AI Learns Insane Monopoly Strategies

@mil3k yes

it rubs the salad on its skin

@Christopher Pan I've tried this and it doesn't work? What am I doing wrong? 😕

@Z3DZ3R0 I've tried this and it doesn't work? What am I doing wrong?

I saw that but didn't understand what it meant lol I just thought it was part of the mumble jumbo on the screen at the time lol but what does it mean tho? anything ?

@b2studios need

Bro, in Monopoly, you can enable bots in local mode, and the bots can be hard to beat, the video kinda makes no sense since they already have it

You're a mad man

@Ajakr yesssss

Then download monopoly I play 3 AI plus me

@b2studios U know AI's are already in Monopoly

Try Diplomacy (the game) you will probably love it and they will either kill you for real or expel you or leave the house. No luck it's rules simpler than monopoly. Making deals then backstabbing is how you win frequently as coming up with superior orders every turn with the limited options really hard. Unlike most game several versions of cheating are allowed. One I love the most is if you can sneak a new army or navy piece on the board and no one catches you it stays.

@Ava I know that Elizabeth Maggie invented monopoly. In any case that's a pretty odd argument against land value taxation considering you can just do both (which is even more relevant of an idea now, when tech companies can make huge money with hardly any footprint, and that's ignoring day traders)

@Colby Boucher Elizabeth Maggie was the inventor of Monopoly. And she was somewhat outspoken against land value taxation as it didn't account for the value that human labor ads. Henry George was pretty much just an inspiration of hers. That would be like crediting Steve jobs with the creation of Apple silicon when he died before that idea was ever thrown around

@Ava *Sort of. It was a critique of land ownership, Henry George wasn't against capital itself per-se. He believed that by distributing wealth generated by the land itself equally (his idea was to do this by making a tax on the value of the land itself the only real tax), the relationship between labor and capital would become something more equitable. There's a theory that his ideas about the value of land were more or less buried by rich landowners paying economists to come up with more favorable explanations for things. Like, not in an intentional, conspiratorial way, but still. He was a fairly popular figure until his ideas mysteriously just faded away.

Yeah welcome to the club pal

Ai: Can't I get a mor fun job, like passing butter?

Hahaha love that Rick and morty reference

Not quite in the same vein but you might like salty bet. It's a 24/7 Livestream of bots fighting each other in a game called mugen

Or concurrently train an AI to commentate on the games.

Chess computers playing against each other is a big thing. Pretty cool to watch.

Golf clapping is a must

@AtomicCortex demon time.

Overtraining

@Jorge Rolando Chávez Mackay Actually, their strategies are not weird. Their strategies are superhuman. They are built on the exact same foundation we already know. However, chess is an insanely complex game. And humans are incapable of thinking in such complexity. Alpha zero has hundreds of years of experience, larger computational + memory abilities and is devoid of emotion. So it finds strategies that are completely understandable from human perspective... but a human would not find them due to lower capabilities.

You pay for the computers then lol

@Seth O'Bannion , then you learn how the enemy fights and keep doing so and keep being far more destructive than them so that way, you can never get conquered.

Tell the A.I. that.

@twizzm It is well documented to use this strategy, and why the number of houses left in the bin is a critical factor to decide whether to build some or not.

Doesn't work

@Awesomekk Me and my boys play monopoly plus on xbox like every week and this strat comes up literally every other game. The easiest set to win with is light blues, because psychologically nobody ever feels threatened by them, but they have the highest return for their cost regarding houses, so it ends up being that we trade properties so that each player gets a completed set and the player with the cheaper houses tends to win because they can more easily and safely build hotels, or easily drop down to 4 houses in order to block a new set from building them.

@Sir Wabaloo My Monopoly set has excess houses so no limit, but that is against the official rules that do limit the total.

@b2studios that's a fatal flaw in your whole simulation start over

Not to mention, the heavy percentage of chance (roll of the dice in every play)

Hey, has anyone built a computer that only answers 42 yet? If not, we better get on that.

@274h auctioning should be more common, most people believe you can simply refuse to buy a property and move on the rules clearly say if you dont want to buy a property (or cant buy) then an auction must take place it is not a choice

@Sawyer Beaton I've had more wins with orange. There are 16 chance cards. Which means you have a 1 in 16 chance of going there, or 6.25%, and that's it. Rolling from jail, even just visiting, puts you at a high chance of landing on orange or red, which is why I got for those 2 properties. And that's also not true, there are plenty of other "Advance to" cards as well.

