Lorem ipsum dolor sit amet, elit eget consectetuer adipiscing aenean dolor

# Holy Avenger (+8) not Vorpaling at 20%

I will defer to someone who can explain, with math, why 300 is sufficient, but I find on the GoW forums people don’t tend to respond well to datasets that are in the low hundreds.

For my purposes, seeing 5-10 people make 50 trials would probably be enough. If 10 people see 7% that’s really suspicious. If 9 people see 7% and 1 person sees 20% that’s, “Whoa, really? More data.”

I’m going by feel, I don’t really know the math behind what makes a sample size strong.

There isn’t any difference here between 6 persons making 50 trials and 1 person making 300 trials, provided all trials were subject to the same conditions. The overall outcome was 24 hits out of 300 attempts.

The question here is “Assuming the lethal rate is 20%, how likely is it to only get up to 8% out of 300 attempts?”. It’s really next to impossible, not just unlucky. The sample size is 60 times the hit chance, so deviation from the mean value will fall off fast. To give a (somewhat unrelated) comparison, this is like testing mythic drop rates with 60000 gem keys.

Edit: For a bit more math, this is like entering “output [count 1 in 300d5]” into https://anydice.com, setting data to “At Most” and checking the percentage number for the #24 row. Unfortunately it times out, I believe the output would be 0.00%.

3 Likes

I’ll say this for maths.
I know that sometimes, there is bias, and intuition doesn’t mean anything. I’m a science kinda guy.

But I also feel like player « experience » is too quickly brushed off.

We play this game. Many of us for a looong time. Our perception must count for something.

For exemple, play 2 dozen games with an Assassin using the last tree bonus.
That’s 10% kill on skulls, and 7% kill on kills.

You’ll see some deaths. Sometimes. In fact, 10% and 7%.

Then play a while with the Holy Avenger. You should « feel » way more kills, since it’s 20%.

But not so.
It’s not science. At all. But when 100% of players using the weapon « feel » it’s not working right, and not a single player experience otherwise, it should be looked into.

Sadly, it’s a purple Ingot investment, and a not so good weapon, so very few players can try it.
If the Vorpal was on Dragon Eye, this forum would be full of crying

Edit: also, I know fixing this sword is low priority, and I agree with that. That is why I suggested to fix the tooltip to 7%, which should take 5 seconds, and will stop leading players astray.

1 Like

My argument is if the RNG that determines this is client-side (and I can’t imagine why it wouldn’t be), then the “same conditions” requirement is invalidated. I’m not sure what kind of entropy sources the GoW RNG uses, but it’s logical different devices, different matches, different moves, etc. don’t count as “the same conditions” when dealing with a PRNG. One forum member (I think it’s akots? I can never remember how to spell it.) has suggested the game doesn’t directly use its RNG output.

So I feel like if multiple people gather the data, we have to consider 3 possible outcomes:

• The probability is strongly indicated to be 7%.
• Some conditions lead to 7%, but some conditions lead to 20%.
• The probability is strongly indicated to be 20%.

I think we’ve only definitively ruled out one of those cases. The middle case is the hardest, because it means it’s possible a dev tested a few hundred matches, saw 20%, and called player bullshit.

It doesn’t matter in this case, the tests prove that there is at least one Holy Avenger that is not vorpaling correctly. There could be any reason, something as complicated as an alternate reality pocket or something as simple as an incorrect text description. It’s up to the devs whever they want to call Ghostbusters or invoke Occam’s Razor and go for the most likely cause.

A very meticulously researched bug report has been filed, going “duh, me no get it, moving this to Feature Requests so it doesn’t cause any work” feels quite insulting.

@Kafka @Cyrup Could you please finally move this back to Bug Reports where it belongs? This kind of community treatment is making the game look exceptionally bad, if you don’t want to handle bug reports, just shut down the whole category.

2 Likes

We’ve done a lot of testing and made a lot of notes about our RNG and statistics around the forum in the past, I’ll link you to a couple of those resources now, some of them may feel irrelevant because they use different game features as context but it all applies the same way to Vorpal:

This fantastic post by Gold Phoenix which explains just SOME of the testing on our RNG that we’ve done:

Basically we’ve tested all sorts of probability based things in the game tens of thousands of times and we haven’t found an issue with our probability based features EXCEPT in any circumstance where the probability was not set correctly in the settings - I’ve repeatedly asked different team members to check the probability set for Vorpal and every time the answer has come back to 20% is set AND is working as intended.

Thanks for the response. I guess we need a more detailed explanation of how Vorpal is intended to work. Because it is (was?) not a straightforward “20% chance for lethal damage any time the spell is cast”. Or we players who tested it need to look into buying lotto tickets.

I’ll dust it off once again, just because the Meta is the worst it’s ever been anyways, might as well play bad cards…
And see if my “luck” changed…

Which reminds me a lot of the Frostmage third trait always hitting the first troop instead of a random one. The team also got asked about this again and again, insisting every time that this is not the case, despite multiple video recordings proving them wrong. Until it finally got fixed about half a year later.

*Sigh* Sorry for feeling passionate about this. Let’s wait and see what percentage chance @Venar gets repeating those tests. Somehow I can’t shake the feeling the team hasn’t really understood the issue being reported.

5 Likes

thanks kafka for the response… I will have to try it out again. I have not used it since my last post, but if nothing has been altered, I must say that I have not once seen it work. not even once. I understand randomness. I even have a passing knowledge of rainbows and unicorns (I have little girls after all). I just cannot fathom that the code is working as developed when I can pay careful attention and not witness it happen even once hours into playing with it. That is the equivalent to hitting heads on a coin flip 20 times in a row (which is more difficult to do than to win the lottery).

I’ve cut a head or two this week. Nothing close to 20%.
I wonder if it works for the AI…
But even putting it on defense, i won’t know.

I suspect the team test was based on the number of troops killed by the spell when casting it at a troop, not the number of troops killed by the spell when casting it at a full health troop. The former approach will show roughly a 20% mortality rate, mostly due to the damage the weapon deals, not due to the instant kill effect.

Could be. A good « non-scientific » test is to use a Team with a Firebomb, and a hero class with the rising shadow trait (20% summon chance when you kill an opponent).

After summoning a dozen Sister of Shadow, but no Vorpal despite numerous casts, you see that something is off.

Of course, with the current meta, such a team will get massacred, so I suggest exploring on a high difficulty level.

You can always do trial battles against a team of your own design. Just set it as your PVP team and test it until you’re satisfied.

I found that the AI acts differently when you test defenses.

I doubt anyone has ever Vorpaled anyone. Sure, they die… X amount of damage at a time (where X is the damage normally causes), but I have yet to see an enemy die from Vorpal. Ever. It is broken.

It has triggered for me before. I killed two troops in one cast. But it took me 60-something casts before the supposedly 20% chance occurred. That is why I was hoping Kafka or someone would explain how Vorpal is intended to work. If the code says 20%, then there must be some additional condition preventing it from happening 20% of the time.

Same here. I can’t say it never happens.