Forum adverts like this one are shown to any user who is not logged in. Join us by filling out a tiny 3 field form and you will get your own, free, dakka user account which gives a good range of benefits to you: No adverts like this in the forums anymore. Times and dates in your local timezone. Full tracking of what you have read so you can skip to your first unread post, easily see what has changed since you last logged in, and easily see what is new at a glance. Email notifications for threads you want to watch closely. Being a part of the oldest wargaming community on the net. If you are already a member then feel free to login now.

 Mezmorki wrote:

My knowledge of military doctrine and terminology is all guesswork, so bear with me...

When I think of "tactics" or "A tactic" I think of it being a specific line of action in order to achieve a specific outcome in support achieving a specific strategic goal. In order to apply a tactic, one has to be confronted with a decision point, where the situational factors/context need to be assessed and a direction for action determined. This direction for action is usually coupled to a specific "tactic."

Let's take a hypothetical example for discussion.

An infantry squad has been tasked with securing a helicopter landing site in an urban plaza across the street with sporadic fire fights breaking in the area. This task is the squads objective or goal, which support some grander strategic need (landing site for additional troop deployments or whatever).

Approaching the street, the squad has a decision to make about how to get across the street, which is likely to expose them to hostile fire when crossing. What do they do? The tactical lines of action could include everyone grouping up and sprinting across the street in a dispersed formation (risky, but faster tactical move). Alternatively, they might leap-frog, with some squad members providing cover fire while others sprint across and then setup their own cover fire (slower, but safer).

The "tactics" under consideration are "dispersed sprint" versus "advance + cover" (or whatever these might be called).The squad leader has to decide what the "best" choice is, balancing the need for speed (can they take the slow safe option or do they only have 30 seconds to get to the drop zone?) versus the risk of taking incoming fire. Either way, the "tactic" is the action that is ultimately employed as a consequence of making a decision.

In reality the tactics available are much more nuanced though, and 40k's lack of nuance is the criticism here. That squad would most likely consider options such as suppressing the enemy while sending some men around the flank. Or maybe remaining stealthy and working their way around as a squad to get the drop on the entrenched enemy, perhaps after leaving a pair of spotters behind to feed them intel. These kind of suppression and positioning tactics are one of the main things 40k lacks. The only way to affect the enemy is to kill them. That's rarely how real world engagements play out.

In 40k terms your options are to stand on the objective because you know you're unlikely to die and can get the points for it or don't stand on the objective and shoot the enemy. There's no way to suppress them and advance under cover of your own fire, or break their morale to make their response less effective.

 Mezmorki wrote:

Even down to things like tri-pointing, or sequences for removing casualties, or setting up effective screens - these are all tactics to be employed in the game. And you're right, the basic toolbox of "tactics" that are allowed are a function of the rules of the game So the tactical tools don't change from game to game, nor would we expect them to. The tactical menu of options doesn't change in a game of chess either. Rather, our moves as players are taken to either set ourselves up to leverage certain tactics or deny tactical options to our opponent. That's the basic foundation of the game.

Tri-pointing isn't really a decision, though. It certainly isn't a meaningful one. This is the key thing I think you're missing. The indicator of depth in a game isn't just about how many total decisions you have to make, but the complexity of those decisions. Once I know how to tri-point there's rarely any reason not to do it and very little my opponent can do to stop me doing it. It's the equivalent of having a unit of Eradicators with an enemy character right in front of them being the only target they can see and calling the decision to shoot them a "tactical choice". Sure, on some ridiculously basic level it is but it's not a meaningful choice.

 Canadian 5th wrote:

You don't need much more than that and a flowchart to pilot a list in 40k so...

See this why this assertion is so effing stupid. Literally anything can be conceivably charted so it stands anything can be "flowcharted" ( chess ai has been around a loooong time ).

So sure there's an order of processing power for chess, but you'll find few games that are not subject to the same potential "flowchart fallacy".

At this point it's time for you guys to put your money where you mouth is, make a flowchart, and play a direct mirror match against a high level player.

But none of you will do that, because it'd actually be really difficult and you like making the assertion more than doing the work ( or actually playing the game).

On the subject of understanding why didn't you catch that the winning list is illegal? Do you even know why it's an illegal list?

Oh yea wicked game changah there.

