cast on hit gem trigger chance

after cast on hit on other sources than attack got fixed i started taking notes on my creatures attacks as before already on hit gems seemed to trigger rarer than they should (which i kinda discounted when the gems could trigger with anything).
probably most of the notes are from 0.4.1 still, so if you changed anything with trigger chances till 0.4.3 just ignore this.

so here are the attacks (not counting any attacks that got dodge or 0 or killing the last enemy in the battle):

6 spell gems (14%, 13%, 11%, 10%, 10%, 10% together a chance to trigger something on attack of 51%[all chances i mention here are always rounded down])
47 attacks 17 triggering a spell 30 no trigger at all
6 spell gems (14%, 12%, 11%, 11% 11%, 11% total chance to trigger something 52%)
45 attacks 17 triggering a spell 28 no trigger at all
6 spell gems (17%, 15%, 14%, 12%, 11%, 10% total chance to trigger something 57%)
45 attacks 19 triggering something 26 no trigger at all
6 spell gems (18%, 14%, 14%, 12%, 11%, 10% total chance to trigger something 57%)
73 attacks 31 triggering a spell 42 no trigger at all
5 spell gems (14%, 14%, 11%, 11%, 10% total chance to trigger something 47%)
140 attacks 35 triggering a spell 105 no trigger at all
6 spell gems (13%, 12%, 12%, 11%, 10%, 9% for a 50% chance to trigegr soemthing + burst of power trait 20%, 20%, 20%, 20%, 20%, 20% for a grand total chance of 87% to trigger something)
180 attacks 137 triggering a spell 43 no trigger at all

yea, i know the sample size could be bigger but i think theres a clear trend here it seems especially weird to me that not a single of my creatures is running overaverage on triggering spells. the 47% chance seems good to look at here to me 140 attacks arent that few and that outcome is either ridiculously unlucky on my part or something is weird with the on hit trigger chance.
please take a look at the spell triggers (ill keep taking notes and update this later on in case thats not enough data for you to warrant looking into this).

I just did a few tests and it seems like everything is working correctly. I’ll take a closer look at this later on.

a quick update: im quite certain by now something is broken with the cast on hit as i have been keeping count and by now it just almost never triggers anymore
(i even stopped counting the nontriggers for a while because i thought all on hit except burst of power trait didnt work anymore until i saw a normal spell gem trigger again).adding the new attacks i noted down to a new total (same spell setup as in the original post)
(51%)
110 attacks 18 triggering a spell 92 no trigger
(52%)
95 attacks 19 triggering a spell 76 no trigger
(57%)
64 attacks 24 triggering a spell 40 no trigger
(57%)
136 attacks 34 triggering a spell 102 no trigger
(47%)
277 attacks 41 triggering a spell 236 no trigger
(87%)
317 attacks 244 triggering a spell 73 no trigger

not even one of those is close to where it should be and again not a single one has an overaverage amount of triggers just only less than it should everywhere.

I’ve done a lot of testing and it all looks good to me. If you attach your save file I could take a look at your specific setup to make sure it’s not just bad luck.

6 spell gems (14%, 13%, 11%, 10%, 10%, 10% together a chance to trigger something on attack of 51%[all chances i mention here are always rounded down])

I think your flaw is that you are adding these together.
It’s 14% independent of 13%, independent of 11%… etc, etc. And sometimes, you’ll get two to trigger (or more).

As an average, 51% is a good barometer, but you must take into account double/triple triggers. So if you get a double/triple/quad trigger, you might not see another trigger for 4 more attacks (or more!).

My best advice is to use things like Multihit, Quadhit, Pummel, or the Chaos Perk to really beef up your return on on-hit gems. At even 14%, it’s very high likelihood of NOT occurring.

[quote=“Psylisa, post:5, topic:2053”]6 spell gems (14%, 13%, 11%, 10%, 10%, 10% together a chance to trigger something on attack of 51%[all chances i mention here are always rounded down])

I think your flaw is that you are adding these together.
It’s 14% independent of 13%, independent of 11%… etc, etc. And sometimes, you’ll get two to trigger (or more).

