Holy Avenger (+8) not Vorpaling at 20%

I am using the Holy Avenger a lot since the weapon upgarde. The Vorpaling is fun, when it happens, and that is very rare.

The upgrade says: 20% chance of lethal to first enemy

Now, since the update i always felt it was not 20%. That would be 1 out of 5 cast, and it’s certainly not that.

Of course, i played long enough to have a feel about %, with so many effect happening 7%, 10%, 15%, 20%, 30%, etc…
But i also knew that coming here complaining would be answered by the “observation bias” theorists.

So i compiled stats.
300 casts, to be exact. None while stunned.
The result is 8% decapitation (24 out of 300).
It’s not 1000s, i know, but i feel like 24 is far enough from the theoric target, 60, to see that it’s not working at 20%.

So, either it’s wrongly coded, OR, it’s not intended at 20% and the text is wrong (could be the later, 20% seemed a bit strong).


How detailed are the stats you’re taking? I’m asking because like, do you know how often the target was barriered? I’ve been playing more than a year and I’m still not sure if barrier will protect against that.

I’m sure someone else will chime in with other potential explanations, it’s just the one that popped into my head.

The weapon deals damage first, removing the barrier, and the Vorpal happens after.
Also, if you kill the first troop on damage, then you Vorp the next one.

Can someone who actually understands statistics comment / explain whether 300 is a significant enough sample size to draw any real views from? I suspect not but my stats knowledge is at least twenty years rusty from disuse.

1 Like

300 can be enough if the departure from expectation is large. Based on a quick check (i.e. I went to a random website), 24 or fewer successes on 300 tries is less than one in a million for a 20% chance of success. I’ll goof around with the weapon and see if I get similar results.

If you do, make sure the kills registered are from Vorpals. If you play on high speed, it can be confusing as 4 things can kill the first troop:
1- The weapon damage
2- Then the Vorpal
3- Then the skulls from the cascade of removing yellows
4- Extra damage from other team members on match (exemple Kracken, Yoa-G, etc…)

1 Like

Smaller samples are tendencious to be streaky in one way or another. For example, in 100 tries you could perform a vorpal strike 25 times and then it would looks like the effect is triggering 25% instead of 20%, by the same token in 100 tries could result in 12 vorpal strikes or less…

For humans/players 300 tries seems like a lot much, but realistically it’s not, the fact remains that we get really bored/tired after so many battles, specially as we build some expectations about th results.

Like I said, 300 tries is plenty if the results are really far from expectation. If I flip a coin and get heads 20 times in a row, I am perfectly justified wondering whether it’s actually a fair coin—and that’s only 20 attempts!

The confounding factor, which I have not confirmed for myself yet, is if Vorpal is a separate damage element from the weapon (or if it’s instead treated as a 20% chance to upgrade from 25 to 9999 damage). Venar says it is separate, but I should test myself to be sure.

What you’ll see is some initial damage done, yellow gems vanishing, and then some « strike » effect that kills the troop, similar to an Ubastet kill or Assassination from hero.
If it died from the spell damage, the « strike » will insta kill the second troop instead.

About the 300, I just want to say that in just knew, from months, that this effect sure does not happen 1 out of 5 cast (20%). Sure, a feeling is not statistic, but when you compare to, say, Magavore kills, it really gives an idea.

I calculated the 300 cast before coming here just to back my feeling with numbers.

Also, if in break it into 3 x 100 samples, I get 8%, 7% and 9%, so it kinda consistent and why I stopped there.

But well it also all depends of luck. In 300 rows it’s possible to have only 24 Vorpal Strikes. Holy Avenger is a cheap/fun weapon, but if anyone wants to rely on “80% chance of nothing different happens!” you are welcome.

Reminds me of grinding drops in MMO’s where sometimes people would kill more than 3K enemies just looking for one specific item. But in the good old Ragnarok On-line a few times i got 0,03% drops in the first ten tries more or less.

Personally my Megavore is rarely “hungry”, and just as well it’s very rare when the enemy Megavore kills something mine, but it obviously has a huge impact in my experience/perception when it does. Even though i’m fully aware that these things happens and that my emotional response is reflexive i still curse and swear a lot about it…

But well, there is a chance this upgrade could be failing and not working as it should, but considering our experiences with Dragon’s Eye vs Delves i’m skeptical the devs would really give a damn about it in the sense of really making it work at 20% chance.