Reds > Oranges because of the Advance to Illinois Ave Chance card makes it the highest landed on single property in the game + higher payouts than the oranges. Yellows are 3rd best and I've won many games with them. The rest of the sets don't matter. I buy them up only to trade for one of the big 3 and I always win if I can get one. Mortgage everything you have to get hotels on it and then try your best to sit in jail as long as possible. Then just wait and win.

Yeah 2nd railroad worse than the 1st one because of advance to go and other cards and doubles from jail cannot land on the 2nd railroad. Though the odds of a 5 being rolled are decently high if you pay to get out of jail with 7 being most likely followed by 6 and 8. Orange having a cheap housing cost and right next to jail has always been the strength of the orange tiles. With pairs of 3 and 4 landing on orange tiles from jail too, making up 33.33% of the pairs you can roll. Last row is either decent to great with blue + enough capital to get some investment in and reap the rewards or just garbage since the green color set would have to be a pair of tiles like the blue's to ever be worth investing into considering the huge housing cost.

@Aryan Toon Magenta is also pretty good.

The more specific the task, the better an AI is able to learn. It's when the task is vague and nuanced that an AI really struggles to develop a strategy or understand what its meant to do

​@b2studios I wonder if in the future humanely-general A.I would keep the same efficiency for the amount of games it can play though, I would assume it would be able to learn with less games like human would be able to learn to play monopoly decently after a single game. also, let's say you give it a task of calculating something. obviously computers do this way faster than humans, however if it just "uses it's computer" would that be the equivalent of a human pulling out a calculator?

​@JWCfive If you gave both sides 100 _minutes_ to learn a task, there's a good chance the AI might end up doing it better because it can use that time far more efficiently than a human could, simulating the task thousands of times in the time it takes a human to physically perform the task just once. However, if you gave both sides 100 _attempts_ to learn a task, there's a much better chance that the human will do it better, because both sides only get to simulate 100 practice sessions and a human will learn more explosively and make far more mental connections in 100 attempts than an AI would. The AIs improvement would be precise, but marginal after 1 attempt, but the human would eliminate a very wide range of possible attempts without testing them purely through prediction. Your claim was that an AI would defeat a human with "the same amount of experience", but 'amount' would imply 'number of attempts' as the metric and that statement only holds true for arbitrarily large numbers. AIs would defeat humans with the same _length_ of experience - excluding arbitrarily small numbers of time - because that brackets it under the constant of time, which is equal for both of them; the AI is simply more efficient in how it can use that time.

@Totally Nothatguy this is untrue. If a human had the same amount of experience as an AI did, the difference is between the extremely precise movements of an AI vs the almost certain variations in play of a human player Edit: actually I think I see what you are saying. But if a trained AI was put up against the best player, it would almost certainly win

@carlina taylor that’s because an AI can tirelessly run tons of games in a short time. If a person played a Mario course 100 times, and so did an AI, the human would win because 100 *attempts* isn’t enough to be better than humans.

lol hahaha nice

Nothing's random.

I called that the hand of God… It's very predominant when my friends and I play risk

You mean realistic

Glad you enjoyed it!

@Hannibal why did you return to this 3 weeks later? it's already out of my head. but imo, I wasn't right but I don't think you're right either, I'd like to think that the majority of fundamental opening theory came from humans, and man sidelines probally came from humans too, I do think that computer occasionally adds new moves to theory, but the more moves, the harder it is to calculate so it doesn't have as much impact as the already established theory that humans developed. If there're any examples of openings that was made from scratch by the computer, please tell me

@Angel Error Ive worked with a good handful of Grandmasters, including some with openings named after them and they’d disagree as well, but sure, if your definition of opening theory is mainlines only the so be it. Go tell all the Benoni players that they’re playing a novelty

@Phil Surtees 99.999% of these possible positions are either bad, really bad or plain stupid, as they lead either to a quick loss or a stalemate. The better ones, and more difficult to find are ones that still lead to a loss after a long and otherwise perfect game. It is quite fair to say that most of the **good** ones have been actually found; some might have not been found yet but it is still fair to say that out of these 70 trillion, it is going to be hard to find one that is actually good, and haven't been found yet.

@Phil Surtees bruh, this is well written. This explanation reminds me of the video I saw about the AI that mastered Go. It (the AI) did the same thing. It made moves “that made no sense” that had devastating effect to the human much later in the game. Many Go masters’ response was summed up to “it was like a human, but also like something else.”

I suppose opening strategy is about providing the most potential for board coverage and access by strategic pieces and optimal opening strategies can be discovered in that way more readily (at least in computational terms) as compared to a neural network. Having not looked into it before, I am curious now if the best opening strategies bear this min/max (achieving optimal “board coverage” in the least moves) principle out versus some other strategy.