As an average, 51% is a good barometer, but you must take into account double/triple triggers. So if you get a double/triple/quad trigger, you might not see another trigger for 4 more attacks (or more!).

My best advice is to use things like Multihit, Quadhit, Pummel, or the Chaos Perk to really beef up your return on on-hit gems. At even 14%, it’s very high likelihood of NOT occurring.[/quote]

sure im adding them and i dont see a problem with that, do you? its just percentage math what you posted there its 0.860.870.890.90.9*0.9=0.485 which means 48,5% chance to not trigger any spell effect, so 51% chance to trigger at leats one the thing you are saying about multiple triggers is just compelte nonsense please check your probability math. (like you say if i flip a coin twice and the first flip is heads the second flip would be more likely to be tails which is just not true)

also attached the save for when u have the time to check on it zack.


slot5.sav (512 KB)

[quote=“Kevin Sith, post:6, topic:2053”]sure im adding them and i dont see a problem with that, do you? its just percentage math what you posted there its 0.860.870.890.90.9*0.9=0.485 which means 48,5% chance to not trigger any spell effect, so 51% chance to trigger at leats one the thing you are saying about multiple triggers is just compelte nonsense please check your probability math.

also attached the save for when u have the time to check on it zack.[/quote]

It’s not nonsense. They are 100% independent chances to happen, and they have no memory of the former result. Additionally, they can happen in multiples, such that Gem B can trigger with or without influence from the result of Gem A.

You’re counting triggers as in “something happened”, but the game is counting triggers as in “this happened X number of times”. So yes, the fact that there are double/triple/quad procs on them does indeed matter.

47 attacks 17 triggering a spell 30 no trigger at all
If out of those 17, 5 were double procs, 3 were triple, and 1 was quad, you'd be much higher on your distribution.

You’d have 17 single + 5 doubles ( 10 ) + 3 triples ( 9 ) + 1 quad ( 4 ) = 40 / 47

If you remove the initial 1 proc (since I’m not sure how you are recording your data), you might have this
8 single ( 8 ) + 5 double ( 10 ) + 3 triple ( 9 ) + 1 quad ( 4 ) = 31 / 47

That’s just an example though. I’m not sure what your actual data is, but the fact that each gem can trigger independently (and multiple gems can activate on the same attack) makes all the difference in the world.

That 51% average reflects that, as one attack can have 6 positive outcomes and thus 6 others have no positive outcome - on a long enough timeline.

Would everyone just please look at the original poster’s math?

6 spell gems (14%, 13%, 11%, 10%, 10%, 10% together a chance to trigger something on attack of 51%[all chances i mention here are always rounded down])

Assuming you have 6 spell gems equipped to a single creature, the chances of NONE of these triggering would be (1-14%)(1-13%)(1-11%)(1-10%)(1-10%)*(1-10%), or ~49%. Therefore, the chance of ONE OR MORE gems triggering is ~51%. If however you observe 30/47 attacks with no event, even though you should only have ~23 attacks without an event, odds are something is amiss with either the random number generator (I worked with game maker and I know it is lacking in certain areas, not sure if this is one as well^^) or someplace in the game engine itself. Keep in mind that this deviation is neither slight nor the exception - all of the original poster’s creatures have uneventful attacks way more often than they should.

That sample size is WAY to small to make any sort of conclusion, thats an avarage of 51%, do at least 1000 hits, then you can START calculating.

look at my first post updating the numbers.
whats it with people responding if they clearly havent even read the thread?
but really thanks for that contribution its immensly helpful…

180 is the biggest sample size you have, even adding up all of them, its still way to small of a sample size to draw any conclusions about random distribution.

Test the null hypothesis and see how unlikely it is even with this sample size. In genetics for example it is not uncommon to have sample sizes <100 and still reach significant conclusions. Even the fact that 6 creatures EACH underperform is already like 1/2^6 since we must assume that any randomness would follow some kind of normal distribution.

