Shotguns Huge Nerf

[nslay]

IMO shotguns worth looking at in RRR are going to have decent crit and status (excluding beam shotguns, I don't understand these yet):

NOTE: Build these as hybrids! (e.g. Viral/Crit, like in the video):

Kuva Kohm
Vaykor Hek
Strun Wraith
Kuva Drakgoon

Borderline (low-ish crit!):
Sancti Tigris
Boar Prime
 

These exclude Astila and Arca Plasmor (1 pellet shotguns).

Lastly, here's hybrid Vaykor Hek and hybrid Kuva Kohm downing a level 140 Exo Gokstad Officer (highest armored enemy in the game):

EDIT: P.S. I use Kuva Chakkhurr as a reference standard (to show how hard Gokstad Officer is to kill!).
 

Spoiler

 

 

Edited March 6, 2020 by nslay

[Kaotyke]

Just now, Skaleek said:

 

Take 5.

You are reminding me of someone.

3 minutes ago, Skaleek said:

I didn't see you post an opinion other then: "because you own a riven you don't have any right to discuss a weapon's viability". Your opinion is literally that other people's opinions don't matter because what, they had a riven for the weapon?

Did I mention that their opinion doesnt matter because they have a Riven? I mentioned that Rivens shouldnt be considered for balance. But it seems you are seeing that as "The person's opnion does not matter either".

Are you trying to start something here?

[Skaleek]

Just now, Kaotyke said:

Did I mention that their opinion doesnt matter because they have a Riven? I mentioned that Rivens shouldnt be considered for balance. But it seems you are seeing that as "The person's opnion does not matter either".

Are you trying to start something here?

No you make a flippant comment to me posting that I have a riven for the strun wraith but its bugged, i'd say you're the one "trying to start something".

Rivens aside, I tested all the shotguns. I find that even the nerfed ones still perform reasonably well. The strun wraith, to me, is an extreme outlier. It's really bad. I hope its because its bugged, but regardless, even if i have a riven for it, and it performs badly, and i say it performs badly... Its still viable weapon feedback, because the riven just ups its power level. Who cares about the riven?

[Kaotyke]

1 minute ago, Skaleek said:

No you make a flippant comment to me posting that I have a riven for the strun wraith but its bugged, i'd say you're the one "trying to start something".

And you are taking this like its a personal crusade. Roll your eyes and move on.

2 minutes ago, Skaleek said:

Rivens aside, I tested all the shotguns. I find that even the nerfed ones still perform reasonably well. The strun wraith, to me, is an extreme outlier. It's really bad. I hope its because its bugged, but regardless, even if i have a riven for it, and it performs badly, and i say it performs badly... Its still viable weapon feedback, because the riven just ups its power level. Who cares about the riven?

Again, if its bugged, I hope it gets fixed. If its really that bad, then we need throw feedback.

[Vahenir]

Don't forget the phantasma in this matter as well. Because that was nerfed into the ground last update. From like 38% status chance to 7% per pellet, even if its a beam gun. Meaning its basically dead right now with its 3% crit chance on top of that.

[nslay]

2 minutes ago, Vahenir said:

Don't forget the phantasma in this matter as well. Because that was nerfed into the ground last update. From like 38% status chance to 7% per pellet, even if its a beam gun. Meaning its basically dead right now with its 3% crit chance on top of that.

They messed up the status calculation... badly. They can't math! I filed a bug report. This is what they did:

per-pellet chance = S / N (this is wrong and it underestimates the true value)

where S is original per-shot status and N is number of pellets.

Then they tripled this value:

new per-pellet chance = 3 * S / N (this is wrong)

 

What it should be (probability theory!):

per-pellet chance = 1-(1-S)^(1/N)

Then you triple this value:

new per-pellet chance = 3*(1-(1-S)^(1/N))

 

The new per-pellet chance will increase with this calculation.

[Renegade343]

12 minutes ago, nslay said:

They messed up the status calculation... badly. They can't math! I filed a bug report. This is what they did:

per-pellet chance = S / N (this is wrong and it underestimates the true value)

where S is original per-shot status and N is number of pellets.

Then they tripled this value:

new per-pellet chance = 3 * S / N (this is wrong)

 

What it should be (probability theory!):

per-pellet chance = 1-(1-S)^(1/N)

Then you triple this value:

new per-pellet chance = 3*(1-(1-S)^(1/N))

 

The new per-pellet chance will increase with this calculation.

I mean, it depends on how you want to do it. The former isn't necessarily wrong, it just depends on how you define variables and values to be.