Yes, that is how that unit should function on that table, where's the depth?

Look at the list. What gets deployed? What can take a hit? When can it take a hit? There's a lot of characters to protect .

And which meta was he busting? The one where he played a DG horde? Or hard-core blood angels? Or harlequins? Or DG Terminator spam? Which one specifically did he bust?

 Daedalus81 wrote:
See this why this assertion is so effing stupid. Literally anything can be conceivably charted so it stands anything can be "flowcharted" ( chess ai has been around a loooong time ).

Flowchart a game of MtG, or Texas Hold 'em, and you'll lose more than you win due to the nature of bluffing and hidden information.

So sure there's an order of processing power for chess, but you'll find few games that are not subject to the same potential "flowchart fallacy".

I've been arguing the chess is also shallow so...

At this point it's time for you guys to put your money where you mouth is, make a flowchart, and play a direct mirror match against a high level player.

This is the issue. I know 40k can be flow charted as most moves are simple, however, I lack the skill to make such a chart. This doesn't disprove anything however as I can neither program a Chess AI nor train a Go AI and yet both of those are possible and have been done such that they beat masters.

Oh yea wicked game changah there.

Not sure if sarcastic or not...

Look at the list. What gets deployed? What can take a hit? When can it take a hit? There's a lot of characters to protect.

Yes and...? That's all stuff that you can only say once you've sat across from your opponent and deployed against them. You'll have the same plan against pretty well all of them but will want to deploy differently for DG than for Harlequins.

And which meta was he busting? The one where he played a DG horde? Or hard-core blood angels? Or harlequins? Or DG Terminator spam? Which one specifically did he bust?

The tournament meta was what he busted and frankly looking at the lists of the attendees it looks like a lot of musty 8e lists getting dusted off as the US opens up a little.

Canadian 5th,

Are you saying that there shouldn't be Tactics articles? What is your tabletop experience with 40K 9th Edition? Are you being forced to write a tactics against your will?

I get the impression that you don't like 40K. That's cool. If you think that it has shallow tactics - OK. Not sure what your objection here is.

I said on the first page of this thread that good old Clausewitz wrote: "Everything is very simple in war, but the simplest thing is very difficult." There is friction is a tabletop game of 40K as well, whether its the dice, terrain, your opponent doing things you didn't anticipate, your own fatigue level, time pressures etc etc. Actual military tactics are fairly simple, although some types of operations have additional complexity of course. I am not saying that 40K is war. But its one thing to say that you have the game figured out, and its another thing to do it on the tabletop with the pressure of time and an actual opponent. I think its useful to have more than list-building advice on Dakka. Insights on how to quickly come up with a sound tactical plan and then make adjustments during the game would be useful, especially when couched in actual experience. Again, nobody is demanding that you write or read a tactics article.

People can chose to participate or not, and I think that Mezmorki's thoughts here are in line with what the Admins were asking for a little while back about the Tactics forum. I'm not saying that people should pay for 40K on-line tutoring, but I think having resources available on Dakka could be a good thing. I see opponents making tactical mistakes in tourneys. I also make bad decisions. The addition of time pressures (both real time and game turn constraints) and having an actual opponent makes it different from Theoryhammer. I think sharing best practices and painful hard-won experience is a good thing for the community.

TangoTwoBravo wrote:
Are you saying that there shouldn't be Tactics articles?

I'm saying that these tactics articles are often either rehashes of previous content or pure math hammer. Not the deep discussions that the OP for this thread had asked for. Do you dispute this?

I get the impression that you don't like 40K. That's cool.

I wouldn't own two armies and be here if I didn't like 40k. I just have major issues with 9th and its lack of depth.

I said on the first page of this thread that good old Clausewitz wrote: "Everything is very simple in war, but the simplest thing is very difficult."

That quote really doesn't hold for a simple wargame where the only variable exists within a toss of the dice.

There is friction is a tabletop game of 40K as well, whether its the dice, terrain, your opponent doing things you didn't anticipate, your own fatigue level, time pressures etc etc.

The only real friction in 40k is the dice and their results as applied to your list. Unlike a battle you always know what forces you have to hand, never run short of ammo, or get bad intel, so any mistakes you make are on you and only on you. That isn't depth that's just a puzzle game with dice being tossed.

