Fixed - Expedition upgrade "Sensors" doesn't work properly.

[Phayte]

At level 10, this upgrade gives a 20% chance for increased returns. I've been keeping track since maxing Sensors, and in 155 total expeditions I have only had 7 with increased returns (4.5%). Even when only counting successful expeditions, that leaves 113, or 6.2%, nowhere near what it should be.

I'm aware my sample size isn't huge, but the roughly 5% has been fairly consistent since I started tracking:
5 in 90 (5.5%)
6 in 114 (5.2%)
7 in 155 (4.5%)

I feel my sample size is large enough to discount 'unlucky' as an explanation.

[universal123]

(04-26-2015, 06:40 AM)Phayte Wrote: At level 10, this upgrade gives a 20% chance for increased returns. I've been keeping track since maxing Sensors, and in 155 total expeditions I have only had 7 with increased returns (4.5%). Even when only counting successful expeditions, that leaves 113, or 5.2%, nowhere near what it should be.

I'm aware my sample size isn't huge, but the roughly 5% has been fairly consistent since I started tracking:
5 in 90 (5.5%)
6 in 114 (5.2%)
7 in 155 (4.5%)

I feel my sample size is large enough to discount 'unlucky' as an explanation.

Agreed, for a 5% margin of error, (which would be between 0% and 9.5%, so well below the 20% expectation), you can be 71.2% confident of being correct with a 113 sample size. If you accept a 10% margin of error (between 0% and 14.5%, still well below 20%), you can be 96.65% confident. The confidence with a 15.5% margin of error (enough to actually hit 20%), is 99.9%.

TL;DR: What you said is backed up by statistics, the odds of being wrong are pretty small.

SOURCE: http://www.raosoft.com/samplesize.html

[WhiteK]

yeah i have the same through sensor 99.99% broke :3 hope Ray fix the rate fast

[Ray]

How long have you been sampling this? It's possible I made a fix in the last couple patches (and don't recollect).

The way the check is done is:

Rand(10000) <= SensorLv*100 ? multiplier = <2 to 4> : multiplier = 1

I do a higher randomization than just rand(100) <= SensorLv because it gives a better, cleaner check with a stronger number range.

[Phayte]

I might be overreading this, but SensorLv is actually 2*Sensors level, right? Since it's supposed to be 2% per level

There are 180 expeditions on my list, I think I've been keeping track for about 6 weeks.

[Phayte]

Since the update 16 failed, 30 succeeded of which 6 with better yield, exactly 20%.
All good now!

[WhiteK]

Agree sensor FIXED topic CLOSED :v