[Vahenir]

Indeed, only too bad that spraying status effects onto a target is no longer as useful unless you have a high crit weapon to actually exploit it as well. So even if the phantasma remained at the original 38% status chance it'd be heavily nerfed just because of that, since it also lacks critical chance and critical damage.

[nslay]

6 minutes ago, Renegade343 said:

I mean, it depends on how you want to do it. The former isn't necessarily wrong, it just depends on how you define variables and values to be.

No, the former really is actually wrong because it will not result in the per-shot status chance to be S. There is no different interpretation you can have. It's wrong.

Example (wrong):

Tigris Prime has a base per-shot status chance of 0.3 and 8 pellets.

Using the wrong value:

per-pellet status chance = 30% / 8 = 3.75%.

Using this value, what's the per-shot status chance? Well, none of the 8 pellets proc status is

probability no pellet procs status = (1-0.0375)^8

The probability at least one pellet procs status is the complement of this probability:

probability at least one pellet procs status = 1-(1-0.0375)^8 = 26%

This value should be 30%, hence, the per-pellet status chance of 3.75% is wrong.

 

Example (right)

per-pellet status chance = 1-(1-0.3)^(1/8) ~ 4.4%

Per-shot status chance = 1-(1-0.044)^8 ~ 30%

This is the same application of probability theory you use for flipping coins and rolling dice. If you did any of this the DE way, your predictions would be wrong.

[AbstractLemons]

3 hours ago, DrBorris said:

1. That is a chart I made before the patch dropped, some shotguns were given different status chances. Notably Kohm and Kuva Kohm were given a very generous buff.
2. With the status changes in general, just looking at the changes to status chance doesn't tell the whole story. Despite some shotguns having a lower chance they still perform quite similarly (Tigris series for example)
3. Most shotguns, even on that chart, saw a major buff to status chance. Everything that could not hit 100% before fared very well.
4. With the status changes dedicating four mods to status chance is not nearly as necessary. There is room for more raw damage mods or QoL.
5. Enemy EHP values were significantly reduced on the high end, where these status shotguns were most valued.

 

Edit: For any wondering, this is the full comment that chart is from.

 

So what you saying is that either OP doesn't know how to read or they are LYING? Makes sense.

[Renegade343]

4 minutes ago, nslay said:

No, the former really is actually wrong because it will not result in the per-shot status chance to be S. There is no different interpretation you can have. It's wrong.

Example (wrong):

Tigris Prime has a base per-shot status chance of 0.3 and 8 pellets.

Using the wrong value:

per-pellet status chance = 30% / 8 = 3.75%.

Using this value, what's the per-shot status chance? Well, none of the 8 pellets proc status is

probability no pellet procs status = (1-0.0375)^8

The probability at least one pellet procs status is the complement of this probability:

probability at least one pellet procs status = 1-(1-0.0375)^8 = 26%

This value should be 30%, hence, the per-pellet status chance of 3.75% is wrong.

 

Example (right)

per-pellet status chance = 1-(1-0.3)^(1/8) ~ 4.4%

Per-shot status chance = 1-(1-0.044)^8 ~ 30%

This is the same application of probability theory you use for flipping coins and rolling dice. If you did any of this the DE way, your predictions would be wrong.

I know what the expected calculation should be. I'm saying that I can create a new sequence that would deviate from the expected calculation and still be consistent with mathematics as we know it.

I can first declare "Let the value of the displayed original status chance be divided equally among all pellets, so that the summation of the status chance of each pellet be equal to the aforementioned value.", then work on from there to get the former result. Basically, I define the value of the status chance differently to what it's expected.

[nslay]

24 minutes ago, Renegade343 said:

I know what the expected calculation should be. I'm saying that I can create a new sequence that would deviate from the expected calculation and still be consistent with mathematics as we know it.

I can first declare "Let the value of the displayed original status chance be divided equally among all pellets, so that the summation of the status chance of each pellet be equal to the aforementioned value.", then work on from there to get the former result. Basically, I define the value of the status chance differently to what it's expected.

Look, if you accept that every pellet is an independent and identically distributed binary random variable for proc/no proc with probability for proc as p, then the various possible outcomes of proc/no proc follow the Binomial distribution which gives the values for per-pellet and per-shot status chances and does not work with your twisted interpretation.

You cannot accept that simple property and then try to play word games to justify DE's incorrect calculation. They did it wrong and it's a bug. It causes all of the shotguns in RRR to have lower statuses than they should. It also against what they claimed they would do... tripling per-pellet chances. They tripled a different value that underestimates the per-shot status.