Insights on how to quickly come up with a sound tactical plan and then make adjustments during the game would be useful, especially when couched in actual experience.

By all means please write them or repost the same from other sources. They won't be deep and that's my only argument.

How many games of 9th Ed have you played?

That you find 9th lacks depth is totally up to you! I won't try to convince you otherwise, because I am not sure what the point would be. I am not trying to convince you to play.

TangoTwoBravo wrote:
How many games of 9th Ed have you played?

You may as well ask how many games of chess I've played as I've said the same about chess as I have of 40k.

In my experience, game decisions are "deep" when there are (1) multiple layers of factors that must be considered' (2) where the ability to pick out the best answer is uncertain/obfuscated to an appropriate degree; and (3) where the results of decisions are consequential. If the margin between the outcomes of good and bad decisions is too narrow, then the game risks having trivial decisions (i.e. all options are nearly equally good or bad so it doesn't really matter what you do). If arriving at the obvious best move is perfectly clear with very little risk involved and not many factors to consider, then the game is one that potentially plays itself.

When it comes to assessing the thought-power needed to play a table-top game, I see three fundamental modes of thinking: spatial , psychological, and logistical thinking. Each of these can be an avenue for adding complexity, layers of factors, or uncertainty to the game. Spatial thinking relates to the complexity of physical board states and relationships between pieces, being able to envision future moves and board states, and mentally permutate through these different physical scenarios. Psychological thinking has to do consideration of your opponent's psychology and personality, their tendencies, your ability to bluff or persuade or negotiate, etc. And finally logistical thinking relates to operational planning, math, probabilities, risk evaluation, etc. We can talk about the overall balance between these three modes of thinking in a game, and we can also talk about the "weight" or level of importance that each one has.

40K is relatively light on the psychological side. There isn't much hidden information to work with or use for bluffing and the like. The spatial thinking demands of the game have shifted considerably over the years, and I'd contend that older editions that, for instance, didn't allow pre-measuring, including more nuance in model placement (vehicle facings), more impactful terrain rules, larger boards, and slower movement probably led to more complex spatial situations emerging during play. But even in 9th I don't think it's as trivial as it's made out to be. It may not be particularly deep, but not is it completely shallow. There is texture. On the logistical side - the execution of 40K is fairly heavy on this - and I concur that much of the decision making process can be viewed in terms of math and flowcharts. But I think that's easier to say in concept than in practice.

I agree with what others are saying (e.g. TangoTwoBravo) that even actual military "tactics" are relatively simple and straight forward. The tricky part lies in the decision of what tactic to employ. I don't think it's always as straight forward as others are claiming - and I think if one were to actually chart out the questions you're asking yourself as you work the mental flow chart, there is quite a bit that's being taken into consideration, even if the end result seems apparent and obvious. That it feels obvious, after having run through the flow chart, is perhaps more a testament to one's experience and knowledge of the game than it is a demonstration that the game lacks depth.

Building off of this, I think what others (e.g. Canadian 5th) are in favor of having is more tactical options available at any given decision point - whether that's more choices of actions to perform or a consequence of a different turn structure (or both!). If the choice of an action for a given unit is only to do A or B, then there is a limit to how much depth you can derive from it. But if they can do A or B or C or D, then those additional options compound across the whole game state and the game becomes "deeper" .... or it might just be more complex and difficult to parse (more illusion of choice to wade through).

This is getting pretty heady and abstract!

I've enjoyed the discussion, even though it's veered away from trying to list and discuss tactics for much of this thread. But it's been thought provoking and illuminating, which I appreciate as well.

I wonder, if summoning gets involved in a game, does that make the game more tactical?

Let's say I have 500 reinforcement points. What units I can/will bring in is "hidden information"?

It may not be optimal for a number of reasons but does it add an element of deeper tactics?

A little off topic, but is AoS, with a wider array of more viable summoning, a deeper tactical game?

Yet more off topic, Warcaster basically functions entirely on "summoning" in that you don't build a list, but gain a number of points with which to add a unit to the board from spawn points. Does having a system like that offer deeper tactics?

Or is it that there is still "optimal decisions" and if you have the skill to make those there's no real choices in summoning?

