• Follow our Instagram!
  • Curious about what we're currently doing about bots? Check out the latest updates from staff in this thread.

Battle Bot Reached VGC #1 with my bot

I know it is not the best time to make this post, and I absolutely agree to enforce regulations on bots on ladder.

This project took me about one week after mikumiku39 reached #1. I first reached #1 entirely by myself the next day with my psych spam team, and then I developed the RL bot that could beat me both in ELO and GXE.

The thing I couldn't expect is that today so many bot developers simultaneously posted on Smogon, so my assumption is that this is the expected training time since the first bot made #1 on VGC ladder. So I changed my plan to not further develop this bot considering its influence on ladder and how it reduces human playing experience. Instead I will focus on human playing experience like match up planner and single player rouge-like OU or VGC battle simulator.
 

Attachments

  • 54E365032DF00A4702004CDBFFD2F5D0.png
    54E365032DF00A4702004CDBFFD2F5D0.png
    757.1 KB · Views: 343
  • x_post_climb.png
    x_post_climb.png
    705.6 KB · Views: 345
  • x_post_teams_light.png
    x_post_teams_light.png
    648.7 KB · Views: 339
It's honestly really cool, even if it's a bit scary ngl, to see a lot of really good ai development. I think the potential to help a lot of players is definitely up there. Have you considered develop the possibility to test against the bot, maybe it's/bo3 as a tournament prepping tool? I'd love testing against a specific team in bo3 and I don't see that kind of project with bots ATM (or I totally misinterpreted and the human playing experience is this Kind of stuff). Big stuff tho
 
It's honestly really cool, even if it's a bit scary ngl, to see a lot of really good ai development. I think the potential to help a lot of players is definitely up there. Have you considered develop the possibility to test against the bot, maybe it's/bo3 as a tournament prepping tool? I'd love testing against a specific team in bo3 and I don't see that kind of project with bots ATM (or I totally misinterpreted and the human playing experience is this Kind of stuff). Big stuff tho
Thanks! Currently my bot is in maintenance and not taking any challenge, but I am sure many other bots are willing to take.

For testing against a specific team, I think the logic is simple. You type "MC999" and then challenge the bot, then the bot will use that team if it is in the pool (not yet implemented in my bot tho).
 
There's like a dozen new bots a day and they all go and say some something along the lines of
I absolutely agree to enforce regulations on bots on ladder.
Just stop making ai bots! There's a dozen people to do this before you, you aren't really proving anything new about ai bots on ladder, you aren't really contributing to solving the issue, and it's not even that impressive. The only people who think this is cool is the other people making the ai bots.

Also the website and probably the bot as well, looks vibe coded
 
There's like a dozen new bots a day and they all go and say some something along the lines of

Just stop making ai bots! There's a dozen people to do this before you, you aren't really proving anything new about ai bots on ladder, you aren't really contributing to solving the issue, and it's not even that impressive. The only people who think this is cool is the other people making the ai bots.

Also the website and probably the bot as well, looks vibe coded
I do not think we should accuse people of vibecoding out of frustration especially when the only information given is just 2 screenshots of the UI
 
i would love for them to prove me wrong if they do that would be awesome
Mine's vibecoded entirely and I'm pretty sure most, if not all, of the rest are too.
There's like a dozen new bots a day and they all go and say some something along the lines of

Just stop making ai bots! There's a dozen people to do this before you, you aren't really proving anything new about ai bots on ladder, you aren't really contributing to solving the issue, and it's not even that impressive. The only people who think this is cool is the other people making the ai bots.

Also the website and probably the bot as well, looks vibe coded
I agree it's not particularly impressive or helpful to do it at this point, but there will always be people that want to do it solely because it's kinda neat if you've never done something like it before, it's very easy to do, and the platform basically gives them no reason not to do it.

For the record, I'm in favor of having a toggle switch that's off by default for users where they can opt into playing bots on ladder, asking every bot to register themselves as a bot, and banning anyone who runs a bot without labelling it as such.
 
Last edited:
I do not think we should accuse people of vibecoding out of frustration especially when the only information given is just 2 screenshots of the UI
Doesn't take a genius to realize that what used to be weeks of writing boilerplate and ensuring ur local selfplay engine works properly is now a weekend if u point a frontier model at it. Transformers have been around for years, there's a reason we're only seeing people apply the tech to Pokemon now - not using a language model to do this would be silly when it's a gigantic force multiplier for exactly this usecase
 
Doesn't take a genius to realize that what used to be weeks of writing boilerplate and ensuring ur local selfplay engine works properly is now a weekend if u point a frontier model at it. Transformers have been around for years, there's a reason we're only seeing people apply the tech to Pokemon now - not using a language model to do this would be silly when it's a gigantic force multiplier for exactly this usecase
i don't have a problem with that but the labeling raffe was using was meant be used to dismiss the project not as a general disclaimer, which i do take issue with
 
it's very easy to do, and the platform basically gives them no reason not to do it.
I would say the general pushback from the community here being “please stop doing this at least until the bot registry is fleshed out” should be reason enough for people to not do this. These projects are cool and all but ignoring the sentiment of the community that actually plays and enjoys this game so you can show of your project that you yourself say isn’t that hard to do just seems rather disrespectful.
 
