|
|||||||
![]() |
|
|
Thread Tools |
|
|
#1 |
|
Clearly cannot choose the wine in front of you
![]() ![]() ![]() Join Date: May 2010
Posts: 1,507
|
Before I begin, I'd just like to make it known that this is not to be taken as a dig at aesoft, Honko, or anyone else who has contributed to research and integration of Pokemon's incredibly complex damage calculator.
If you're a regular here in LC, you've undoubtedly noticed the delay in any movement by the council. This has been, frankly, due to concern by several members raised on LC damage calculations. Essentially, the question lies within the fact that level 5 calcs are simply different than level 100 ones, and we just aren't sure that PS! is calculating these correctly. Let me explain why these concerns have been raised (by people much smarter than myself, might I add), to the best of my ability, and what you can do to help. A while back, it was noted that masara's damage calc gave slightly different results than Honko's calculator for LC. These discrepancies most likely arose due to the fact that level is taken into account early in the damage formula, which is problematic due to how often the formula rounds. Further concerns and suspicions were raised when some strange discoveries were made regarding multi-hit moves in LC on Pokemon Showdown!; that is, one could almost effortlessly sweep a team with Mincinno not just due to the tendency of the sim to favor better procs (that is, 4 and 5 time occurrences are supposedly much more frequent on PS! than in game), but supposedly also due to just a higher damage output than was correct on each hit as well. You'll notice I'm saying "supposedly" a lot, and that's because we really just don't know. Certain things of the above paragraph are most definitely irrefutable, but others are a mystery. That's why we need you, the community. Initially, a friend of mine and I were going to take confirmation calculations, but it quickly became apparent neither of us have the time or capability to produce the amount of calculations needed to draw an adequate conclusion. Here's how you can help. -Find a friend. Get a team of level 5 Pokemon with known IVs and EVs on cartridge. -Battle, making sure to note the damage outputs received. Do this multiple times--damage outputs always vary, as any calculator will show you--and remember, the more data we can collect, the better -Plug in your Pokemon's IVs and EVs into Honko's damage calculator. You are also welcome to plug it into masara damage calc as well. -Recreate the battle you had on PS!, making sure to note each damage output. -Post your results here from all 3 (or 4) sources so we can confirm and analyze.
__________________
Don't count the days, make the days count. Katakiri: How did you even get into this university?
|
|
|
|
|
#2 |
|
perpetual semifinalist
![]() ![]()
Join Date: Jun 2009
Posts: 1,981
sirndpt <3
|
It would also be nice if we manually did the calc as well, although that should match the cartridge. I have two DSes, so I'll probably be doing a few tests this weekend.
|
|
|
|
|
#3 |
|
Clearly cannot choose the wine in front of you
![]() ![]() ![]() Join Date: May 2010
Posts: 1,507
|
Askaninjask was a true bro tonight and helped me finally do some testing. The results were... unsettling, to say the least.
What we used: Larvesta Stats: 21 / 13 / 12 / 10 / 11 / 9 (Level 5; spammed Ember [the attacking Pokemon]) Lillipup 20 / 11 / 9 / 7 / 11 / 11 (Level 5; spammed Odor Sleuth [the defending Pokemon]) What's important: 10 Special Attack stat using STAB Ember against 11 Special Defense and 20 HP at Level 5.
Honko's calc predictions
masara calc predictions
PS! Results
Cartridge Results
Obviously, we need many, many more tests to confirm results that are as startling as these, so please help if you have some spare time.
__________________
Don't count the days, make the days count. Katakiri: How did you even get into this university?
|
|
|
|
|
#4 |
|
Join Date: Mar 2009
Posts: 1,145
na na la la la hey hey doot do
|
Having done the cartridge battle before the masara calc, you can imagine our surprise when we found that neither Hanko's calc nor PS's calc was right.
...
This is really quite surprising. If anyone needs anyone to help them test, let me know and I'll be on it!
__________________
two antennae get married on top of a roof the ceremony wasn't much, but the reception was excellent |
|
|
|
|
#5 |
|
>> BEGIN POSTBIT
![]() ![]() ![]()
Super Moderator
Join Date: Aug 2011
Posts: 1,131
>> END POSTBIT
|
I have discovered a relevant bug in Pokemon Showdown. Namely, the random factor was being rounded to nearest like most modifiers, but apparently it is not a modifier and should actually be rounded down.
This bug has been fixed, and you can test the new formula on the Pokemon Showdown main server right now. I will be doing all I can to ensure a cartridge-accurate implementation of the damage formula.
__________________
Pokemon Showdown | Replay player | No, I am not impressed that you know that my name is Guangcong Luo. The PS website says "Copyright Guangcong Luo" at the bottom, ffs. |
|
|
|
|
#6 |
|
Clearly cannot choose the wine in front of you
![]() ![]() ![]() Join Date: May 2010
Posts: 1,507
|
New PS! Results
__________________
Don't count the days, make the days count. Katakiri: How did you even get into this university?
|
|
|
|
|
#7 |
|
no abuse pls
![]() ![]() ![]() ![]()
|
gg masara wins
Anyone feel like comparing these results to PO to make sure we weren't playing a lie?
