GamePress

Gym Battle Simulators

I've been developing a couple of models, one for ranking movesets based on DPS versus particular opponents and another that simulates a gym battle.

I'm in the backtesting portion of the development and I'm trying to find a gym battle simulator which produces an accurate simulation. Strangely, the two that I've found online (including the one that's been my go to up until now) have behaviors that I can't explain.

Has anyone else who's used these understand the below?

I have always relied upon the PokeAssistant.com Battle SIM as my go-to simulator. But when seeded it as follows:

Attacker: Level 35 Ursaring (15/15/15) with Metal Claw/Close Combat
Defender: Level 35 Blissey (15/15/15) with Zen Headbutt/Hyper Beam

I see something I don't understand starting with 90.9 seconds remaining. Ursaring uses a Metal Claw at 90.9--that has a cool down of 0.7 seconds, so it's next attack should be at 90.2. However, it's next attack is shown at 90.899 seconds--a mere .001 second later--and it's the charge move Close Combat. It's next attack should be Close Combat since it's saved up 100 energy after the prior Metal Claw, but it shouldn't be able to use it until 90.2 seconds, right?

You see a similar pattern between Blissey's quick attack at 89.1 and charged attack at 89.099. It appears as if the sim executes a charge move the instant there is enough energy without regard for cool down.

Am I right in assuming this is a bug or is there perhaps something about the gym rules that explains this?

Once I noticed that (potential) issue with pokeassistant's sim, I tried using PokeBattler.com with the same mon setup, but I'm also seeing some things I don't understand.

First, the initial attack by the Attacker isn't until 98.9, when I thought it should be at 99.3. Similarly, the first attack by the Defender isn't until 97.5 when I thought it should be at 98.4.

Secondly, even though I selected "Expected" for defender behavior, Blissey appears to use Zen Headbutt every 1 to 3.2 seconds. I thought with the "Expected" option, it would consistently use the cool down plus 2 seconds. Is that not the case?

You can see this one yourself easily via the link below:

http://www.pokebattler.com/fights/attackers/URSARING/quickMoves/METAL_CLAW_FAST/cinMoves/CLOSE_COMBAT/levels/35/ivs/FFF/defenders/BLISSEY/quickMoves/ZEN_HEADBUTT_FAST/cinMoves/HYPER_BEAM/levels/35/ivs/FFF/strategies/DODGE_SPECIALS2/DEFENSE

I would also love is someone can confirm the gym behavior of charge moves. I'm currently assuming an average of a 2 second penalty to each charge move--meaning it's cool down is what the move's cool down is plus 2 seconds. This is based on the 1.5-2.5 figures I've seen here. While I've seen good correlation between my simulator and others in regards to quick moves, I haven't seen the same in regards to charge moves. PokeBattler.com in particular mentions that for their "Expected" defender behavior, the "special attacks are always 1 attack late". Does anyone know what that means or why that's the approach they took? Is that how gym battles work for charge moves?

As I mentioned, I'm not doing this to discredit these sites, I've relied on pokeassistant.com for quite some time, but I want to make my simulator as accurate as I can, so it'd be great if there were something to compare it to.

Any ideas/suggestions?

My plan, by the way, is to release this once it's completed. I'd guess in the next week or so.

Asked by TotoroWrangler7 years 1 month ago
Report

Answers

by Puppi 7 years 1 month ago

Sorry mate I'm not gonna go digging through all this stuff and unfortunately can't answer your questions off the top of my head with certainty. Although I agree with both of your observations, they both seem odd.

If you don't find the validation that you are looking for either here or elsewhere on the net, then go with your head.

I notice a lot of people are happy to slap together a spreadsheet with a bunch of assumptions that they probably don't even realize they made... and ultimately do injustice to certain pokemon/moves.

Based on the issues you have raised in your post, I have confidence that you're gonna get it pretty close. And if not... put it down to a learning experience. Seem's like you're having fun. :)

Up
0
Down

Thanks at least for the reply. I guess my post was too long for most to go through, digest and respond to. :D

And I agree with your assessment. My best guess is that my models are accurate, perhaps more so than others that are out there, so I'll release it soon. If I've made mistakes, I'm sure I'll hear about those once others get their hands on it. :)

Up
0
Down