I just tested confuse ray rate in 70 itterations: Hit Itself 60% of the time.

[FNTCZ]

Alright so I had the feeling that confusion definitely did not feel like a 30% so after seeing @Darkshade's response on the other thread I said fuck it and went on and test it myself. with the help of @pachima and @ozzie1550 I wasted 45 minutes of my life spaming confuse ray and rapid spin. I kept track of the results in a spreadsheed because I definitely do not have to study for a final this friday; anyways,

 

I did this 70 times with the following results: 42 times my pokemon hit himself in confusion and only 28 times the attack went through, giving an experimental confusion hut rate of 60%

 

now I am no mathemathician but I think this does mark a trend, and I have heard from multiple people too that confusion rate doesnt feel nowhere near that 30% it supposedly is.

https://en.wikipedia.org/wiki/Birthday_problem

 

 

 

Spreadsheet:

Spoiler

 

 

 

Edited November 29, 2017 by FNTCZ

[Spaintakula]

70 times of the event happening can't really prove anything though, unless the numbers are really onesided?

Try passing atleast 3-4 more the times you'd actually be able to get an actual % of how often confusion hits.

So like, 400 confusion turns minimum let's say.

Ideally, you really want to have it tested atleast enough where you can post a percentage that's above the 100 mark.

And do keep in mind it doesn't have to be 33% exactly for it to be proven you're wrong. It can be 40% of the times you hit yourself on a base of 100 confusion hits, and you'd still be wrong because of insufficient data to actually calculate it over and over again.

And nice spreadsheet lol

[FNTCZ]

8 minutes ago, Spaintakula said:

70 times of the event happening can't really prove anything though, unless the numbers are really onesided?

Try passing atleast 3-4 more the times you'd actually be able to get an actual % of how often confusion hits.

So like, 400 confusion turns minimum let's say.

Ideally, you really want to have it tested atleast enough where you can post a percentage that's above the 100 mark.

And do keep in mind it doesn't have to be 33% exactly for it to be proven you're wrong. It can be 40% of the times you hit yourself on a base of 100 confusion hits, and you'd still be wrong because of insufficient data to actually calculate it over and over again.

And nice spreadsheet lol

I think you got it the other way around. Confusion is supposed to be 30% yet the test showed 60% thats an error margin of 100% (2x)  which is fucking huge even for a sample of 70 itterations.

 

I did want to go for 100 tests but eh, I got really tired and the result seemed pretty clear: It was nowhere close to the value provided by the devs.

Edited November 29, 2017 by FNTCZ

[Spaintakula]

Just now, FNTCZ said:

I think you got it the other way around. Confusion is supposed to be 30% yet the test showed 60% thats a difference of 100% (2x)  which is huge for a sample of 70 itterations.

 

I did want to go for 100 test but eh, I got really tired and the result seemed pretty clear: It was nowhere close to 30% (The value devs provided).

I know confusion is supposed to be 30, I'm well aware of pokemon's statistics, was 50%, now it's 33% since new gen mechanics.

And that tiredness is probably what's missing from you getting a bit closer of a answer than 60%. Percentage on that small of a scale is not huge at all. I could've done 20 confusion hits lets say and only gotten hit by 2, or gotten hit by 15, it's a small scale to really define it's 33%  or 50% or 100%. Plus, I really don't see why they'd really change the equation of confusion hitting back to 50 or whatever you think it is they changed it to.

[FNTCZ]

18 minutes ago, Spaintakula said:

I know confusion is supposed to be 30, I'm well aware of pokemon's statistics, was 50%, now it's 33% since new gen mechanics.

And that tiredness is probably what's missing from you getting a bit closer of a answer than 60%. Percentage on that small of a scale is not huge at all. I could've done 20 confusion hits lets say and only gotten hit by 2, or gotten hit by 15, it's a small scale to really define it's 33%  or 50% or 100%. Plus, I really don't see why they'd really change the equation of confusion hitting back to 50 or whatever you think it is they changed it to.

I think it is just a bug.

 

a sample of 70 events is enough given that we are dealing with a flat percentage instead of some complex variable like the shiny rate would be.

 

For instance, I just coded a small simulation that throws a dice 70 times and counts how many times the number 1 rolls. Given that there is 1/6 (17%) chance of this event occuring, across a 70 repetitions the aproximate value that should be obtain would be 11.6. I ran this small simulation 10 times and got the following values: 16, 10, 11, 11, 12, 11, 12, 8, 10, 12. yeah, 16 and 8 do seem off but they represent 22% and 12% instead of 17%. that is nowhere as big of a difference as the one between 30% and 60%. Also, the other 8 times the simulation was ran the value obtained was completely spot-on. the margin of error should never be near the 100% obtained in the confusion tests, because if you roll a dice 70 times and get the number one 24 times please go to a casino or buy a lotto ticket.

 

What I am trying to say is: for this case, the sample size of 70 is fair and enough, specially considering the drastic difference obtained between the tested value(60%) and the theoretical value (30%) that suggests that something is indeed off.

