- Quick overview of the best monsters, type-wise and stats-wise, at a maximum tier.
- Computation of the best party in the top of the pool.
- Monster bans.
- Cook an egg on your laptop.
- Simulate a combat.
- Take into account cancer tactics, specific skills or items.
- Use your GPU (for now?)
Note
Requires the user to input the monsters data (monsters.json)
which is not provided here.
It can be dumped out of PokeMMO very easily.
More on that below.
Warning
Will max and heat up your CPU. Spreading the workload across a lot of threads increases RAM usage. Expect the computation of all the combinations of 6 in 70 monsters (131 115 985 parties) spread across 12 threads to require 14Gb of free RAM.
- Go to settings
See image
- Go to utilities, Dump Moddable Resources
See image
- Select Monsterdex Data, click Yes
See image
- Check confirmation popup
See image
- Find the archive in dump/resources
See image
- Extract the archive with your favorite file archiver
- Open monsters.json, copy all its content
- Paste it in the relevant prompt, on the live demo or locally
- Follow instructions there
- It makes a pool out of all the monsters of a tier and the one tier lower.
- It sorts the pool based on the arithmetic average of the normalized edges of one monster against all the other monsters of the pool.
- It takes the top X monsters of the pool, makes as many combinations of 6 different elements in that set as possible.
- It compares the performance of each party against all other monsters of the pool, so not only against the monsters in the top X, and gives the best party.
- The stats difference of a monster against another is the difference of the stats of both, ignoring the worst attack stat of an attacker (attack or sp_attack) and ignoring the corresponding defense stat of the defender (ex: higher attacker's attack -> ignore defender's sp_defense)
- The normalized stats difference of a monster against another is the fourth root of the relevant stat difference, which makes it closer to 1.
- The type edge of a monter (attacker) against another (defender) is the best damage multiplier of the attacker against the defender based on their types, divided by the best damage multiplier of the defender against the attacker based on their types.
- The normalized edge of a monster against another is the type edge multiplied by the normalized stats difference.
- After making all the parties, we give them a score, so that the party with the highest score is the best party.
- We get the score of the party against one monster by calculating the geometric average of all the normalized edges of the monsters in the party against the monster in the pool, but we gave twice as much weight in the average to the best monster in our team. In the context of a geometric average, it makes the highest normalized edge of the party (best monster to pick) as important as the ones of all the other monsters in the party. It is relavant because we can expect to be fighting a monster with our best counter around half the time(?).
- The final score of a party is the geometric average of all the scores of the party against each monster of the pool.
-
Reasons to make it party vs monster pool instead of party vs parties? Useless redundancy, since comparing a monster combination against all possible combinations is the same as comparing a monster combination against all other monsters a really big number of times. So we compare each party against the whole monster pool instead, only once, which gives the same result but in minutes instead of weeks.
-
Why so slow with big numbers? Javascript.
-
Why Javascript? You can use this in your web browser without downloading any file.
-
Why is the demo page so ugly? Make the HTML yourself. I hate it. CSS? never heard of that.
- Tommy Odland for his paper on the topic which gave me the inspiration to do that.
- The PokeMMO team, obviously. (Which this project is not affiliated with)
- Some japanese company that likes balls.
PokeMMO Optimal Party by Maxime Ghazarian is licensed under CC BY-NC-SA 4.0
