RBY Random Battles - Level Balancing

[Crystal_]

Even though randbats aims to be as close to competitive as possible, it still leaves a lot to be desired. Some of the possible movesets and move combinations are very suboptimal, to the point that some Pokemon such as Fire-types or Water-types often come without a STAB move, and a few Pokemon may even come with no damaging move other than Explosion/Selfdestruct. A lot of Pokemon may also come with moves that have no use whatsoever even though there are far more useful options in their movepool. I've been working to improve the efficiency of the movesets while still keeping some variety in Pokemon that can successfully use different kind of sets.

In addition to movesets, I also intended to balance the levels using my own criteria. Instead of being based off just the tier, each Pokemon now gets a level based on its overall strength in a metagame where all 151 Pokemon may show up. The tiers are accurate when the player is given the choice to select 6 Pokemon, but when Pokemon are assigned randomly, other factors such as BST should be taken into account. Based on my knowledge of this tier, as well as RBY OU, I've tried my best to estimate each Pokemon's strength when assigning the levels. This level balance also opens the possibility of making OUs not so blatantly common as they are.

When I submitted the pull request to the Pokemon Showdown repository with the changes, some problems with balanced the levels were brought up:
- Firstly, it was always my own criteria. Even though I have knowledge in this format and in all the RBY tiers, if someone else was given the task to balance out the levels, they'd probably choose slightly higher or lower levels in some cases, probably up to a 4-level difference.
- As it stands right now, levels are based on tiers. This means that Pokemon in the same tier retain their relative speed tiers as we're mostly used to.

My goal with this thread is to hear opinions on whether the levels should be kept as they are, or should be further balanced according to overall strength rather than just its RBY tier. If you think the latter and would be interested in suggesting a level spread for each Pokemon, I'd love to hear your opinion on it.

The levels I suggested can be seen here: https://github.com/xCrystal/Pokemon...fc675076a64c/mods/gen1/scripts.js#L1286-L1311 The Pokemon that aren't referenced in the custom scale use the default level given to the tier it belongs to. RBY tiers can be seen here: http://www.smogon.com/dex/rb/pokemon/

I've opened a poll regarding this matter in this thread, and also in the RoA room of Pokemon Showdown.

The pull request can be seen here: https://github.com/Zarel/Pokemon-Showdown/pull/3210

 

[Crystal_]

In regards to the moveset changes.

Proposed changes can be seen here, compared to the old movesets: https://github.com/Zarel/Pokemon-Showdown/pull/3210/files
And here, just the new proposed changes alone: https://github.com/xCrystal/Pokemon...0248752fc675076a64c/mods/gen1/formats-data.js

Here's how it works:
- comboMoves have a 50% chance to be picked, either all of them or none
- If there's still room for more moves, pick one of exclusiveMoves at random
- If there's still room for more moves, add essentialMove
- Fill the moveset with randomBattleMoves (pick at random, with some obscure restrictions that are contemplated here)

 

[Mirabel_]

Can't be overstated that the existing RBY tiers should have absolutely no bearing in randbats

 

[Crystal_]

Can't be overstated that the existing RBY tiers should have absolutely no bearing in randbats

They have, under the premise that factors such as movepool, type and stat spread are what makes each Pokemon belong to the tier it belongs to. The reason I gave Alakazam a higher level than Golbat is not because of their tier difference, but because of their movepool, stats, and type differences. These are, in turn, the factors that make Alakazam belong to OU and Golbat belong to bottom UU/NU.

 

[Mirabel_]

They have, under the premise that factors such as movepool, type and stat spread are what makes each Pokemon belong to the tier it belongs to. The reason I gave Alakazam a higher level than Golbat is not because of their tier difference, but because of their movepool, stats, and type differences. These are, in turn, the factors that make Alakazam belong to OU and Golbat belong to bottom UU/NU.

