In the Maximizing Candies article, should it be Floor or Ceiling function?
This is the article: https://pokemongo.gamepress.gg/maximizing-candy-raid-bosses-pinap-or-golden-razz
I created my own calculator, and the results disagree with what's written in your article.
https://www.reddit.com/r/TheSilphRoad/comments/9ky3tw/raid_candies_calculator/
Ignoring Silver Pinaps, my calculator finds that using 9 Pinaps and 3 GRB on a Level 20 Kyogre (2% base rate) with Excellent Curveball Throws and Psychic Medal is Gold will yield 3.25208 Kyogre Candies. But the GamePress recommendation of using 4 GRBs (and thus 8 Pinaps) yields 3.23666 Kyogre Candies. (Edit: I initially used the base 5 candy Mewtwo results. But I figured I should use base 3 results and change it up to a Kyogre. The ultimate rounding question still applies.)
The gamepress article clearly shows the Ceiling Brackets for rounding k, where k was 3.7 something.
But I'm also curious about their calculation for the GRB/Pinap Excellent Throw Non-weatherboosted rate. Their table displays 14.668% / 6.154% for GRB / Pinap, while my own calculator (which may be flawed in some way) uses (when rounded to same number of digits) 14.662% / 6.145%. I couldn't quite get both numbers to match; if I get the GRB to match at 14.668% by using a throw multiplier of 1.7007, I get for the Pinap 6.147%.
So maybe the discrepancy comes from either of our individual catch chance estimates being off.
If anyone wants to check the maths and make sure I implemented the formula correctly in Cell C13 in my spreadsheet, I'd appreciate that.
Answers
Hi, I'm the writer of the article.
As you pointed out, the differences in our calculations for catch rates under each scenario caused the discrepancies. My best guess is that my choice of r that is different from yours, and I forget which r I chose at that time.
But the formula for k is under strict math derivation, and I'm more confident on that part. I just double checked my result.
Hey, thanks for writing it up! It's been fascinating to me.
What's odd to me is that if I could find an r that works to get a % catch value for one of the berries, then it should also get me the value for the other berry, but I couldn't do it. So there may be two slightly different r's, or there was premature rounding involved that's yielding differences.
I liked when the math said you'd ceiling it, as it meant "We want 3.7 throws to use GRBs. We can't do 3.7, and 3 would be too "late", so we compromise and go with a 4." But even when ceilinging both a 3.7 or a 3.2 result by changing that r, and comparing it to flooring it, the floor result of 3 seems to always give just a bit more in terms of berries.
I've tried following the math in the reddit post linked at the end of the article, but that seemed like a different approach (and doesn't help that it appears to use a bunch of different variables). I've been out of calculus for 7 years, so, I'm rusty on my maths and manipulation to follow it readily.
Maybe I have a flaw in the expected overall catch % and the expected number of berries, so the k is being *misused*, because I tried to work in the silvers into the equation. I'll try creating that formula again from scratch and compare the two.
Hi, follow up. I found a really good suspect for a mistake... My expected candies equation failed to account for the transfer candies. Oops. I'd have traced dependents to figure that out, but, I had moved from Excel to Google Drive before doing that final testing. Was too lazy to copy back from Drive to Excel. That fault rests on me.