__________________
|
|
|
|
|
#8 |
|
>> BEGIN POSTBIT
![]() ![]() ![]()
Super Moderator
Join Date: Aug 2011
Posts: 1,131
>> END POSTBIT
|
For the record, PS moved to the newly discovered BW damage formula a while ago: http://www.smogon.com/bw/articles/bw...damage_formula
I'd imagine that before that change, PS (using the gen 4 damage formula) would be emitting mostly 4's, too. I'd also imagine that if PO is emitting 4's, that's probably because they're still using the gen 4 damage formula, not because they have the correct formula.
__________________
Pokemon Showdown | Replay player | No, I am not impressed that you know that my name is Guangcong Luo. The PS website says "Copyright Guangcong Luo" at the bottom, ffs. |
|
|
|
|
#9 |
|
<Feranfell> punbot irl aka virginity protector
![]()
Join Date: May 2010
Posts: 1,652
us best
|
Tested on PO using the Larvesta and Lillipup RayJay and askaninjask had.
4, 4, 4, 4, 10 (crit), 4, 4
__________________
![]() |
|
|
|
|
#10 |
|
>> BEGIN POSTBIT
![]() ![]() ![]()
Super Moderator
Join Date: Aug 2011
Posts: 1,131
>> END POSTBIT
|
I just manually worked through the BW formula.
Normally: 93.75% chance of doing 4 damage 6.25% chance of doing 6 damage On crit: 18.75% chance of doing 9 damage 75.00% chance of doing 10 damage 6.25% chance of doing 12 damage Other things I've learned today: Game Freak is ridiculous about their formulas. Half the multipliers round down, the other half round to the nearest, but 0.5 gets rounded DOWN, not up. Anyway, PS now calculates damage accurately.
...
__________________
Pokemon Showdown | Replay player | No, I am not impressed that you know that my name is Guangcong Luo. The PS website says "Copyright Guangcong Luo" at the bottom, ffs. |
|
|
|
|
#11 |
|
Clearly cannot choose the wine in front of you
![]() ![]() ![]() Join Date: May 2010
Posts: 1,507
|
This is fixed, the next test we need to run is a test of a multi hit move (something like Rock Blast from Dwebble or Mincinno Tail Slap) on PS! and cartridge. If that goes well, then I think we're done here.
__________________
Don't count the days, make the days count. Katakiri: How did you even get into this university?
|
|
|
|
|
#12 |
|
she's probably sexting nprtprt
![]() ![]()
Join Date: Aug 2011
Posts: 654
|
Well all I can say is that while I'm not completely sure that multihit moves are completely fixed, they're definitely not as ridiculous as before. Minccino's Tail Slap no longer 2HKOes Aron for one, but checking up on multiple situations with the multihit would still be good to be safe.
A good place to start on: ![]() Minccino used for this (196 Atk Adamant) ![]() Hippo used for this (132 HP / 212 Def Impish @Eviolite) ![]() On this attempt did 16 damage with all 5 hits, which was in the range of the calcs, like expected Calcs for this:
Just need to test some situations on PS to see if we get similar results
__________________
Last edited by EBeast; Aug 18th, 2012 at 2:50:25 AM. |
|
|
|
|
#13 |
|
there is actually no underscore in my name
![]() ![]() ![]()
Moderator
Join Date: Jan 2008
Posts: 5,484
|
EBeast, you rock. Can you do me a huge favor and also do those calcs with Minccino holding a Life Orb and a Choice Band?
EDIT: I know it's dumb, but could you also do those calcs without the Eviolite on Hippopotas? Or at least the one you did above? |
|
|
|
|
#14 |
|
Clearly cannot choose the wine in front of you
![]() ![]() ![]() Join Date: May 2010
Posts: 1,507
|
Yeah em's also right because if Hippo doesn't have Eviolite then there should be a greater range of possible outputs iirc but you do indeed rock man
__________________
Don't count the days, make the days count. Katakiri: How did you even get into this university?
|
|
|
|
|
#15 |
|
Join Date: Apr 2011
Posts: 418
North Korea
|
BS Ebeast stealing my pics without giving me credit for taking them over a month ago.