Well you improve everything you touch and it's not saying much to declare this no exception, but... given that randbats is the most popular RBY meta, you intend to design randbats mons into the roles they have in metas many of which barely even exist today? Without even unbanning wrap as per Smogon OU? I thought from the original post you weren't going to prioritize the identity some mons accrued under smogon tiering over the best fit for this 151 mon wrapless meta.

 

[Crystal_]

Well you improve everything you touch and it's not saying much to declare this no exception, but... given that randbats is the most popular RBY meta, you intend to design randbats mons into the roles they have in metas many of which barely even exist today? Without even unbanning wrap as per Smogon OU? I thought from the original post you weren't going to prioritize the identity some mons accrued under smogon tiering over the best fit for this 151 mon wrapless meta.

The main goal is to have every Pokemon have similar strength and usefulness, and that's achieved by carefully assigning a level to each Pokemon. Whether they keep their role in the metagame they are often played or not is secondary. No Pokemon will have exactly the same role or utility as in one of the actual metagames, since in randbats each Pokemon has to expect to be paired up against any combination of 6 of the 151 Pokemon, and ideally where each of those 151 are about as useful as each other. It's not about assigning levels so that each Pokemon fits its original role as close as possible, but so that no Pokemon is underpowered and therefore the difference between the strength of both player's teams is as low as possible. Also, yes, I decided to keep trapping moves as well as Fly/Dig, evasion moves, and OHKO moves all banned, because I thought that was what most people would prefer. Granted, Wrap moves are no longer banned in the main tiers, but I believe bringing them back in randbats would make more than one unhappy (personally I don't really care).

 

[Bughouse]

Can you provide a speed tier list with these new levels? I see for example that Gengar now outspeeds Tauros, for example. Since Tauros is 58 and Gengar is 60.

Moves largely look like an improvement.

 

[Mr.E]

The only seeming disadvantage of manually adjusting levels is that it's more work, so if someone reliable and skilled is willing to do that...

 

[Crystal_]

Can you provide a speed tier list with these new levels? I see for example that Gengar now outspeeds Tauros, for example. Since Tauros is 58 and Gengar is 60.

Moves largely look like an improvement.

It seems like the custom levels are going to be declined, but if they are accepted I'll try to work on those speed tiers.

The only seeming disadvantage of manually adjusting levels is that it's more work, so if someone reliable and skilled is willing to do that...

You can see my proposed levels here: https://github.com/xCrystal/Pokemon...fc675076a64c/mods/gen1/scripts.js#L1286-L1311

 

[Zarel]

The reason randbats in PS gives every Pokémon in the same tier the same level is so speed tiers are retained within a tier. This is useful because then playing randbats can help you learn the speed tiers for OU, etc.

Is there that much more value in subjectively deciding every Pokémon's level individually? I can be convinced that way if there are good reasons for it, but I'm not really seeing good reasons.

We don't need every Pokémon at exactly the same effectiveness, they can differ a bit, and tier-based level adjustment is usually enough not to make them differ too much.

 

[Crystal_]

The reason randbats in PS gives every Pokémon in the same tier the same level is so speed tiers are retained within a tier. This is useful because then playing randbats can help you learn the speed tiers for OU, etc.

Is there that much more value in subjectively deciding every Pokémon's level individually? I can be convinced that way if there are good reasons for it, but I'm not really seeing good reasons.

We don't need every Pokémon at exactly the same effectiveness, they can differ a bit, and tier-based level adjustment is usually enough not to make them differ too much.

I think it has to do with the fact that I always understood the goal of randbats differently. I always thought the goal was to approximate each Pokemon to have the same effectiveness, and that the fact that levels were merely based on tiers with very few exceptions was that nobody had bothered to work on them.

I never considered randbats as a way to become familiar with the generation and the speed tiers, but I can see the point on it, and if that's the case, then yeah, I guess it'd be good to have levels consistent across tiers (even though then you could also make a point about having all fully evolved Pokemon at the same level, since lower tier Pokemon may show up anytime in the higher tiers).