I would say the general pushback from the community here being “please stop doing this at least until the bot registry is fleshed out” should be reason enough for people to not do this. These projects are cool and all but ignoring the sentiment of the community that actually plays and enjoys this game so you can show of your project that you yourself say isn’t that hard to do just seems rather disrespectful.
Community pushback is not enough reason to not do it for at least some people. Like idkup said this is basically a weekend project of pointing an AI at it and saying "Make a bot that plays better than every human in this format. Make no mistakes" and it just kinda does it.

I'm not saying it should or shouldn't be done btw, just that some people have different value systems, and if allowed they will do it for no reason other than "idk, seems neat", and whether they show their work to the rest of the community or keep it hidden once it's been done already is ultimately up to them. I expect people to find it much less appealing once a bunch of these have been made already, or one gets released open source and everyone has access to it so there's not much point to training your own.
 
I know it is not the best time to make this post, and I absolutely agree to enforce regulations on bots on ladder.

This project took me about one week after mikumiku39 reached #1. I first reached #1 entirely by myself the next day with my psych spam team, and then I developed the RL bot that could beat me both in ELO and GXE.

The thing I couldn't expect is that today so many bot developers simultaneously posted on Smogon, so my assumption is that this is the expected training time since the first bot made #1 on VGC ladder. So I changed my plan to not further develop this bot considering its influence on ladder and how it reduces human playing experience. Instead I will focus on human playing experience like match up planner and single player rouge-like OU or VGC battle simulator.
Does the bot build the team and moves or does it just battle? I think exploring optimal team space is a lot more interesting than perfect battling but you kind of need to master the latter to do the former.
 
Community pushback is not enough reason to not do it for at least some people. Like idkup said this is basically a weekend project of pointing an AI at it and saying "Make a bot that plays better than every human in this format. Make no mistakes" and it just kinda does it.

I'm not saying it should or shouldn't be done btw, just that some people have different value systems, and if allowed they will do it for no reason other than "idk, seems neat", and whether they show their work to the rest of the community or keep it hidden once it's been done already is ultimately up to them. I expect people to find it much less appealing once a bunch of these have been made already, or one gets released open source and everyone has access to it so there's not much point to training your own.
don't even quantify it with 'if allowed'

maybe it will be easier to pretend it isnt happening if people arent posting but this requires no real commitment, very limited pokemon knowledge, and very limited technical ability
 
Does the bot build the team and moves or does it just battle? I think exploring optimal team space is a lot more interesting than perfect battling but you kind of need to master the latter to do the former.
Currently the bot only uses public teams. In internal self-play some teams look very strong, but those rankings only loosely predict ladder results. Since the bot's view of the meta isn't the real meta, I think it is still a supportive tool which relies on human to make the team building decisions.
 
Smogon openly supports collaborating with people flooding this forum with AI garbage and botting the ladder, they don't care that you ruined hundreds of peoples games on the ladder. So this isn't surprising in the slightest.

I get the frustration, especially if bots have ruined the experience for people. But I think there's a difference between Smogon's approach and supporting botting on the ladder.

They've confirmed that site leadership is discussing concerns around "AI/bots, laddering, tiering, tournaments, anti-cheating measures" and more, with the goal of finding solutions (https://www.smogon.com/forums/threads/the-future-of-smogon.3787747/).

Their announcement (https://www.smogon.com/forums/threads/initial-policy-on-bots.3789528/) also states:

"We will use that registry to preserve the parts of the PS / Smogon ecosystem which by definition have to be human based (suspect reqs, usage stats, general laddering).

All of these key parts of how we run the website today are affected by bots. This is where a lot of policy and technical discussion is focused right now, so there will be more announcements on precise details in the future."

Want to criticize them for not taking immediate action or question whether their approach will work? Sure, you're entitled to that. But not having a finalized solution doesn't mean they support botting on the ladder (seriously, they literally mentioned this topic lol...). Whether their approach will be effective is a separate question.

Given all that, I'm not sure how fair it is to say that "they don't care".

----------

"Smogon this, Smogon that".

Folks, let's give them some time to figure things out. They've already addressed the situation and made it clear that they're looking into it.

I get that this is a heated topic and people are bound to feel frustrated. Still, since this thread is primarily about the bot itself, how effective are these comments compared to reaching out to Smogon directly? They feel a bit out of place, though not entirely unrelated to the topic.
 
Last edited:
Back
Top