Sparky Bot

[Imnuktam]

Please put more effort into this bot, she could be a lot of fun if you cared to work on her. Older magic games had great AI and a huge variety of decks to play against. Where did all the guys that coded those go?

Probably you intentionally put little effort into it thinking you are forcing people to play online multi matches instead, but with the majority of people who play that bot, that is not the case.

I have a buddy who owns an arcade, another who works at a hardware store and another at a hobby shop. All of them like myself would love to play this at work during downtime but will need to jump up and dont want to forfeit every single time because of it or get timed out. Playing those same 5 mono colored easy mode decks gets old and then instead of logging on and playing and maybe buying more cards, we just do something else.

[Izzet Firemind]

If you ask a lot of people on this forum, they play against dozens of bots every day. I don't believe that, but I have no proof, and I'm not here to spread misinformation.
I heard that it's almost impossible to program a decent AI into this game. The only reason it was able to be worked in previous games seems to be the limited card pool. Each deck had to be programmed individually and the AI itself was actually a very simple program.
I asked Google AI why AI has such a hard time with MTG. I figured if anyone would know, it would be another AI. Here is what it said:

Magic: The Gathering (MTG) struggles to have a truly "good" AI because of its immense complexity, with a vast number of cards, potential deck combinations, and intricate strategic possibilities, making it nearly impossible for an AI to calculate all possible outcomes and consistently make optimal decisions in every situation; essentially, the game's complexity approaches a theoretical limit where even the most advanced AI would still struggle to play perfectly.
Key reasons why MTG AI is challenging:
Huge state space:
The sheer number of possible board states, card combinations, and potential plays creates a massive space for an AI to analyze, making it computationally expensive to evaluate all options.
Imperfect information:
Players don't know their opponent's hand, creating uncertainty and requiring strategic decision-making based on limited information.
Delayed rewards:
It can take many turns to win a game in MTG, making it difficult for an AI to learn from its actions as the consequences of a decision might not be immediately apparent.
Strategic depth:
The game encourages complex decision-making, including bluffing, resource management, and understanding the opponent's strategy, which are hard for AI to replicate at a high level.
Card interactions:
With thousands of unique cards, each with its own abilities and interactions, an AI needs to understand the nuanced play patterns of every card to make informed decisions.
While current MTG AI can play the game, it often makes sub-optimal choices due to these complexities, resulting in an experience that can feel less challenging or strategic compared to playing against a human opponent.

[Minmataro]

Sparky is there for brand new players and to test a brand new deck brew just to see if your mana curve is ok. Why would you use it to play magic normally? Just use casual queue.

[Izzet Firemind]

Originally posted by Minmataro:

Sparky is there for brand new players and to test a brand new deck brew just to see if your mana curve is ok. Why would you use it to play magic normally? Just use casual queue.

I get why you would want to play the bots. The OP's example is at work where he would have to AFK a lot and it would rope him in a normal game. It's also a good way to hone in new game mechanics and get things down to memory so you don't miss upkeep triggers and the such before you take a deck to FNM.

[Imnuktam]

Originally posted by Izzet Firemind:

If you ask a lot of people on this forum, they play against dozens of bots every day. I don't believe that, but I have no proof, and I'm not here to spread misinformation.
I heard that it's almost impossible to program a decent AI into this game. The only reason it was able to be worked in previous games seems to be the limited card pool. Each deck had to be programmed individually and the AI itself was actually a very simple program.

I wouldnt even mind if the bot didnt play its decks perfectly, just so long as it had a lot more for sake of variety. Making it capable of playing all the starter decks would be a good first step.

Games like magic 2015 (10 year old game) that actually put effort into its bot had so many more to play against and so many of them were better than what is provided in Arena, and it played them all much better.

With only 5 mono color decks, it honestly just feels like they just dont want to because they think it will funnel everyone into online multi and cut down que times if people have no other choice.

[Izzet Firemind]

Originally posted by Imnuktam:

Games like magic 2015 (10 year old game) that actually put effort into its bot had so many more to play against and so many of them were better than what is provided in Arena, and it played them all much better.

That's why I made mention of card pool. That game had a limited set of cards, so the decks could be better balanced. It wasn't a matter of better AI, it was card pool tweaking.

Edit: From what I can tell there are 300 unique cards in that game. That's an absolutely tiny card pool when compared to MTG as a whole.

[Tom the Mime Artist]

You could make a competent AI for certain archetypes relatively easily and for other archetypes it'd be a nightmare. The less a deck cares about what its opponent is trying to do, the easier is it to get a bot to play it decently. That also makes it less interesting to play against a bot though.