However, computer random generators are known for their streakyness, especially the lower quality ones, like the one included in game maker. Still needs more sample size to verify. Doing a small sample size myself, just to see. To small for conclusions, but will let you know anyway.

the first post updating the numbers which doesnt mean the opening post but whatever the biggest single size there is 317 and 999 attacks alltogether.

by now i even stopped noting because its quite obviously something wrong but sure i have some more counts now already and can easily resume.
it feels like the chance to trigegr gets worse the longer i play, as im now in artifact realm 42 without leaving and even the guy with 87% chance to trigger
only triggers like every 3rd attack atm. (yes these are guesstimates i am aware ill resume noting it downbut really its already ridiculous)

Assuming you have 6 spell gems equipped to a single creature, the chances of NONE of these triggering would be (1-14%)*(1-13%)*(1-11%)*(1-10%)*(1-10%)*(1-10%), or ~49%. Therefore, the chance of ONE OR MORE gems triggering is ~51%.

Except… it’s not 51%. Each spell gem has a 10-14% chance of triggering independently.
When you flip a coin, the chances are 50/50 of heads or tails (we’ll ignore magical edge lands).
When you flip it again, it’s still 50/50. The odds don’t change based on what you did earlier. Or what you’ll do next.

The same is true for the spell gems.
The first has 14% activation. That’s 86% of nothing happening.
The second has 13% activation. That’s 87% of nothing happening - REGARDLESS OF WHAT THE FIRST OUTCOME WAS.

That’s the main issue; and why the 51% (average) isn’t a good barometer, unless you’re also taking into account the double/triple/quad/etc procs since it’s independent chance. When you do, you’ll get much closer to the 51% distribution.

[quote=“Kevin Sith, post:10, topic:2053”]

look at my first post updating the numbers.
whats it with people responding if they clearly havent even read the thread?
but really thanks for that contribution its immensly helpful…[/quote]

Your math checks out so my guess is that the “attack hit” from some certain sources (spells, traits, or some combinations of them) didn’t check the spellgem procs.
Your sample size is admittedly small but significant enough.
I’m not 100% sure how the spell procs work when you stack many spells with attacks from spells and traits.
I noticed in some occasions the spells will not proc when the target is dead so this will greatly decrease the chance of “something happens”.

If you would be a bro to Zack, help him by testing it with without any spell. Equip all your 6 gems and do many single-normal attacks. See if the math checks out. My guess is it will.
Next is to test each spell/trait without stacking them all together so you know for sure what went wrong.
Pay attention to how many attacks you hit the target, and whether the target is dead or not. Because that will affect your sample size.

Your post is very long so it is easy to be misunderstood. Just let it go man. Lets keep the board friendly here.

Sorry missed that, thought you’d modify the OP.
Well yeah, thats enough sample size to maybe start looking into it. What i am noticing however, is that on hit artifact seems to trigger way more consistently than on hit gems.

@Psylisa: Kevin’s math is correct he just shortened it a bit.

The chance of a proc is 1 - (chance of no proc).
Chance of no proc is independent which means for the gems 14%, 13%, 11%, 10%, 10%, 10% you get

0.860.870.890.90.9*0.9 = 48.5%

Chance for at least once proc is 1 - 48.5 = 51.5%.

However if the “one proc” thats counted actually was a triple, that should be counted as 3 procs. Which does change the resulting distribution. Maybe the RNG is extremly streaky, the distribution still matters.

[quote=“Kejal, post:18, topic:2053”]@Psylisa: Kevin’s math is correct he just shortened it a bit.

The chance of a proc is 1 - (chance of no proc).
Chance of no proc is independent which means for the gems 14%, 13%, 11%, 10%, 10%, 10% you get

0.860.870.890.90.9*0.9 = 48.5%

Chance for at least once proc is 1 - 48.5 = 51.5%.[/quote]

Yes, and I get that.

But WITHIN that 51.5% are double/triple/quad procs that aren’t accounted for - which is my point. You MUST account for those.
So when you make 1 attack and get all 6 procs, your ratio is 6:1
And when you make 11 more attacks that result in nothing, your ratio is 6:12 (thus getting you 51%)

Which explains why the OP is seeing what he sees. Each spell gem has a 10-14% chance to proc independent of each other. On a long enough timeline, he should average of about 51% proc chance. But this doesn’t mean he has a 51% chance per attack to see 1 gem activate.