Sand Shark ability to devour way below stated "25%" rate


#1

When using Sand Shark in pvp battles I activated his ability.

Underbite says “Deal 5 damage to a random enemy. 25% chance to devour the enemy.”

So I’ve been frustrated with the low chance of devour, so I decided to track them and see how frequent it really hits. Here’s the results (recorded over 10 matches):

4 devours with 30 skill activations, which would be 13.33% barely half the stated devour rate of 25%.

Here’s the numbers by match (with match %) to those curious:

  1. 4 activations 0 devours = 0%
  2. 4a 1d = 25%
  3. 1a 0d = 0%
  4. 2a 0d = 0%
  5. 4a 0d = 0%
  6. 1a 0d = 0%
  7. 3a 0d = 0%
  8. 1a 1d = 100%
  9. 4a 0d = 0%
    10.6a 2d = 33%

totals: 30a 4d = 13.33%

Please update the tooltip or update the % to devour to match the current tooltip. thanks.


#2

Ten matches is not a sufficient sample size.


#3

That is pretty normal variance for such a small sample size. I’m afraid you just got unlucky.


#4

It does seem off. Be great of you continued testing with him!


#5

The sample size is 30 skill activations, not the number of matches. The skill activating determines the probability not the number of matches played.

If you feel that my numbers are off, test it yourself and post your results here.

More data will prove or disprove this skill not activating correctly.


#6

The odds of achieving exactly k occurances of an event in n trials, given a probability p, is a very well known statistical distribution known as a “binomial distribution.” It can be expressed as:

f(k, n, p) = n!/(k!(n-k)!) * p^k * (1-p)^(n-k)

Substituting in values from your anecdote:
n = 30
k = 4
p = 0.25

f(4, 30, 0.25) = 30!/(4! * 26!) * 0.25^4 * 0.75^26 = ~0.0604, or ~6%.

So you got unlucky, but not unreasonably so. You would need a much larger sample size to draw any meaningful conclusion that the numbers skew low. For example, if you actually saw 40 devours over 300 samples (a tenfold increase in data points at the same purported ratio), the probability of that would be so small as to defy reason, and we could begin to question the stated value of p.


#7

Feel free to try the skill out on your own. This has been going on for days.


#8

The numbers aren’t cumulative. Each time it activates it has the same chance to devour or not.


#9

Okay, I decided to humor you and do some testing of my own.

Battle 1 - 2/4 devours
Battle 2 - 3/9 devours
Battle 3 - 0/3 devours
Battle 4 - 2/7 devours
Battle 5 - 1/5 devours
Battle 6 - 1/8 devours
Battle 7 - 4/7 devours (!)
Battle 8 - 2/4 devours
Battle 9 - 2/8 devours
Battle 10 - 3/9 devours

Total: Total 20/64 or around 31%. The sample size is still much too small for certainty. You’ll notice I was at exactly 25% after the 6th battle. If I had just taken results for battles 3 through 6, I’d be at 4/23, or 17.39%. Taking the last four consecutive would have me at 11/28, or approx 39.28%. Point is, my results vary wildly in this small set just based on where I decide to keep records from. The streaks happen both ways, and I’ve had much much worse. In the past, I’ve distinctly recall having gone 9 casts in a row in the same battle without devouring with Kerberos, and he has exactly the same chance as Sand Shark to trigger the devour. It still could be less than 25% overall, you are just going to need a lot more data points for it to count as evidence.

My advice would be to leave the RNG based troops on your defense team, because it always feels like 100% chance for the AI.


#10

I wonder why people do not make a shand shark team that can quickly spam sand shark?


#11

Because Hunger is so much better.

Look at Kerberos, even with Arcane, you almost never see him in play. His spell cost is a little high for it’s effect.

I had a team with Sandshark, Great Maw, Alocyte, and Kerberos on defense. Then you realize it could be Great Maw, IK, Mercy, Sheggra.


#12

With the recent changes devour won’t work against impervious enemy and since you cannot aim it there’s chance AI will target such enemy. So when fighting against teams that contains impervious creatures devour chance will be indeed lower.