For a lower tier burn deck, against most decks it's basically:
Can I cast a burn spell? If yes, do they have a creature with lifelink? If yes, target the creature, if no, target their face. It still won't play it optimally and would make bad calls in mirror matches. When the deck starts caring about what the opponent is doing, that's where the effect of the larger card pool kicks in.

[Imnuktam]

Originally posted by Izzet Firemind:

Originally posted by Imnuktam:

Games like magic 2015 (10 year old game) that actually put effort into its bot had so many more to play against and so many of them were better than what is provided in Arena, and it played them all much better.

That's why I made mention of card pool. That game had a limited set of cards, so the decks could be better balanced. It wasn't a matter of better AI, it was card pool tweaking.

Edit: From what I can tell there are 300 unique cards in that game. That's an absolutely tiny card pool when compared to MTG as a whole.

There are a definitely a huge amount of cards released on arena compared to 2015. That said, this game has made at least 100x more money, just a guess but I know Arena has generated over 500 million some years. It has probably generated somewhere around 2-3 BIllion so far.

Wish they would use some of that to hire a lot more coders to make sparky work. She doesnt need to know all 15,000 or play them all perfectly, but they could surely (IF they really wanted to) code arenas sparky to effectively play 1000 which is just a little more that 3x the 300 2015 had.

Honestly just feel they dont care. 2015 had dozens of decks, this sparky has 5 really basic mono colored decks and she plays them more poorly than 2015 played theirs.

[Zlehtnoba]

Yes, they do not want to make a better Sparky. It's just a tool to teach beginners some basics, not a way to pass time as OP wishes for.

This is sound economic thinking, beginners then go to play in PvP, and start spending. Which is not inherently a bad thing, Arena needs to make money. Those of us who enjoy Magic know that it's a very expensive hobby. Still, since the change in management at Hasbro, it feels the company is more about making money fast than about sustainability and player enjoyment. We have power creep, a boring Standard, and release after release after release ... Tons of ways to play, if of course you are ready to spend gems to get the cards.

[Imnuktam]

Originally posted by Zlehtnoba:

Yes, they do not want to make a better Sparky. It's just a tool to teach beginners some basics, not a way to pass time as OP wishes for.

This is sound economic thinking, beginners then go to play in PvP, and start spending. Which is not inherently a bad thing, Arena needs to make money. Those of us who enjoy Magic know that it's a very expensive hobby. Still, since the change in management at Hasbro, it feels the company is more about making money fast than about sustainability and player enjoyment. We have power creep, a boring Standard, and release after release after release ... Tons of ways to play, if of course you are ready to spend gems to get the cards.

Is it really sound economic thinking though? Two of the 5 friends I have that used to play this a lot to 'pass time' simply stopped playing because they mainly played the bot and grew tired of playing against the same 5 decks forever. They now spend their time and money in a different game and spend 0$ a month here compared to the 50-100 $ a month or more they used to spend.

They would also pvp here and there when their schedules permitted and cut down que times of the rest of Arenas player base, but now they do neither of those.

Devs should want to make a better sparky for everyone if making more money and having more players is the driving force behind their decisions.

[Zlehtnoba]

Originally posted by Imnuktam:

Originally posted by Zlehtnoba:

Yes, they do not want to make a better Sparky. It's just a tool to teach beginners some basics, not a way to pass time as OP wishes for.

This is sound economic thinking, beginners then go to play in PvP, and start spending. Which is not inherently a bad thing, Arena needs to make money. Those of us who enjoy Magic know that it's a very expensive hobby. Still, since the change in management at Hasbro, it feels the company is more about making money fast than about sustainability and player enjoyment. We have power creep, a boring Standard, and release after release after release ... Tons of ways to play, if of course you are ready to spend gems to get the cards.

Is it really sound economic thinking though? Two of the 5 friends I have that used to play this a lot to 'pass time' simply stopped playing because they mainly played the bot and grew tired of playing against the same 5 decks forever. They now spend their time and money in a different game and spend 0$ a month here compared to the 50-100 $ a month or more they used to spend.

They would also pvp here and there when their schedules permitted and cut down que times of the rest of Arenas player base, but now they do neither of those.

Devs should want to make a better sparky for everyone if making more money and having more players is the driving force behind their decisions.

I'm not sure how typical your friends are. If playing in free time pvp is not interesting enough for them, they probably are not the target audience. I find that not playing when I can't dedicate, say, half an hour to Arena, makes me more interested in playing when I do have the time.

As an aside, a thing you can do is play bot draft. You can stop drafting in the middle of the draft, and come back later to continue where you left off. A draft game in best-of-one should not take more than ten minutes either. I hesitate to mention this because drafting can be expensive if you can't average three to four wins per run.