Anyway, next pics are LO vs Eviolite, LO vs no item, and CB vs no item: LO vs evio Hippo: ![]() ![]() DamageLO vs no item I think I got a crit on this one or a low damage roll. Band vs no item Hippo: ![]() ![]() DamageCalcs: Masara: 252 Atk Life Orb Minccino (+Atk) Tail Slap vs 132 HP/212 Def Eviolite Hippopotas (+Def) : 16% - 16% (7 hits to KO). All 5 hitting should be 20 252 Atk Choice Band Minccino (+Atk) Tail Slap vs 132 HP/212 Def Hippopotas (+Def) : 16% - 24% (5-7 hits to KO) 252 Atk Life Orb Minccino (+Atk) Tail Slap vs 132 HP/212 Def Hippopotas (+Def) : 16% - 16% (7 hits to KO) Kalashnikov 252Atk Life Orb lvl 5 Minccino (+Atk) Tail Slap vs 132HP/212Def Eviolite lvl 5 Hippopotas (+Def): 60% - 80% (15 - 20 HP). Guaranteed 2HKO. 252Atk Life Orb lvl 5 Minccino (+Atk) Tail Slap vs 132HP/212Def lvl 5 Hippopotas (+Def): 60% - 80% (15 - 20 HP). Guaranteed 2HKO. 252Atk Choice Band lvl 5 Minccino (+Atk) Tail Slap vs 132HP/212Def lvl 5 Hippopotas (+Def): 80% - 120% (20 - 30 HP). Guaranteed 2HKO. 50% chance to OHKO. PS: NU God sent out Minccino! vergil must die sent out Hippopotas! Turn 1 The foe's Minccino used Tail Slap! Hippopotas lost 12% of its health! Hippopotas lost 12% of its health! Hippopotas lost 12% of its health! Hippopotas lost 12% of its health! Hippopotas lost 12% of its health! Hit 5 time(s)! The foe's Minccino lost some of its HP! LO vs evio hippo Minccino used Tail Slap! The foe's Hippopotas lost 20% of its health! The foe's Hippopotas lost 20% of its health! The foe's Hippopotas lost 20% of its health! The foe's Hippopotas lost 20% of its health! The foe's Hippopotas lost 20% of its health! Hit 5 time(s)! Minccino lost some of its HP! LO vs no item Hippo The foe's Minccino used Tail Slap! Hippopotas lost 20% of its health! Hippopotas lost 20% of its health! Hippopotas lost 20% of its health! Hippopotas lost 20% of its health! Hippopotas lost 20% of its health! Hit 5 time(s)! Band vs no item Hippo
__________________
16:46 <TLCNU> Esca literally is 2HKO'd by Stoutland Fire Fang |
|
|
|
|
#16 |
|
she's probably sexting nprtprt
![]() ![]()
Join Date: Aug 2011
Posts: 654
|
To follow up on Trop's calcs
From the above calcs it showed that Masara had some strange calculations at LO vs no Eviolite, and I have noticed some calcs from it being a bit off when doing a few on my own. So for the sake of this we should use Kalashnikov or Honko as our damage calculators. From Trop's Ps calculation we can see that multihit moves are still off in some way as LO vs no Eviolite Hippo on PS was doing 20% per hit, which all 3 calcs say isn't possible. Also on CB vs no item, each hit is doing 20% while Honko's calc says the possible variations are (4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 6) 20% of Hippo's HP is 5 damage, a variation that is not possible.
__________________
Last edited by EBeast; Aug 18th, 2012 at 4:14:17 PM. |
|
|
|
|
#17 | |
|
pronounced "Honko"
![]()
Join Date: Dec 2009
Posts: 1,023
|
Quote:
I've updated my calc with this, it's correct for the Larvesta vs Lillipup calc now, and hopefully for everything else as well. Please let me know if you find any other problems. EDIT: oops, I meant unrounded. Basically my question is, does the remainder from divmod get used or not? My problem was that I was assuming it was used, but it sounds like it gets dropped at least for the base damage and random factor. Is it dropped everywhere else that uses ÷ as well?
__________________
![]() Last edited by Honko; Aug 18th, 2012 at 5:26:28 PM. |
|
|
|
|
|
#18 |
|
there is actually no underscore in my name
![]() ![]() ![]()
Moderator
Join Date: Jan 2008
Posts: 5,484
|
Huh? From my understanding, everything that uses the ÷ is floored, but modifiers (stab, choice band/specs to [sp]atk, eviolite to defenses, the random factor, LO, etc) are rounded.
On that note, doing it out by hand you get the following for all the Tail Slap vs. Hippo calcs: No item vs. no item: 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4 No item vs. Eviolite: 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4 Life Orb vs. no item: 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 5 Life Orb vs. Eviolite: 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 5 Choice Band vs. no item: 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 6 Choice Band vs. Eviolite: 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 4 ----------- EDIT: that seems inconsistent for the LO vs. no item pic that Trop posted though. Is there any way you could redo that ingame? Because even with a crit (should do 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 12), there's no way to get 22 damage. I guess it's possible with two high rolls, but that seems unlikely? Last edited by Elevator Music; Aug 18th, 2012 at 5:10:30 PM. |
|
|
|
|
#19 |
|
Join Date: May 2012
Posts: 45
|
Life orb does the same against eviolite and no item? Weird.
also, ban mincino. Too powerful! BS probably OHKO's then.
__________________
Terrakion, Infernape Keldeo. The trio of destruction. RMT up (GothiSun) |
|
|
![]() |
| Thread Tools | |
|
|