[Renegade343]

Just now, nslay said:

Look, if you accept that every pellet is an independent and identically distributed binary random variable for proc/no proc with probability for proc as p, then the various possible outcomes of proc/no proc follow the Binomial distribution which gives the values for per-pellet and per-shot status chances and does not work with your twisted interpretation.

You cannot accept that simple property and then try to play word games to justify DE's incorrect calculation. They did it wrong and it's a bug. It causes all of the shotguns in RRR to have lower statuses than they should. It also against what they claimed they would do... tripling per-pellet chances. They tripled a different value that underestimates the per-shot status.

I'm not saying that I'm accepting what DE's doing. I'm saying that it's not necessarily incorrect because of the starting definitions. For example, in our everyday algebra system, we have a set of definitions such that dividing by 0 is undefined. But we can tweak that set of definitions, and make dividing by 0 have a value. And we do have such a system: https://en.wikipedia.org/wiki/Wheel_theory. Would you then say that wheel theory is incorrect because it works with a different set of definitions than what we typically use?

Saying that 'DE's way of doing the tweaked status calculations isn't exactly wrong because they started on an entirely different set of definitions' and 'DE's way of doing the tweaked status calculation is unintuitive and stupid' are two different things. I'm saying both.

[MHz3000]

The strun wraith saddens me.  It had a 40% status chance, but the good news?  I own 3 or 4 other shotguns and a slash viral kohmak.

[nslay]

5 minutes ago, Renegade343 said:

I'm not saying that I'm accepting what DE's doing. I'm saying that it's not necessarily incorrect because of the starting definitions. For example, in our everyday algebra system, we have a set of definitions such that dividing by 0 is undefined. But we can tweak that set of definitions, and make dividing by 0 have a value. And we do have such a system: https://en.wikipedia.org/wiki/Wheel_theory. Would you then say that wheel theory is incorrect because it works with a different set of definitions than what we typically use?

Saying that 'DE's way of doing the tweaked status calculations isn't exactly wrong because they started on an entirely different set of definitions' and 'DE's way of doing the tweaked status calculation is unintuitive and stupid' are two different things. I'm saying both.

OK, show me the distribution for pellet random variables that justifies DE's approach. Let's see if it makes any sense.

Because the sensible N p/not p experiments follows the Binomial distribution which governs all of these status chance values. It is a matter of fact... something even observable in reality! not just wishful thinking.

[Renegade343]

11 minutes ago, nslay said:

OK, show me the distribution for pellet random variables that justifies DE's approach. Let's see if it makes any sense.

Because the sensible N p/not p experiments follows the Binomial distribution which governs all of these status chance values. It is a matter of fact... something even observable in reality! not just wishful thinking.

And I'm saying that you can first say "Let the value of the displayed original status chance be divided equally among all pellets, so that the summation of the status chance of each pellet be equal to the aforementioned value.". The "Let" part of that sentence already implies that a new definition is made, and in this case, the value of the original status chance itself (i.e.: 20) is divided equally among the pellets (i.e.: 4), then that value is then made to be the status chance of a pellet.

Is it counter to what we normally do (binomial distribution calculations)? Yes. Is it wrong? Not really (in the sense that the application of mathematical systems used in that method isn't violated at least), just really, really weird and rare to do it in that manner. And what I'm saying isn't justifying DE, it's explaining the thought process of DE, because that's more interesting to figure out, wondering what system they're using. Goes in general for anything in life for me, really.

Edited March 6, 2020 by Renegade343

[nslay]

Just now, Renegade343 said:

And I'm saying that you can first say "Let the value of the displayed original status chance be divided equally among all pellets, so that the summation of the status chance of each pellet be equal to the aforementioned value.". The "Let" part of that sentence already implies that a new definition is made, and in this case, the value of the original status chance itself (i.e.: 20) is divided equally among the pellets (i.e.: 4), then that value is then made to be the status chance of a pellet.

Is it counter to what we normally do (binomial distribution calculations)? Yes. Is it wrong? Not really, just really, really weird and rare to do it in that manner. And what I'm saying isn't justifying DE, it's explaining the thought process of DE, because that's more interesting to figure out, wondering what system they're using. Goes in general for anything in life for me, really.

That doesn't tell me anything about how the pellet proc/no proc chances are distributed. The burden is on you to prove to me that DE did something reasonable.

Do the probability theory, give me the per-pellet probability mass function (whether it independent or not) that follows your vague description Let the value of the displayed original status chance be divided equally among all pellets, so that the summation of the status chance of each pellet be equal to the aforementioned value. And let's see if that makes any reasonable sense. I don't think the pmf is going to be very simple and I don't expect the pellets will be i.i.d. based on your vague description.

