MobileGamer Rey's Hero's Event w/CrazyExcuses

Can someone explain why I should watch this? I thought more information on Rey's event would be provided since Kozi posted it, but it took about 10 seconds to realize that's not the case.

I took it as a hint...

Yeah, you're probably right.

It’s worth watching for the explanation of why doing 3-sims at a time is just silly.

"NicWester;c-1332991" wrote:
It’s worth watching for the explanation of why doing 3-sims at a time is just silly.

While I 100% agree that the 3-sim myth is rubbish, she really offered no explanation other than "I don't even know how that is possible." Not a convincing argument.

Any theory about sims is automatically silly.

"EventineElessedil;c-1333013" wrote:

"NicWester;c-1332991" wrote:
It’s worth watching for the explanation of why doing 3-sims at a time is just silly.

While I 100% agree that the 3-sim myth is rubbish, she really offered no explanation other than "I don't even know how that is possible." Not a convincing argument.

Yep. The video was not very compelling on that point (or any point). I guess I can buy it being Kozi's stealth hint that all the speculation about Rey's journey is correct. Seems like the type of thing she would do.

Although I respect her theory and I even support it, having a coding hack for simming is not that hard: the server receives a request for a number of trials, nothing prevent the servers to have different distribution of rewards for number received.

In a more technical terms, nothing prevents devs to create conditional distribution given the number of sims and make sure that the join distribution still has an overall drop rate of 1/3 by punishing even numbers of sims.

I agree it would totally silly and unprofessional to do it, but this is not impossible, differential pricing exists after all.

And if you really want a code snippet (in R as I assume she uses it):

# Let's say you get a shard if the sample gives a 1 
# assume p < 0.5
compute_shard_rewards <- function(sim, p) {
  
  m <- sim*p + min(1, 0.1/abs(sim - 1/p+1e-8))
  s <- sim*p*(1-p)
  return(max(0, min(sim, floor(rnorm(sim, m, s))))
}

I did not tune it to have expected value of one third, but I guess you can do with a bit more of time :-)

I still say this is complete non sense and my guess devs just draw from multiple random independent coin flip.

reworked algorithm to take out the hard coded equality.

I hope the mobile gaming guy is okay he didn't look well. And yes I think we will be needing that line up I will catch it next time

"Neo2551;c-1333040" wrote:
Although I respect her theory and I even support it, having a coding hack for simming is not that hard: the server receives a request for a number of trials, nothing prevent the servers to have different distribution of rewards for number received.

In a more technical terms, nothing prevents devs to create conditional distribution given the number of sims and make sure that the join distribution still has an overall drop rate of 1/3 by punishing even numbers of sims.

I agree it would totally silly and unprofessional to do it, but this is not impossible, differential pricing exists after all.

And if you really want a code snippet (in R as I assume she uses it):

# Let's say you get a shard if the sample gives a 1 
compute_shard_rewards <- function(sim, p) {
  rewards <- rep(0, sim) 
  if (sim == 3) {
    rewards <- c(1, rbinom(2, 2/3*p)) # make sure you have 1 shard but scale down the odds to not make it too obvious
    return(sum(rewards))
  }
  return(sum(rbinom(sim, 19/20*p))) # max number of sim is 20 so   
}

I did not tune it to have expected value of one third, but I guess you can do with a bit more of time :-)

I still say this is complete non sense and my guess devs just draw from multiple random independent coin flip.

She did mention that she believed the only way this was possible was if it was hard coded in. Which is exactly what you’re suggesting with that code