On the subject of flowcharting a game of 40k, can we all agree that with sufficient computing power that would could train an AI to play and win a game of 40k? If so, this proves that 40k can be boiled down to a series of logical decisions that can be performed by anybody given sufficient time. This is literally a flowchart with each branch contingent on the board state after each toss of the dice. If desired, we could also build a robot to house our AI and ensure that it can roll dice perfectly forcing us to use truly random elements for determining actions rather than pseudo-random ones such as thrown dice.

If we can agree to the above does that not prove that 40k is more shallow than as seemingly simple a game as Poker and much more simple than a game of MtG?

 Canadian 5th wrote:
On the subject of flowcharting a game of 40k, can we all agree that with sufficient computing power that would could train an AI to play and win a game of 40k? If so, this proves that 40k can be boiled down to a series of logical decisions that can be performed by anybody given sufficient time. This is literally a flowchart with each branch contingent on the board state after each toss of the dice. If desired, we could also build a robot to house our AI and ensure that it can roll dice perfectly forcing us to use truly random elements for determining actions rather than pseudo-random ones such as thrown dice.

If we can agree to the above does that not prove that 40k is more shallow than as seemingly simple a game as Poker and much more simple than a game of MtG?

Yeah if we don't understand what training an AI means, it's the same thing as a flowchart.

Like in chess when a computer rummages through a database to find the best move to play, totally the same thing as a flow chart.

Getting an AI to play the game is theoretically easy. Getting an AI to always do the mathematically optimal move - i.e. having an actually mathematically solved sequence of decisions - is quite another.

Realistically I don't think you can create a game an AI couldn't ultimately beat humans at - but that's because, in the case of 40k, it would be "playing" by processing huge amounts of geometry and statistics. Most (all?) humans are not. They might approximate to it due to some maths - and more often experience - but its very different form of "intelligence".

dhallnet wrote:
Like in chess when a computer rummages through a database to find the best move to play, totally the same thing as a flow chart.

That was the old way of building a gameplaying AI and it works well enough for Chess because we have a lot of data to work from. You could get the same results by letting a learning AI play millions of rated games on an online Chess game and this form of learning is what made Alpha Go work.

https://en.wikipedia.org/wiki/AlphaGo

Why can't this same approach be, theoretically, applied to 40k?


Automatically Appended Next Post:

Tyel wrote:
Getting an AI to play the game is theoretically easy. Getting an AI to always do the mathematically optimal move - i.e. having an actually mathematically solved sequence of decisions - is quite another.

Realistically I don't think you can create a game an AI couldn't ultimately beat humans at - but that's because, in the case of 40k, it would be "playing" by processing huge amounts of geometry and statistics. Most (all?) humans are not. They might approximate to it due to some maths - and more often experience - but its very different form of "intelligence".

If this isn't possible how does Alpha Go work?

 Canadian 5th wrote:
On the subject of flowcharting a game of 40k, can we all agree that with sufficient computing power that would could train an AI to play and win a game of 40k? If so, this proves that 40k can be boiled down to a series of logical decisions that can be performed by anybody given sufficient time. This is literally a flowchart with each branch contingent on the board state after each toss of the dice. If desired, we could also build a robot to house our AI and ensure that it can roll dice perfectly forcing us to use truly random elements for determining actions rather than pseudo-random ones such as thrown dice.

Yes

If we can agree to the above does that not prove that 40k is more shallow than as seemingly simple a game as Poker and much more simple than a game of MtG?

No

Reason #1: Because given sufficient computing power I think an AI could turn a game of poker or magic into an optimization exercise too, and likewise "solve it" and beat most players. As an example, Race for the Galaxy is as well known, and quite complex boardgame, with a good dose of double-think and reading your opponent. A programmer made a digital version using a neural network AI algorithm and it is absolutely brutal in its ability to win. That doesn't entirely prove the point - but I don't see how AI having to account for how a human might use or might not use one of the dozens of stratagems on hand in 40K is much different from an AI working through the logic of MtG. In terms of Poker an AI is going to have an huge advantage because they are never going to make a tell. Which leads to the next point...

Reason #2: Uncertainty and unpredictability are all factors that can be accounted for. Whether that's coming from the randomness of die rolls, the randomness of what of cards in magic you happen to draw, the randomness of what good/bad moves your opponent might dream up, or anything else. It's all math - which is what you've been saying the whole time.