And yes, given reasonable expectations about how pellets proc status, the most simple/believable explanation is that DE made a mistake (a lot of posters also make the same mistake ... look at posts in the past couple says S/N written everywhere!). I doubt they formulated a brand new (and probably complicated!) pmf that allows them to simply divide by the number of pellets. I think they just forgot basic probability theory. It's a honest logic bug on their part.

[Renegade343]

Just now, nslay said:

That doesn't tell me anything about how the pellet proc/no proc chances are distributed. The burden is on you to prove to me that DE did something reasonable.

Do the probability theory, give me the per-pellet probability mass function (whether it independent or not) that follows your vague description Let the value of the displayed original status chance be divided equally among all pellets, so that the summation of the status chance of each pellet be equal to the aforementioned value. And let's see if that makes any reasonable sense. I don't think the pmf is going to be very simple and I don't expect the pellets will be i.i.d. based on your vague description.

And yes, given reasonable expectations about how pellets proc status, the most simple/believable explanation is that DE made a mistake (a lot of posters also make the same mistake ... look at posts in the past couple says S/N written everywhere!). I doubt they formulated a brand new (and probably complicated!) pmf that allows them to simply divide by the number of pellets. I think they just forgot basic probability theory. It's a honest logic bug on their part.

I have never argued about whether DE's method is reasonable. All I've said is "Based on this definition made by DE, the resultant calculation from the definition isn't wrong.", not "DE's definition is reasonable in the first place.". There's a difference between the two.

And what I said with the first step wouldn't work with probability theory, because it's not even using probability theory to start with. The first step is working with just the value, not the value and the % together.

[schilds]

Corrosive caps out at 10 procs now. It may be counterintuitive, but you typically won't want to trade off too much damage for higher status chance on weapons built for statuses that have a cap, and you won't want the weighting for those statuses to be too high. Exceptions will be when you're building to get the procs applied as fast as possible, or have another status like slash or toxin. That also means that weapons that trade off damage for higher status chance in their base stats will fall off in effectiveness if built only for status effects that cap out. All that extra status chance is doing nothing unless you're building for status effects that can use it.

Edited March 6, 2020 by schilds

[nslay]

17 minutes ago, Renegade343 said:

I have never argued about whether DE's method is reasonable. All I've said is "Based on this definition made by DE, the resultant calculation from the definition isn't wrong.", not "DE's definition is reasonable in the first place.". There's a difference between the two.

And what I said with the first step wouldn't work with probability theory, because it's not even using probability theory to start with. The first step is working with just the value, not the value and the % together.

If every pellet is an i.i.d. random variable with probability p to proc status chance (in the new UI, you are even explicitly given this value), then your description cannot work. If you accept this,  the binomial distribution dictates this:
S = 1 - (1-P)^N
P = 1 - (1-S)^(1/N)

where P is per-pellet chance, S is per-shot chance, N is number pellets. From this, it cannot happen N*P = S unless N = 1... You need N*P = S to satisfy your description.

So the only way for your abstract thinking to work is that DE derive a new pellet proc/no proc distribution so that N*P = S. I am asking for very basic information, what is the pmf for an individual pellet proccing status and does it make sense for players?

Your way of thinking is like "another possible explanation":

You wake up in the morning and you find a burned tree. There was a thunderstorm last night.

Simplest explanation: The tree was hit by lightning and caught fire.
Another possible explanation: An old Soviet satellite re-entering the atmosphere that was super-heated crashed into the tree and the tree caught fire.
Another possible explanation: Aliens decided to shoot the tree with their space ship.

I mean, these are all possible explanation, just like DE could have derived a brand new pmf that deviates with how players reasonably understand pellets proccing status so that they could redistribute status with arithmetic division. But it's far more likely DE made a mistake because they forgot probability theory... just like all these forum posters make the same mistake.

EDIT: P.S. And in the most likely event that they did make mistake, then by the definition of mistake, they did the calculation wrong. It could happen that there could have been a hypothetical DE that made a brand new pmf for pellets proccing status so that N*P = S. That doesn't make the real DE here and now any less wrong... doing P = S/N for pellet procs that follow the Binomial distribution B(N,P) is in fact wrong.

