An alternative to the Random Number Generator use in game

[MsSmart]

Many games and even scientific analysis had always had difficulties with the Random Number Generator (RNG) being less than random, and seeming to get stuck in a much narrower band than 0 to 100. As a result in many applications the use of the RNG has been abandoned for the use of Tables. These tables have randomly generated values, but each table ensures that no number is repeated, and thus all values can be found on the table.

 

If the system were to use tables instead of RNG, it would get rid of that pesky "I have a 95% chance to hit" but I am missing 20% of the time, and boy those mobs seems to be nailing me much more often than they should (may be a perception) and also would eliminate to need the so called streak breaker, since the 5% or less would happen guaranteed 5 times in a hundred, which would award the auto-hit.

 

Only a suggestion...

 

In Dark Age of Camelot, players are allowed to enter the RNG seed of their choice in order to attempt to change where the RNG is stuck on stupid. But I prefer the use of tables if possible.

 

Sue

[Sunsette]

I don't think a table would fully address the issues with CoH's randomness, as the issue is confounded by the fact that the game makes a lot of secret to-hit checks which streakbreaker can get broken on as well as the fact that it struggles to update quickly enough in fast-paced scenarios and will sometimes pull the same seed multiple times in a short period of time. 

 

But I do think a table, if it wouldn't somehow break the game, would be a great idea. 

 

To be honest, I just think the forced 95% hit cap is a bad gameplay experience that would be best eliminated, but I'm sure that would catch someone's ire. 

Edited March 11 by Sunsette

[American Decoy]

"These tables have randomly generated values, but each table ensures that no number is repeated, and thus all values can be found on the table.

 

If the system were to use tables instead of RNG, it would get rid of that pesky "I have a 95% chance to hit" but I am missing 20% of the time...."

 

So, if the table system will not allow repeated values. the result for a hit-or-miss check would be "I have a 95% chance to hit" but I am missing 50% of the time."

 

RNGs are not perfectly random, but if they were we would still get repeated numbers.  And we would still get 95% chance items that are not 95% results in our limited sampling.

[megaericzero]

Wouldn't tables introduce the potential of RNG manipulation? At least inasmuch as card counting.

 

Not that it would necessarily be easy but that it would become possible.

[Sunsette]

5 minutes ago, megaericzero said:

Wouldn't tables introduce the potential of RNG manipulation? At least inasmuch as card counting.

 

Not that it would necessarily be easy but that it would become possible.

 

Theoretically, but not practically, to a point where the amount of effort required would be immense. Purchased random generation sets are *extremely* large, it's not a small lookup table.

[RelativeQuanta]

More recently, there are true random number generators that use stuff like thermal readings of your CPU as another option.  

[lemming]

Curious on implementation:

Would each entity get a table that they would then run thru, or are tables shared?

 

I suppose what one could do is when an entity is generated, a list is generated that has unique 1-100 (or 1000, so 0.1 to 999.9) and then you pop off the numbers as they are used.  If list gets to under a threshold, a new list gets tacked on.  (And these would be generated constantly and given out to requesting entities)

 

Not sure if this would really solve stuff, but maybe?

[MsSmart]

19 hours ago, megaericzero said:

Wouldn't tables introduce the potential of RNG manipulation? At least inasmuch as card counting.

 

Not that it would necessarily be easy but that it would become possible.

They would not, the use of tables actually replaces RNG, it guarantees no repetition of values yielded by an RNG source, and thus over time, as you play you will experience all 100 values, and then table repeats itself or a new table is used. In general it is good practice to have multiple tables, so "card counting" is pointless.

[MsSmart]

17 hours ago, lemming said:

Curious on implementation:

Would each entity get a table that they would then run thru, or are tables shared?

 

I suppose what one could do is when an entity is generated, a list is generated that has unique 1-100 (or 1000, so 0.1 to 999.9) and then you pop off the numbers as they are used.  If list gets to under a threshold, a new list gets tacked on.  (And these would be generated constantly and given out to requesting entities)

 

Not sure if this would really solve stuff, but maybe?

Excellent question, if tables are used, there is no need for the streak breaker. In live days, the reason for the streak breaker was because the RNG was less than random and it itself would get stuck on narrow bandwidth of results and thus prompted the need for a streak breaker. With the streak breaker software removed, then the reading of the tables is smooth. I would prefer a table to an entity to ensure everything has the right variety of rolls and no odd repetitions occur.

[MsSmart]

17 hours ago, RelativeQuanta said:

More recently, there are true random number generators that use stuff like thermal readings of your CPU as another option.  

it still does not ensure proper number distribution, its just another seed.

[MsSmart]

21 hours ago, American Decoy said:

"These tables have randomly generated values, but each table ensures that no number is repeated, and thus all values can be found on the table.

 

If the system were to use tables instead of RNG, it would get rid of that pesky "I have a 95% chance to hit" but I am missing 20% of the time...."

 

So, if the table system will not allow repeated values. the result for a hit-or-miss check would be "I have a 95% chance to hit" but I am missing 50% of the time."

 

RNGs are not perfectly random, but if they were we would still get repeated numbers.  And we would still get 95% chance items that are not 95% results in our limited sampling.

I am not sure, I understand you. Your third statement would be erroneous in its assumption, if you have a 95% chance to hit, using a table, you will get 95% of the time hits.

[megaericzero]

2 hours ago, MsSmart said:

They would not, the use of tables actually replaces RNG, it guarantees no repetition of values yielded by an RNG source, and thus over time, as you play you will experience all 100 values, and then table repeats itself or a new table is used. In general it is good practice to have multiple tables, so "card counting" is pointless.

Right. If someone can keep track of when their table is rolling over and knows whether they've gotten more desirable or undesirable numbers since the last rollover, they can extrapolate the probability of the upcoming rolls. You could also potentially poke at it with things like Secondary Mutation - you'd know when you got the low roll because you turned into a monkey.

 

Again, not that it would be easy. I suppose the biggest counterargument is that the sheer number of evasion checks happening in TF/trial content would continually roll your table over, making tracking it moot. At that point there's also so many people, though, that tables are also probably moot since you would average as a group anyway from the sheer number of people doing things that need rolls.

 

I never took statistics in school, though. Probably should've. 😛

Edited March 12 by megaericzero

[RelativeQuanta]

3 hours ago, MsSmart said:

it still does not ensure proper number distribution, its just another seed.

It's not quite that simple.  The thermal values are truly random and seed a CTR_DRBG with AES random number generator which provides values that are indistinguishable from true random numbers.  At least, Intel's does in the link I gave earlier.  You could also use the thermal values to create your random numbers directly, if you want something non-deterministic.

[MsSmart]

On 3/12/2024 at 1:07 PM, megaericzero said:

Right. If someone can keep track of when their table is rolling over and knows whether they've gotten more desirable or undesirable numbers since the last rollover, they can extrapolate the probability of the upcoming rolls. You could also potentially poke at it with things like Secondary Mutation - you'd know when you got the low roll because you turned into a monkey.

 

Again, not that it would be easy. I suppose the biggest counterargument is that the sheer number of evasion checks happening in TF/trial content would continually roll your table over, making tracking it moot. At that point there's also so many people, though, that tables are also probably moot since you would average as a group anyway from the sheer number of people doing things that need rolls.

 

I never took statistics in school, though. Probably should've. 😛

I did, it was part of my master degree. But I picked up the tables, when studying cyber courses, in where they clearly teach the superiority of the use of tables

 

Hugs

 

Sue