That's why human meta dictates that you should buy three houses on properties (best return for cost) and never hotels. The fourth house is for draining the supply if you get to it.

@Mercy nope!

@prod savvy no just me being nice lol, does sound a bit botty reading it back

@b2studios The thumbnail seems to be a little bit sus.....

@b2studios but the question is: When?

@Matt Brown youre the dope that missed the "seemingly" aspect of the OPs comment😂 Also, great to hear that we have someone in the comments that appears to have mastered understanding of human consciousness, surprised you're here and not currently presenting your findings to the other top minds in the field...

Ai doesn't get bored, it's just randomly trying everything by brute force and when the right values go up, it uses that stuff a bit more often and then does that a jillion times. It's mindless brute force.

It could be that it needed to expel its accumulation, so that it could continue what it was programmed to do, play the game? I think once it had all it could get, it saw this as “not a great place if it wished to continue” and decided that selling/auctioning was the best way to get back to actually playing fully.

The AI realized that taking ridiculous and nonsensical decisions for the “fuck of it” payed off beyond belief. If you’re constantly throwing bizarre Hail Marys for no reason, it’s hard to defend against that.

Tha'ts so true hah.

@backelie ​The risk of being caught will not prevent people from trying the only option they can see. It is fine to reward people for working hard, we do not all need to have the exact same income. We should all have the same or similar chances. I wouldn't even have needed to gift my brother anything to prevent the robbery / tantrum. If I'd only let something go to auction and allow him to get it for the minimum price he'd have had a way out. All I needed to do is not bid. Or even lose some more money than I strictly needed by paying fair value for it. We could have continued playing and both have fun.

@Sheka88 That's a valuable life lesson; The less someone has to lose the less incentive to follow the rules. Or, put more succinctly: Inequality breeds crime.

Auctioning is usually bad for the simple reason that overpaying slightly for the ability to acquire properties outside your own turn is absolutely worth it, so it doesn't make sense to give your opponents that opportunity willingly.

@allen swanson Thank you

@Merennulli Parker Brothers, actually.

My thoughts exactly, felt clickbaited.

@b2studios Yes please! This was a great video and I definitely would love to see a follow-up. Subbed either way :)

@b2studios or you just payed someone to code the program for you

I buy everything I land on, but strive to get the orange properties.

Ahahhahah

If you are going to exploit human traits use the ability to not charge people.

Correct but because there are so few variables you can actually control in Monopoly, a theoretically optimal strategy wouldn't really have that much of a statistical advantage over a 'decent' strategy (humans using common sense for example)

@Nolan Armstrong AI is more efficient and humans have the capacity to make mistakes through their free will. Free will gives an advantage in creativity, in developing something entirely new whether it is art or philosophy. Although neural networks can simulate a new work of art, they cannot create a new form or style in the way humans like to do.

@Godowsky Godowsky but the problem is playing with humans the game is highly randomized. In the beginning of the game ppl tend to buy almost every property they land on because they have the cash on hand and the goal is to "monopolize" the board, obviously. That's why you have to think ahead multiple moves but it's also extremely risky to not acquire properties at the start of the match. Although I have found out playing conservatively and not super aggro is usually more efficient especially when playing against a wildly enthusiasticly buyer

@Zorro9129 I'd argue that free will in this circumstance would make the human less efficient than the ai. Humans also sometimes react on emotions and many variables enter playing even a simple game. For example if two humans are betting on a game I'd argue they're likely to play more conservatively. These variables don't exist with ai. Would love to hear your thoughts on this. Cheers

@Godowsky Godowsky I'd go with first semester students as I believe this is a simpler introduction they can understand and use as a template/tool going forward.

Yeah, not the most exciting ending ever. But it’s basically a science experiment, right? With hypothesis “this AI can significantly improve on typical human player strategy,” and negative results. As a young nerd-ling, I was taught that negative results are important and useful, but often under-appreciated. That description hit awfully close to home, so I resolved to give them the respect they deserve! Must’ve been more than [mumble] years ago, but ya know, I still have a soft spot for them 😂

