Displayed mod fusing success rates are grossly wrong.

squirrelloid

Suspicious that i was failing far more often than i should be, i decided to keep track of my recent batch compression of R4 mods. The character doing the fusing had 400 mysticism and used 7% catalysts throughout, for a nominal success rate of 80%.

Actual success rated differed significantly. To achieve 71 successful fusings, 164 fusing attempts had to be made. That's an actual success rate of ~43%.

Significance was tested with a Chi-squared test. Chi-squared works by summing (observed-expected)^2/expected. This is a "two box" test with one degree of freedom. Observed successes were 71, and observed failures were 93. Expected successes in 164 attempts are .8*164 = 131.2, and expected failures are 32.8. This produces a Chi-squared value of 110.489.... I couldn't actually find a program which would tell me just how significant that was quickly, but https://www.swogstat.org/stat/public/chisq_calculator.htm did let me figure out that the 1 degree of freedom critical value for p < 0.000000001 (!!!) is only ~36. (At which point it refused to calculate anything higher). That's a 1 in a billion event for a chi-squared value less than 1/3rd our actual chi-squared!

The only sane conclusion is to reject the null hypothesis that the given success rates are accurate. The observed behavior is so ridiculously unlikely if they are accurate that there probably haven't been enough mods dropped *in the lifetime of the game* to make such an event believable even if the mods had all been fused.

roughbearmattach

I must concur. Even a recently PTS update mentioned that the Devs will be looking at having the displayed success rate match the actual rate.

Here is an even better test: Find a character whose percent change at fusion is less than 100% at say, R2 mods, but would be 100% with catalysts. Then begin fusing. If any of the fusions are failures, we will have definite proof of the fusion rate being incorrect.

pantagruel01

squirrelloid said:

Actual success rated differed significantly. To achieve 71 successful fusings, 164 fusing attempts had to be made. That's an actual success rate of ~43%.

Which means you actually used 448 mods, which at 0% failure rate would produce 89.6 fusings; multiply that by 0.8 and you get 71.68. Which means the displayed rate is accurate for the percentage of mods expected to become components of the next category.

Which is... not at all what one would expect. I wonder if that's a coincidence or it's accurate for higher rank mods.

squirrelloid

pantagruel01 said:

squirrelloid said:

Actual success rated differed significantly. To achieve 71 successful fusings, 164 fusing attempts had to be made. That's an actual success rate of ~43%.

Which means you actually used 448 mods, which at 0% failure rate would produce 89.6 fusings; multiply that by 0.8 and you get 71.68. Which means the displayed rate is accurate for the percentage of mods expected to become components of the next category.

Which is... not at all what one would expect. I wonder if that's a coincidence or it's accurate for higher rank mods.

That would be very strange. Certainly it's not what they're claiming when they say there's an 80% success rate - it's clearly a per fuse attempt rate, not a per mod consumed rate.

More testing is apparently warranted. Sadly, i don't save mods below R4 - it just isn't worth the space or effort. If anyone wants to dump half a thousand R1 or R2 mods on me, i'll do the test.

roughbearmattach said:

I must concur. Even a recently PTS update mentioned that the Devs will be looking at having the displayed success rate match the actual rate.

Am I reading you right? They're going to fix the *display* by making it match the rates, not fix the *rates* so they match the display? Because the displayed rates are already punitively low to the point of fusing not being worthwhile. (Heck, they could remove failure entirely and fusing still wouldn't be worthwhile - AH prices for a rank N mod are less than 5x that of a rank N-1 mod)

roughbearmattach

squirrelloid said:

pantagruel01 said:

squirrelloid said:

Actual success rated differed significantly. To achieve 71 successful fusings, 164 fusing attempts had to be made. That's an actual success rate of ~43%.

Which means you actually used 448 mods, which at 0% failure rate would produce 89.6 fusings; multiply that by 0.8 and you get 71.68. Which means the displayed rate is accurate for the percentage of mods expected to become components of the next category.

Which is... not at all what one would expect. I wonder if that's a coincidence or it's accurate for higher rank mods.

That would be very strange. Certainly it's not what they're claiming when they say there's an 80% success rate - it's clearly a per fuse attempt rate, not a per mod consumed rate.

More testing is apparently warranted. Sadly, i don't save mods below R4 - it just isn't worth the space or effort. If anyone wants to dump half a thousand R1 or R2 mods on me, i'll do the test.

roughbearmattach said:

I must concur. Even a recently PTS update mentioned that the Devs will be looking at having the displayed success rate match the actual rate.

Am I reading you right? They're going to fix the *display* by making it match the rates, not fix the *rates* so they match the display? Because the displayed rates are already punitively low to the point of fusing not being worthwhile. (Heck, they could remove failure entirely and fusing still wouldn't be worthwhile - AH prices for a rank N mod are less than 5x that of a rank N-1 mod)

I think the devs meant that they will make it so the displayed rate is correct--that 80% truly is 80%.

We'll see.

pantagruel01

Okay, testing 100 R7 mods, nominal fuse chance 35%. Result was 16 R8s with 1 left over, so 19 were lost, meaning a rate of 16/35, or 45%. It appears my observation before was as coincidence. In fact, it's consistent with the thesis that success chance is fixed, since it's actually a better rate than R4s using catalysts.

chaelk

as far as I know, success rate is per fusing.
same as RNG rewards, some people will get a lot of successes. others won't despite having the same success chance per fusing.​​

squirrelloid

chaelk said:

as far as I know, success rate is per fusing.

same as RNG rewards, some people will get a lot of successes. others won't despite having the same success chance per fusing.​​

Which is why i did a statistical test to determine how likely it was i simply experienced bad luck. The probability is... very very tiny.