 

Edited March 6, 2020 by nslay

[Renegade343]

3 minutes ago, nslay said:

If every pellet is an i.i.d. random variable with probability p to proc status chance (in the new UI, you are even explicitly given this value), then your description cannot work. If you accept this,  the binomial distribution dictates this:
S = 1 - (1-P)^N
P = 1 - (1-S)^(1/N)

where P is per-pellet chance, S is per-shot chance, N is number pellets. From this, it cannot happen N*P = S unless N = 1... You need N*P = S to satisfy your description.

So the only way for your abstract thinking to work is that DE derive a new pellet proc/no proc distribution so that N*P = S. I am asking for very basic information, what is the pmf for an individual pellet proccing status and does it make sense for players?

Your way of thinking is like "another possible explanation":

You wake up in the morning and you find a burned tree. There was a thunderstorm last night.

Simplest explanation: The tree was hit by lightning and caught fire.
Another possible explanation: An old Soviet satellite re-entering the atmosphere that was super-heated crashed into the tree and the tree caught fire.
Another possible explanation: Aliens decided to shoot the tree with their space ship.

I mean, these are all possible explanation, just like DE could have derived a brand new pmf that deviates with how players reasonably understand pellets proccing status so that they could redistribute status with arithmetic division. But it's far more likely DE made a mistake because they forgot probability theory... just like all these forum posters make the same mistake.

 

I feel like you keep thinking I believe DE's definition is correct, and keep showing me the correct method, when I already know that from the start (why do you think I'm able to separate the concepts that clearly?). I made my explanation because I enjoy creating new math explanations for things, even if it's unintuitive or weird.

[nslay]

21 minutes ago, Renegade343 said:

I feel like you keep thinking I believe DE's definition is correct, and keep showing me the correct method, when I already know that from the start (why do you think I'm able to separate the concepts that clearly?). I made my explanation because I enjoy creating new math explanations for things, even if it's unintuitive or weird.

I feel like you're trying to persuade me that DE was maybe possibly feasibly probably using a different distribution for pellet procs in the old system that would let them correctly distribute pellet status like P = S/N (because, for sure, just from the new UI, it's a binomial distribution now).

I like creating new math too. I don't believe DE did that here... it's possible, but I think it's akin to one of those very unlikely scenarios I listed above. And I think that for multiple reasons:
1) It's easy to programmatically sample binomial distribution. Just sample x ~ U[0,1], if x < p : status proc, x >= p : no status proc (do this N times).
2) It's easy to do statistics with the binomial distribution. It's well documented and understood. This property is good for game developers that need to do some math for balancing.
3) The pellets proccing status individually in the same way as an unfair coin flip is both simple and intuitive for both players and the developers alike. It behaves very similarly to other games where you roll multiple dice (only your random variables are no longer binary).

I would very much like to see the PMF that gives P = S*N that is so easy and self-explanatory that it would be completely reasonable to expect DE to have concocted such an unusual distribution. I'd like to be proven wrong.

Edited March 6, 2020 by nslay
Got the conditions backward!

[Renegade343]

3 minutes ago, nslay said:

I feel like you're trying to persuade me that DE was maybe possibly feasibly probably using a different distribution for pellet procs in the old system that would let them correctly distribute pellet status like P = S/N (because, for sure, just from the new UI, it's a binomial distribution now).

I like creating new math too. I don't believe DE did that here... it's possible, but I think it's akin to one of those very unlikely scenarios I listed above. And I think that for multiple reasons:
1) It's easy to programmatically sample binomial distribution. Just sample x ~ U[0,1], if x < p : status proc, x >= p : no proc status (do this N times).
2) It's easy to do statistics with the binomial distribution. It's well documented and understood. This property is good for game developers that need to do some math for balancing.
3) The pellets proccing status individually in the same way as an unfair coin flip is both simple and intuitive for both players and the developers alike. It behaves very similarly to other games where you roll multiple dice (only your random variables are no longer binary).

I would very much like to see the PMF that gives P = S*N that is so easy and self-explanatory that it would be completely reasonable to expect DE to have concocted such an unusual distribution. I'd like to be proven wrong.

I was never trying to persuade you. Just explaining that this is the definition that DE's using, and under said definition, the calculations make sense. I never said the definition itself made sense, or is correct.

Edited March 6, 2020 by Renegade343

[nslay]

Just now, Renegade343 said:

I was never trying to persuade you. Just explaining that this is the definition that DE's using, and under said definition, the calculations make sense.

P = S / N does not make sense and is almost certainly an error.

[Renegade343]

2 minutes ago, nslay said:

P = S / N does not make sense and is almost certainly an error.

Oh, by making sense, I mean the steps that come that definition do follow each other, and it's at least valid (under the logical sense).