Well I didn't really learn anything about monopoly I didn't know....I am really good at monopoly and win over 90 percent of the time and my wife has never beat me In 20 games over the years....she obviously doesn't enjoying playing as much anymore.....here are the real secrets to the game and I think most ppl do not play the auction style because that changes the game drastically along with certain house rules about free parking and how much goes in there......ppl that like to put 500 plus all the fines fees in free parking really make the game more luck based as it can be used to keep ppl in the game when thier on the verge of bankruptcy....number 1 rule when u start the game buy everything!!! At least first few turns intill your money starts to get low ...and how do you know your money is getting low?...you only need enough cash on hand to pay for the most costly rental places or cc or chance cards requirements that are currently on the board , so if no one has a monopoly yet you only need about 200 In cash to save and all that other money should be put I. Towards property........so moving on so your almost are broke but u got about 200 dollars and a few properties....u want to get a monopoly as fast as u can!!!! Whatever it takes !!!! And u wanna build fast!!!! You need to mortgage every other property and liquidate all your assets and build at least 3 houses on your monopoly....u want hotels but at least 3 houses or it isn't worth doing it. Obviously boardwalk park place are the beds property but honestly anything past the jail on the board is good enough with a hotel to bankrupt others, also the 2nd set of properties after each right turn are the better property the light blue the orange the yellow and and dark blue, houses cost the same so if you have a choice always go for the 2nd property. Also the brown properties are great to snag if you can because what they do is when your opponent goes past goes and lands there with a hotel all the money they collect will go to you and a Lil more. Now this is important because for some ppl that 200 is the only cash they have rolling in that gives them purchasing power , so taking that away with a hotel on the browns is a great strategy.

The end just plain sucked. I feel like I wasted my time.

Same here. I would've liked a chart for wins and looses and the best strategy for winning. Had trouble keeping up the fast paced speech as my superpower is limited to speaking Finnish fluently.

@Matt Brown Chess is a great use case for ML and Monopoly isn't. ML works so well for chess because the number of operations for brute analysis of a move is very big, and ML can find latent patterns to shortcut that. For Monopoly, the amount of analysis for what to do in a given turn isn't that much (or more specifically, the marginal increase of the objective diminishes quickly with the more computation you do) Which means that humans can and do perform the analysis themselves in a very short time and there isn't anything new ML can really bring to the table. I wouldn't say ML is infinitely better than brute force; there are best use cases for all types of problem solving.

10:17

Basically the bots started off playing the game like 8 year olds, then discovered that the point of the game was to prove that landlords are leaches. Nice

_(AIs get smashed with a hammer)_

I was looking for this comment too to see if anyone was as stupid as you in intentionally misinterpreting the obvious meaning. Do you really think that the video was saying what you accuse it of saying? Do you believe it's more likely that you alone figured this out, than the possibility that your interpretation is wrong?

@ZeroKitsune ​ @ZeroKitsune I was also expecting someone to have objected to this false statement. I stopped watching the video after seeing this. Programmers not knowing math? It's like a programmer who can't type - ridiculous. I suppose this will change when AI and speech recognition get better, and, as it is, most of the programming being done for business hardly requires more than a rudimentary grasp of even simple arithmetic.

I was looking for this comment. Thank you. It may be more likely to land on even spaces if you wait to roll doubles (because you only get out if you roll doubles) but it's completely wrong to imply that the chances of rolling doubles somehow increases by waiting in jail.

@Beardosa There's a plugin for that. 150K 👍 but only 4K 👎

It's more exciting, but unlikely. Humans aren't exactly dumb, so to be beaten by something that actually is VERY dumb, is pretty hard. AIs that beat humans usually do so by having an inherent "AI" advantage. Like being able to simulate millions of turns in seconds or just having wikipedia and thousands more websites memorized (to win Jeapordy) or just being able to consistently have several hundreds precise actions per minute (in videogames)... And they can even end up with minor flaws which simply didn't cause a losing problem but any human would quickly realize there is potential for optimization. I've seen an AI play Starcraft 2 and there is a number indicating the best number of units to gather resources. The AI always produced more. But it never lost because of it, so this trait survived.

@C4Oc And you'd lose all your friends only if you lose rather than if you win, which is another incentive to win the game.

@Seth Rose if you listened to the Deep Mind podcast, this isn’t the case. The IBM ones play like machine/alien, Alpha Zero ones actually play like very strong human player according to chess champions who played against them.

Yeah what a waste of time. Shame because I would've known instantly that the video wasn't about that if I could see the dislike button

@Ohrami ah i see. Its all about power, thanks!

@Jotarozee Orange is the strongest but realistically with good trading most properties are equally viable if the trades are balanced properly.

@Ohrami are there any properties that you prefer to get than others

@Jotarozee It's hard to define my entire strategy because it's extremely complex. Monopoly is a very complex game where every situation is unique. Generally, you want to use your actions to maximize your leverage and use your leverage in trade deals with other players to try to gain as much power within the game as possible to make it as likely as possible to win the game.

what is your go-to strategy?

I don't understand his "pay off debt to the bank" comment. Too many jokes make it hard to understand what's going on. Since you start with $1500 you can't buy anything for $3000, unless you can borrow from the bank?

I should release the results of training sometime...