Pokémon Catch Rate Formula Conflict?

[Hubologist]

It seems that Serebii and Bulbapedia have released two different formulas for calculating the catch rate of Pokémon, and they seem to give different results given the same ball, the same Pokémon, the same amount of current HP, and the same status.

Here are the links: Bulbapedia Formula | Serebii Formula

I used my own level 50 Uxie to test each formula, assuming I'm using a Dusk Ball inside a cave, that I've put Uxie to Sleep, and also assuming that I've gotten Uxie's HP down to 1 using False Swipe. My particular Uxie has a max HP of 137, and the catch rate for all Uxies is 3.

Here's what I came up with:

Bulbapedia Formula: 9.3% catch rate (I had to use the approximation formula to solve for p... I'm smart, but math isn't my strongest suit.)
Serebii Formula: 35.7% (or, more likely, 3.6%, but going by the instructions, it's 35.7%).

35.7% can't be right, because it's way too high. It took me forty Dusk Balls to catch Uxie (all of them while she was asleep, and at 1 HP), so 35.7% is out of the question.

So, which formula is right, and which is wrong? And how can EITHER of them be right when it took me forty tosses of the Dusk Ball to catch Uxie under those conditions? XD (Actually in that case, the corrected Serebii formula seems more accurate).

 

[Zhen Lin]

The Bulbapedia formula is taken from the linked site. 40 tries before success for an event that has a constant 9.3% probability isn't actually anomalous - if we were working at a 5% significance level with a two-tailed test, 40 tries is just within the 95% confidence interval.

You'd have to conduct a lot more experiments before having any reasonable data, however. On average it should take just 11 or so tries - assuming all conditions are constant.

 

[Hubologist]

The Bulbapedia formula is taken from the linked site. 40 tries before success for an event that has a constant 9.3% probability isn't actually anomalous - if we were working at a 5% significance level with a two-tailed test, 40 tries is just within the 95% confidence interval.

You'd have to conduct a lot more experiments before having any reasonable data, however. On average it should take just 11 or so tries - assuming all conditions are constant.

Oh, I know that. Forty balls easily allows for a ~10% overall probability; on another occasion, I might have gotten a catch on the first or second toss. I'm mostly just bemoaning my forty-toss catch. ;-)

What do you mean when you say that the Bulbapedia data is taken from the linked site? Do you mean it comes from the Serebii page? My primary question was which one of the two formulas are correct. I lean towards the Bulbapedia one, since it seems more complete.

 

[Zhen Lin]

Go to the bottom of the page, you will see a Japanese link. It comes from there.