[Imnuktam]

Originally posted by Zlehtnoba:

Originally posted by Imnuktam:

Is it really sound economic thinking though? Two of the 5 friends I have that used to play this a lot to 'pass time' simply stopped playing because they mainly played the bot and grew tired of playing against the same 5 decks forever. They now spend their time and money in a different game and spend 0$ a month here compared to the 50-100 $ a month or more they used to spend.

They would also pvp here and there when their schedules permitted and cut down que times of the rest of Arenas player base, but now they do neither of those.

Devs should want to make a better sparky for everyone if making more money and having more players is the driving force behind their decisions.

I'm not sure how typical your friends are. If playing in free time pvp is not interesting enough for them, they probably are not the target audience. I find that not playing when I can't dedicate, say, half an hour to Arena, makes me more interested in playing when I do have the time.

As an aside, a thing you can do is play bot draft. You can stop drafting in the middle of the draft, and come back later to continue where you left off. A draft game in best-of-one should not take more than ten minutes either. I hesitate to mention this because drafting can be expensive if you can't average three to four wins per run.

Yeah I am not sure how much of the population is like the people I know, but 100% of the people I know who play arena would love that bot to have more decks and be given more love.

All of them like to play in situations like jobs with a lot of free time but may have to jump up at any second and deal with something that will make them forfeit every game if against real people. Sadly the majority of the ones I knew that used to play have all quit and moved on to something else.

WOTC probably has no idea how much they are losing every month from people like that. My buddy that owns the arcade used to play it at work all the time with it up on the 60" screen on the wall for all to see giving free advertising and sucking other people into the game. He can only play the bot with how often he will be needed to do something else and after a while he just got sick of playing against those same 5 decks. Now it is usually just a random fighting game on when I go down there and he never logs in anymore.

[Corey Trevor]

Originally posted by Izzet Firemind:

If you ask a lot of people on this forum, they play against dozens of bots every day. I don't believe that, but I have no proof, and I'm not here to spread misinformation.
I heard that it's almost impossible to program a decent AI into this game. The only reason it was able to be worked in previous games seems to be the limited card pool. Each deck had to be programmed individually and the AI itself was actually a very simple program.
I asked Google AI why AI has such a hard time with MTG. I figured if anyone would know, it would be another AI. Here is what it said:

Magic: The Gathering (MTG) struggles to have a truly "good" AI because of its immense complexity, with a vast number of cards, potential deck combinations, and intricate strategic possibilities, making it nearly impossible for an AI to calculate all possible outcomes and consistently make optimal decisions in every situation; essentially, the game's complexity approaches a theoretical limit where even the most advanced AI would still struggle to play perfectly.
Key reasons why MTG AI is challenging:
Huge state space:
The sheer number of possible board states, card combinations, and potential plays creates a massive space for an AI to analyze, making it computationally expensive to evaluate all options.
Imperfect information:
Players don't know their opponent's hand, creating uncertainty and requiring strategic decision-making based on limited information.
Delayed rewards:
It can take many turns to win a game in MTG, making it difficult for an AI to learn from its actions as the consequences of a decision might not be immediately apparent.
Strategic depth:
The game encourages complex decision-making, including bluffing, resource management, and understanding the opponent's strategy, which are hard for AI to replicate at a high level.
Card interactions:
With thousands of unique cards, each with its own abilities and interactions, an AI needs to understand the nuanced play patterns of every card to make informed decisions.
While current MTG AI can play the game, it often makes sub-optimal choices due to these complexities, resulting in an experience that can feel less challenging or strategic compared to playing against a human opponent.

They are easy to spot because they usually have a name like Dave or John, have zero cosmetics, very basic decks. I have seen a few that have a companion or a basic card sleeve but rarely. I will take a screenshot next time I play one. I have never seen them in competitive though that I can remember; I think they don't use the unranked bots for sparky simply because they want you to play pvp. The unranked bots play a little bit like sparky though; they can get confused and do really dumb things like trying to play something against hexproof then directing it at their own creature when they can't. They are very basic.

[Zlehtnoba]

Did you consider these players may be new to Magic? Everything you mention, except maybe the usernames, points to that. But yes, it's also how a basic bot would look like.

[Corey Trevor]

Originally posted by Zlehtnoba:

Did you consider these players may be new to Magic? Everything you mention, except maybe the usernames, points to that. But yes, it's also how a basic bot would look like.

Bro nobody with a name and no numbers in their name and a generic name to boot. I have seen some that have names like cannon etc they aren't all just personal names. I have some old screenshots of it one is Mario I think. They will have some semi complex or recent cards, but they usually have zero synergy to their deck as well. One I played which I think was Mario had the one ring in the deck and nothing that really went with it; just basic ass cards. If thats a human I will play a red blue deck with no draw cards or removal for the rest of my life.

[Corey Trevor]

Found an old screenshot of a unranked bot

https://steamcommunity.com/sharedfiles/filedetails/?id=3420889632