Edited November 29, 2017 by FNTCZ

[XelaKebert]

1) Your sample size is way too small. In order to see the percentages correctly you would have to repeat this test thousands of times over. Just because you have a static percentage that does not excuse poor testing.

 

2) This is a broad generalization. You are taking your own small sample and applying the results to the whole.

 

For example, the odds of rolling any given number on a 6 sided die is 1 in 6. You can roll and get vastly different results in one study of the die. Repeating that test over and over you will start to see the numbers point to roughly a 1 in 6 odds for rolling any number on a single die.

 

[Laz]

I don't agree with his testing, I just think he just wants to bring this to your attention so you can double check if the move works properly, and I agree with this. Due to some moves being broken previously, you never know. It doesn't hurt to double check right.

[FNTCZ]

1 minute ago, Lazaro23 said:

I don't agree with his testing, I just think he just wants to bring this to your attention so you can double check if the move works properly, and I agree with this. Due to some moves being broken previously, you never know. It doesn't hurt to double check right.

Yeah this please.

 

Also, 

5 minutes ago, XelaKebert said:

1) Your sample size is way too small. In order to see the percentages correctly you would have to repeat this test thousands of times over. Just because you have a static percentage that does not excuse poor testing.

 

2) This is a broad generalization. You are taking your own small sample and applying the results to the whole.

 

For example, the odds of rolling any given number on a 6 sided die is 1 in 6. You can roll and get vastly different results in one study of the die. Repeating that test over and over you will start to see the numbers point to roughly a 1 in 6 odds for rolling any number on a single die.

 

using only the number one as reference (rolling the dice 70 times). after 100 simulations an error of 100% was only obtained, on average, after the 58th simulation. I mean that if the confusion rate is indeed 30% then the probability of my excersise dropping the results it dropped was 1/58.

 

Of course I would've loved to do 100 or 200 confusion hits but I do not have that kind of time. It has been multiple people who have brought in multuple instances the fact that confusion does not seem to be 30%, not even close.

 

Could you please check? perhaps the next afternoon that I am free I will try doing this 200 times because I am pretty convinced that something is indeed wrong.

[XelaKebert]

10 minutes ago, FNTCZ said:

Yeah this please.

 

Also, 

using only the number one as reference (rolling the dice 70 times). after 100 simulations an error of 100% was only obtained, on average, after the 58th simulation. I mean that if the confusion rate is indeed 30% then the probability of my excersise dropping the results it dropped was 1/58.

 

Of course I would've loved to do 100 or 200 confusion hits but I do not have that kind of time. It has been multiple people who have brought in multuple instances the fact that confusion does not seem to be 30%, not even close.

 

Could you please check? perhaps the next afternoon that I am free I will try doing this 200 times because I am pretty convinced that something is indeed wrong.

As with any percentage given that is per roll. Each roll is a separate roll.

[FNTCZ]

28 minutes ago, XelaKebert said:

For example, the odds of rolling any given number on a 6 sided die is 1 in 6. You can roll and get vastly different results in one study of the die. Repeating that test over and over you will start to see the numbers point to roughly a 1 in 6 odds for rolling any number on a single die.

This is exactly the example I wrote to test the odds of obtaining the results I did.

 

I rolled a dice 70 times, counted how many times I rolled the number one, the theoretical value should be 11.6, each simulation had its own margin of error ofcourse. Then I ran that simulation thousands of times to see, on average, how many attempts would it take to obtain an error of 100% (such was the one in the confusion tests). The result obtained was ~57. That means that >2% of the tests made on a sample sized of 70 itterations would result in an error of 100%. so either my test was that 1/57 or there is something actually off..

 

Spoiler

Public times As Integer
Public bol As Boolean

 

'rolls the dice 70 times and counts how many times the number 1 is obtained. the theoretical value should be 11.6 times per simulation
Sub test()
    y = 0
For i = 1 To 70
    x = WorksheetFunction.RandBetween(1, 6)
    
    If x = 1 Then
        y = y + 1
    End If
Next i

If y >= 24 Or y <= 5 Then
    bol = False
End If

End Sub

 

'runs the dice simulation untill an error of 100% is obtained, saves the itteration on which this error was obtained
Sub test2()
bol = True
times = 0
While bol = True
    Call test
    times = times + 1
Wend
   ' MsgBox (times)
End Sub

 

'obtains an error of 100% 100 times. then averages how many tries it took to get such a giant error.
Sub finalfinal()
    For i = 1 To 200
        Call test2
        Count = Count + times
    Next i
    
    MsgBox (Count / 200)
End Sub
 

 

Edited November 29, 2017 by FNTCZ

[Kyu]

We've confirmed that Confusion is proccing more often than is intended. It will be fixed in the next update. The following moves have been disabled and banned in PvP until the next update, which should be out later this week:

• Chatter
• Confuse Ray
• Confusion
• Dizzy Punch
• Dynamic Punch
• Flatter
• Hurricane
• Psybeam
• Rock Climb
• Signal Beam
• Supersonic
• Swagger
• Sweet Kiss
• Teeter Dance
• Water Pulse

 

In the future, please report suspected exploits to the Bug Reports section, email [email protected], or PM a Developer directly instead of posting it publicly. Thank you.