 100% a contract card since it has no other type.

How This Affects Playing ToT

Just from an intuition standpoint this can be confusing. We see the animations of the cards being shuffled and we tend to assume the probabilities would be like a typical card draw. If there are 1 to 4 copies of each card within the starting 100 cards (I believe it's 100 at the start but could be wrong), we tend to associate them with initial probabilities on the order of 1/100 to 4/100. For example, the relatively common occurrence of a repeat Hlaalu Kwama Egg/Ebony Mine:

Typical Card Draw Probability for Drawing the 2nd Kwama Egg/Ebony Mine (assuming 4 copies, and): 3/99 = 1/33 or once in 33 attempts, given close to a full deck.
Categorical RNG for Drawing the 2nd Kwama Egg/Ebony Mine : [(20%)*(33%)*(3/3)] = 0.066 or 1/15 or once in 15 attempts.

1/15 feels closer to what I encounter versus 1/33. These numbers will change depending on how the cards are drawn and discarded, but it gives an example of why this event could be more common for categorical RNG. Keep in mind the categorical RNG system can also have a tendency to keep the chances around 1/15, where the typical RNG system might go down to something like 1/60 as the game progresses.

For those that are unfamiliar, I chose this example because repeat Hlaalu contract cards when combo'd with just one other Hlaalu card can award around 9-12 gold just from the contract cards themselves. This is an absurd amount of gold especially for the early game and can easily lead to a game winning advantage.
From the example above, chaining contract cards is more potent and common than one might expect. This isn't just for Hlaalu that effectively only has one type of contract card. It is also for decks like Orgnum and Mora where the contract cards are similar in effect (power generation) and combo off each other.
A patron's card type might still appear even though players are buying all the patron's cards. Think of games where Crow cards constantly show up.
I could go on, but I'll leave the rest of it up to the reader to ponder.

Possible Reasons

My guess is that all of ESO uses the same categorical RNG system. I don't have a background in software, but I believe this would help standardize the RNG implementation and make it easier to troubleshoot and update systems, versus having to go through each team's unique RNG configuration.
"Truly random" RNG in ToT might lead to odd games. I haven't really investigated this but one could use some RNG simulator for numbers 1-100 (take the chosen number out and decrement by 1 each time to account for card draw to make it more accurate) and see what the game could look like.
There might be modifiers appended to the RNG numbers to make this closer to what could be expected from shuffling cards.
I also suspect the 5 card draw spots in the Tavern were initially there to communicate a more categorical approach (1 spot for each Patron + 1 Tavern spot). However, it may have been removed for simplicity of presentation.

Conclusion

Again, this is a guess at the RNG system.
A categorial RNG system is not intuitively expected and has different chances (often higher chances) of drawing certain cards. This is especially important when chaining contract cards can completely swing the direction of the game. Therefore, players have a critical need to understand what is the underlying RNG system.
It may still be an RNG seeding issue causing the duplicates, or both.
To prove this further would take more work than I'm willing to do for ToT.

Personofsecrets · May 28

Thank you very much for reviewing this aspect of ESO. It certainly is possible that this type of RNG system is the TOT tavern in a way that may bias it toward certain outcomes and away from a more true RNG.

And I think that if there is categorical RNG, that there is something beyond that. The reason for thinking this is that it isn't just contracts that seem to show up as repeats. As you noted in a video, the game was just repeating all cards. Indeed, I've had a large number of games where all 3 Philanthropy cards show up quite early.

The categorical RNG ideal would cause repeat contracts to happen more since most decks only have 1 or 2 different contract cards. Why there are repeats of non-contracts though seems to be a different RNG seeding issue. My theory is that the seed isn't properly refreshed.

Also, I haven't written this before, but I would like to point out that some cards seem to full under the same name, tag, category or however one wants to put it, even though the cards are different. For example, Plunder and Pilfer always seem to show up along with Squawking Oratory. Because the cards are similar, I feel as though something during the design process, at a very base level, is common among these cards and whatever that is is a factor that the RNG generator is looking at. I've noticed this with Stonelore Rockseer and Haruspex as well. They always seem to show up together.

I'll keep my official repeat card theory to only have to do with same name cards and their morphs because going beyond that feels too speculative to me. That said, I notice what I notice.

Necrotech_Master · May 28

dont want to poke too many holes, but it wont be an even 33% chance between card/agent/contract as each deck varies in how many contract cards it has

i also wouldnt differentiate between agents and cards/contracts, as an agent can be a normal card or a contract card

my other post in another thread breaks down each deck and amount of contract cards (i count contract agents as contract cards, im only differentiating between contract/non-contract)

sayswhoto · May 29

Personofsecrets wrote: »

The categorical RNG ideal would cause repeat contracts to happen more since most decks only have 1 or 2 different contract cards. Why there are repeats of non-contracts though seems to be a different RNG seeding issue. My theory is that the seed isn't properly refreshed.

That's a good point, and I agree that the RNG seeding issue is also a likely cause. It's difficult to tell without some sort of official response from ZOS.

Personofsecrets wrote: »

Also, I haven't written this before, but I would like to point out that some cards seem to full under the same name, tag, category or however one wants to put it, even though the cards are different. For example, Plunder and Pilfer always seem to show up along with Squawking Oratory. Because the cards are similar, I feel as though something during the design process, at a very base level, is common among these cards and whatever that is is a factor that the RNG generator is looking at. I've noticed this with Stonelore Rockseer and Haruspex as well. They always seem to show up together.

For a while, I also suspected the cost of the card played a role in influencing the next draw. I stopped looking into it since it's just too much time spent trying to figure out ToT.

Necrotech_Master wrote: »

dont want to poke too many holes, but it wont be an even 33% chance between card/agent/contract as each deck varies in how many contract cards it has

i also wouldnt differentiate between agents and cards/contracts, as an agent can be a normal card or a contract card

my other post in another thread breaks down each deck and amount of contract cards (i count contract agents as contract cards, im only differentiating between contract/non-contract)

The 33% chance is from the division of categories (3 categories being Card, Agent, and Contract). The final RNG roll would take into account number of cards per category.

Yes, there are contact cards that can be agents. The table I gave was an example of what the system could be like. I mainly wanted to show how a categorical RNG system could lead to different probabilities versus if it was just a random draw from about 100 cards. What the actual RNG is (if it is categorical), would probably be different.

Personofsecrets · May 29

sayswhoto wrote: »

Personofsecrets wrote: »

The categorical RNG ideal would cause repeat contracts to happen more since most decks only have 1 or 2 different contract cards. Why there are repeats of non-contracts though seems to be a different RNG seeding issue. My theory is that the seed isn't properly refreshed.

That's a good point, and I agree that the RNG seeding issue is also a likely cause. It's difficult to tell without some sort of official response from ZOS.

It could also be that if something like contracts and non-contracts have their distinct categories that there are sub categories within them or just the non-contract cards.