Agreed, but if it’s, say, 7%, like many similar effect, then maybe just change the text to 7%, as purple ingots are not that easy to come by.

Testing an hypothesis is not a trivial matter in statistics. If you use this link, you will see the OP’s result is significant with a significance level of 0.01 (which means that there is still a 1% chance that the OP’s result is due to random bad luck).

1 Like

Yeah I’m usually fast to jump on “that’s probability, you got ‘lucky’” but in this case I feel like OP’s done just enough trials to warrant someone else trying too. Unfortunately “300 matches at 1x with a specific weapon” isn’t exactly everyone’s idea of fun.

Most interesting to me is how consistent the percentages stay if he segments his 300 trials into 100. My personal opinion of the GoW RNG is it’d have been more likely to see wild variance, not +/- 1%, if the implemented percentage were 20%.

So yeah, it’s still “a small sample set” but there’s enough factors lined up here to make me :thinking: and believe it’s worth a look.

What’s a “fun” Holy Avenger team, Venar? If I commit to some trials daily I can build up a similar data set “soon” I guess.

You could play 1000 games and have the same percentage and it won’t be changed because it’s too easy for the powers to be just say “Well, that’s just RNG”.

Nah, they’ve checked the code (and corrected bugs) before when players pointed out that a set of results was extremely unlikely.

1 Like

What did you input as parameters? I can do a straightforward binomial calculation, but I don’t understand what the population variance should be in the link you gave.

I like Hero, Asha, Worldbreaker, Possesed King.

But that is to win. You kill fast and quick, and rarely use the HA more than twice.
If you just want to cut heads, use HA, double Dwarven Gates and Loyalty, and stay away from Dragon Eye teams.


My understanding of percentages n statistics in general:
20% chance doesn’t mean it will happen 1/5 times, whether if u do it 5 or 50000000 times
It just means every individual try, u have a 20% chance
Not the overall average
No one should ever expect an average result as that’s not what a % chance means
U could literally go 0/100 and it doesn’t mean the % is incorrect

Here’s the thing about that, going back to the coin example someone else said:

If you flip a coin 20 times and get 20 heads, that’s pretty astounding, but possible. I’ve done simulations before, and the end result was I was surprised how likely 20-in-a-row was over trials of even 1,000. IIRC it was almost always likely to happen if I did about 3,000 trials. Neat. Anyway, I’m digressing.

Now imagine you pick up the coin again the next day and flip it 20 times, and get 20 heads again. This is not impossible, but still weird even though we’re only talking 40 trials.

So you pick up another coin and flip it 20 times. You get 14 heads and 6 tails. That’s lopsided, but not as astounding as 20 heads. You try again. 9 heads with 11 tails. OK, we’re looking at random results!

So you pick up the first coin again. This time you get 19 heads and 1 tails. You get up to 39 heads and 1 tails.

So you call a friend and have them come over to do 40 flips with both coins. The first coin flips 38 heads 2 tails. The second flips 29 heads 11 tails. You ask them to try again and they get 40-0 and 18-22.

Something is wrong with that coin. Maybe if you flip it 10,000 times you’ll see it right itself. But if you can go 80 trials and get 79 heads, you get suspicious. It’s especially suspicious if when you flip other coins you get “expected” results, but when you resume with the suspect coin they’re tilted again.

What’s going on here is we keep trying over and over and over again. The more that one coin produces results that favor heads, the more confident we can become that it is “fixed” and we aren’t having “weird luck”.

So ideally, the way to do this experiment is:

  • Have multiple people use Holy Avenger and see if they get very consistent results.
  • Have the same people try another weapon with a 20% chance of something and see what results they get for the same number of trials.

If every person who tries Holy Avenger is seeing 7-9%, and the same people are seeing more like 18%-22% off of the other weapon’s effect, that can make us very suspicious something is wrong.

I think the main reason we don’t already have that is Holy Avenger isn’t exactly a popular meta weapon, and as has been pointed out 300 matches aren’t exactly a large sample set.

It’s still possible Venar just had “really bad luck”. To know that, someone has to try.

I’m not on board with “it’s definitely a bug” but I am on board with “if this is just luck, it’s exceptionally strange luck”. So I’m going to try an experiment. It doesn’t hurt you for other people to try and validate.