Remove Bots from MTGA !

[taba]

Remove Bots from MTGA!

I have been playing various online versions of MTG for ten years.
I think I can tell when I'm playing with a human player and when I'm playing with a Bot.

If the game detects that a player has won too much, Bots will be the opponents.

The Bot is impossible to beat.
Bot always has a better cards.
The game cheats.

I don't think it's fair to the players.

(Please share this for MTG Twitter and other social media)

[Doctor Zalgo]

NObodies_heRE původně napsal:

I don't know about bots but when im on a win streak i tend to end up with opponents that have the exact deck needed to counter or just destroy my deck within the the first 4 turns. then I consistently get fight those decks for 5-8 rounds, get tired of losing instantly then switch decks and repeat the process of winning then being forced to lose.

Assuming perfect randomness from the matchmaker, how many games would you reasonably expect to play before running into someone who has a counter for your deck? Like, what number would make you think 'this isn't rigged'. Because I suspect the answer is 'any time I lose a game its rigged'.

[John-Silver]

Doctor Zalgo původně napsal:

NObodies_heRE původně napsal:

I don't know about bots but when im on a win streak i tend to end up with opponents that have the exact deck needed to counter or just destroy my deck within the the first 4 turns. then I consistently get fight those decks for 5-8 rounds, get tired of losing instantly then switch decks and repeat the process of winning then being forced to lose.

Assuming perfect randomness from the matchmaker, how many games would you reasonably expect to play before running into someone who has a counter for your deck? Like, what number would make you think 'this isn't rigged'. Because I suspect the answer is 'any time I lose a game its rigged'.

every 100 or so games created by the matchmaker not counting just your games (commonly with so many games), you should be asking the rate the matchmaker creates a counter scenario match-up across all games it creates not yours exclusively rng for this doesn't stop at you.

[anaris]

yup! 1 game in 100-10000 being a weird coincidence is a ballpark for the kinds of things most people talk about, and like... there are 1.4 million accounts logging in every day and playing 1-100 games each, so 1 in a million chances can crop up a hundred times a day

[anaris]

congrats ur today's lucky user

[Imnuktam]

taba původně napsal:

Remove Bots from MTGA!

I have been playing various online versions of MTG for ten years.
I think I can tell when I'm playing with a human player and when I'm playing with a Bot.

If the game detects that a player has won too much, Bots will be the opponents.

The Bot is impossible to beat.
Bot always has a better cards.
The game cheats.

I don't think it's fair to the players.

(Please share this for MTG Twitter and other social media)

I wish WOTC were capable of making a good bot these days. The majority of games I play are against Sparky since I am often in situations where I need to afk frequently and the AI is trash. They put so little effort into it and the 5 mono decks it can play. Half of its moves make no sense and the lack of a functional AI bot to play just drove most of the people I know who liked to play this while idle at work out of the game.

[anaris]

it's fundamentally not possible. a couple of recent papers have looked into this and the best systems available - various neural nets trained specifically on Magic, for example - still struggle to make simple evaluations like "which player is currently winning" that we all make constantly to steer our game plan *even when the ai has all the information possible including both decks and both hands*

[Doctor Zalgo]

anaris původně napsal:

it's fundamentally not possible. a couple of recent papers have looked into this and the best systems available - various neural nets trained specifically on Magic, for example - still struggle to make simple evaluations like "which player is currently winning" that we all make constantly to steer our game plan *even when the ai has all the information possible including both decks and both hands*

Both Claude and R1 with reasoning on are actually really solid at understanding game state and making moves (call it 'skilled amateur' level) and even limited fine tuning has really impressed me. [Ask me about baking a R1-style reasoning layer into Mixtral 8x7 to teach it the multi-mode reasoning necessary to build decks and play MtG if you have an hour to kill].

However. even if WotC did invest in the order of $200m +++ for a system capable of sustained medium-high level play for the number of concurrent users that they have (because that's how much it would cost just for the hardware alone), it would be the sort of project that would be ground breaking from a gaming and AI perspective and would generate global headlines and numerous research papers. We would all know about it because everyone would be telling us about it.

So, maybe its not bots with special decks, maybe its 'there's only a fixed number of deck archetypes and statistically you're going to run into all of them if you play long enough.

[anaris]

i absolutely have time for that, tell me more. I'm only familiar with it as a complexity study, which is mostly focused on solving, rather than in terms of "good enough"; how good is "skilled amateur"? Are misplays or missed opportunities more common fail states, and can you fine tune for either and which do you think "feels more significant" to the player? Does it build deck synergies/combos directly from text, or just by proxy of stuff like mana values and types or via netdeck associations, and are you using the full rules, and do you see novel play from it?

[Ha 何豪源]

one day the truth will come out, and then i want to see the reaction of the people who keep defending wotc. whats so bad about admitting it if they really use bots?

[anaris]

literally nothing lol that's why it's absurd that you think we're "defending" wizards when we say "you are imagining things"

[Kurt Angle's Neck]

Ha 何豪源 původně napsal:

one day the truth will come out, and then i want to see the reaction of the people who keep defending wotc. whats so bad about admitting it if they really use bots?

Still waiting for you to post a single quote of me saying anything positive about Wizards...

[Doctor Zalgo]

anaris původně napsal:

i absolutely have time for that, tell me more. I'm only familiar with it as a complexity study, which is mostly focused on solving, rather than in terms of "good enough"; how good is "skilled amateur"? Are misplays or missed opportunities more common fail states, and can you fine tune for either and which do you think "feels more significant" to the player? Does it build deck synergies/combos directly from text, or just by proxy of stuff like mana values and types or via netdeck associations, and are you using the full rules, and do you see novel play from it?

You can try it for yourself just by firing up deepseek and talking it through the first few turns of a game.

It is able to use the card text and abilities to put together a game play. However, in my experience, the main mistakes it seems to make relate to missing synergies on cards that require you to make an underlying decision. E.g. it will underrate a particular move because it misses an underlying beneficial interaction like 'I play X, declare creature rat and now all my rats get +1/+1'. Once you point that out, it's more than happy to step through why that's the best move and long term game implications, but that's definitely something that needs to be finetuned.

Back end databases with rules and cards also help keep it on track as it isn't familiar with more recent releases. (Although chatgpt 4.5 is current up to DFT).

[Doctor Zalgo]

Ha 何豪源 původně napsal:

one day the truth will come out, and then i want to see the reaction of the people who keep defending wotc. whats so bad about admitting it if they really use bots?

It's not bad that they use bots, it's that any bots they might use are still theoretical, or at least experimental. And also they'd cost a staggering amount of money to run compared to the forecast benefit (e.g. there's no benefit to them running an army of bots).

[anaris]

Doctor Zalgo původně napsal:

It is able to use the card text and abilities to put together a game play. However, in my experience, the main mistakes it seems to make relate to missing synergies on cards that require you to make an underlying decision. E.g. it will underrate a particular move because it misses an underlying beneficial interaction like 'I play X, declare creature rat and now all my rats get +1/+1'. Once you point that out, it's more than happy to step through why that's the best move and long term game implications, but that's definitely something that needs to be finetuned.

i have wondered if we could pull some of this stuff from Arena's card programming; like they've spent time coding the individual cards in ways that might be able to elucidate the functions for a deep learning model

[nul]

Win = skill
Lose = rigged

flawless logic. i get it, youre god's gift to gaming