"RNG" at its best


#1

I’ve been playing arena with Black Menacles on hero. I used 22 times in 15 battles (still one remaining in second run). Devour fired only onceONE time! It is not the same ratio AI gets on devours…
Maybe our beloved devs will finally admit something is seriously wrong with this game? And those “random” number it generates are not equally random for players and for AI? And these numbers differ between game modes (GW favours AI a lot p.ex. in comparison with regular pvp)?


#2

Are you positive that devour fired only once? A lot of times troops are immune to devour and on the new UI it is sometimes hard to see when devour fires, but then gets immediatly nullified by immunity.

I’m not sure if this what is happening or not, but keep an eye out for it. You might be surprised.


#3

But in arena the troops don’t have traits, and devour immunity is only obtained via traits, correct?
Just asking for clarification that it wouldn’t actually be something someone would need to factor in if gathering data in arena. Definitely that’s a factor in regular PvP. =)


#4

Another thing to keep in mind if the damage kills the troop that would have been devoured that would for obvious reasons cause devour not to proc


#5

@Elarcadia Yes, I am 100% sure it fired once. And no, it had nothing to do with immunities…

@Vangor You are right. Still… when Menacles are used by AI it may kill troops and very often devour another. But when I play 16 arena games, and get one devour in 20+ shots - something is seriously wrong.

Menacles are only one example. Just look at Kraken, at Kerberos. Compare their behaviour for player and for AI, in different game modes… and some interesting num
bers are coming up :slight_smile:


#6

Please feel free to share your numbers.

I for one find kraken devours quite often and no more for the ai than it does for me. As for BM and kerby… I havent really used kerby since they nerfed him into the toliet and i never really used BM…


#7

I agree. When I have tracked this, I generally find that after I get to a reasonable sample size, my results are pretty-much as advertised. The problem is getting to a reasonable sample size for the AI’s attempts, since I generally play to keep them from attempting to devour me.

It’s the same thing as when I was tracking TDS resurrections. It was easy to get a good number of kills against the AI, but I hardly ever had troops die, so I got very few resurrection opportunities for myself.


#8

I don’t think this is true. Target lock doesn’t exist for any other multi-step spell in Unity like it did in Adobe, so the devour portion would more than likely just select a random target from the survivors.


#9

Thats a good point and yet another thing we could use some dev clarification on… @Ozball?


#10

Don’t think I need to, I think everybody sees what it is like.
If you want to defend game mechanics and say thay work well - it is your right and your choice :slight_smile: I am not here to create some sh*tstorms :slight_smile:


#11

But think of all those brown gems you’d get.


#12

Buckle up and get ready for some

M A T H E M A T I C S

There is a 4/5 chance that Black Manacles does not devour a troop. The chance of not devouring on all of N casts in a row is 4/5 to the power of N.

(0.8)^22 = 0.00738

That’s slightly more than 1 in 136 (0.00735). If you have a very very large number of players who each cast Black Manacles 22 times, you can expect roughly 1 player with zero successes for every 136 players with at least one success each.

But we can go deeper. If a group of players all cast black manacles 22 times, how large does the group need to be for a zero-success to occur more often than not?

The chance of a random set of 22 casts containing at least one success 1-0.00738 = 0.99262. The chance of two sets is 0.99262 squared, or 0.98529. Here is where we use a logarithm.

If x = b^y, then log b (x) = y. We want to solve 0.5 = 0.99262^y. log 0.99262 (0.5) = 93.575.

If 94 or more players all cast black manacles 22 times, there will probably be one of them who succeeds zero times.


#13

See, that’s where you have it exactly backwards. You are the one making a claim (“something is seriously wrong with the RNG”), so now you have to prove your claim - the default assumption otherwise is that your claim is false.


#14

Lol you are the one who said “some interesting numbers are coming up”

Saying that and refusing to share those numbers leads me to think that youbdont have any numbers at all…

Its not about defending anything. I made no claims here. I didnt open a topic to say something wasn’t working right and then refuse to show any evidence of it.

My whole point, and ppl are getting tired of these two words, was that what you are experiencing is called recall bias.

Now since it is clear to me now that your only objective here was to complain and you wont listen to any kind of logic without getting indignant i will be muting this thread.

Have a lovely afternoon
Vangor


#15

how many players must cast manacles 22 times, for it to be “probably one” of them who suceeds in all 22 times?
just out of curiosity


#16

Well, we start with the chance of just one set being all successes.

0.20^22 = 0.0000000000000004194304

That is 1 in 2,384,185,800,000,000. There is then a 2,384,185,799,999,999 in 2,384,185,800,000,000 chance (0.9999999999999995805696) that a set will have at least one failure. Now it’s time for a logarithms. Praise wolfram alpha for being able to handle these numbers, because every other calculator I’ve tried just doesn’t work here.

1,652,591,658,973,563

If 1.653 quadrillion players all cast black manacles 22 times each, there will probably be at least one player who devours an enemy every time. That is over two hundred thousand times more people than the entire world population.


#17

So… you’re telling me there’s a chance. Yeah!


#18

i have been playing this game for a month now and i will state categorically that there is a built in bias in the programming for the computer player,

Ally


#19

Extra, extra, read all about it! The plural of “anecdote” has changed; it’s now called “data” - You heard it here first!


#20

Step 1 - “make a claim” - is the easy part. Check. Now time to move on to step 2 - “prove it”.