If you want to prove that 40K is objectively shallower than poker or MtG, I don't think the argument you're laying out helps support your claim.

By the way, AlphaGo uses Montecarlo. I'm not o an expert, but that method doesn't lend itself to non discrete options like 40k movement.

 Canadian 5th wrote:

dhallnet wrote:
Like in chess when a computer rummages through a database to find the best move to play, totally the same thing as a flow chart.

That was the old way of building a gameplaying AI and it works well enough for Chess because we have a lot of data to work from. You could get the same results by letting a learning AI play millions of rated games on an online Chess game and this form of learning is what made Alpha Go work.

https://en.wikipedia.org/wiki/AlphaGo

Why can't this same approach be, theoretically, applied to 40k?

Yeah I know. Chess AIs can also use databases though (openings and final states).
It's not that the same approach can't be used, rather that a trained AI as nothing to do with a flow chart and its principles means that you can apply it to any game ever (or any task really, since that''s why they are researched).
So since now you know how what a trained AI is, why do you compare this to a flow chart ? It doesn't even use a flow chart.
While I'm sure it has already been debated here, Alphastar is an AI from the same company as AlphaZero & AlphaGO which plays Starcraft 2 on the ladder and is highly ranked (and they had to limit its abilities to make it fair). Is this also a "flow chart able" game ? Do you think it would be harder to make it play MtG ?

Almost any game can be "learned" by a computer nowadays.

 Canadian 5th wrote:
If this isn't possible how does Alpha Go work?

Sorry, maybe I was confusing.

1. I think you can create an AI to play 40k.
2. I think you could create an AI that ultimately wins most 40k games versus humans - or at least has a sufficiently high win percentage for us to say it is better.
3. I'm not however sure this proves 40k is simple, because I think you could do that for all such games.

Basically as per the post above.

FWIW, I think an AI's advantage in poker isn't not making tells - which are more "movie poker" than real life. Its being able to process and remember huge amounts of information beyond what most players can manage. I.E. What is the game state, how are people acting, what was the game state (at every stage), how did people act then, how have they tended to act in previous hands (while baring in mind behaviour can change) and so on. That's what professionals are doing - not staring into each others eyes and hoping their nose twitches, although it may happen from time to time. People obviously manage to play online with no faces or bodies to speak of.

 Mezmorki wrote:
Reason #1: Because given sufficient computing power I think an AI could turn a game of poker or magic into an optimization exercise too, and likewise "solve it" and beat most players.

They could only solve for so much of the meta in a game like MtG where top decks often make significant changes between major tournaments. It's possible that a very powerful AI could also account for how the meta is likely to shift and build for that as well, but that is more effort than it would take to perfect a game with as little hidden information as 40k. Likewise, they could probably learn to read tells in poker by doing things like reading heart rates, measuring temperatures, tracking eye movements, etc. but by the same token they could also potentially read marks on cards imperceptible to a human and 'cheat' at poker. I'd argue that needing these two extra steps makes the AI required to play these games more complex.

As an example, Race for the Galaxy is as well known, and quite complex boardgame, with a good dose of double-think and reading your opponent. A programmer made a digital version using a neural network AI algorithm and it is absolutely brutal in its ability to win.

If the deck is fixed from a known selection of cards this makes the AI's job far easier than in a deck-building game such as MtG. It may also be easier than poker, though it may also be that the level of play for a niche boardgame is lower than the skill level in professional poker players.

Reason #2: Uncertainty and unpredictability are all factors that can be accounted for. Whether that's coming from the randomness of die rolls, the randomness of what of cards in magic you happen to draw, the randomness of what good/bad moves your opponent might dream up, or anything else. It's all math - which is what you've been saying the whole time.

It might be possible that an AI could look at the pool of cards in any given MtG format, the established meta for said format, and its own ability to read an opponent and reduce MtG to a game with very few variables. However given the pool of played cards - let alone total cards - in any given format is vast I'd argue that learning to read a magic card and analyzing the card pool alone is already more difficult than solving a game of 40k.

 Canadian 5th wrote:

It might be possible that an AI could look at the pool of cards in any given MtG format, the established meta for said format, and its own ability to read an opponent and reduce MtG to a game with very few variables. However given the pool of played cards - let alone total cards - in any given format is vast I'd argue that learning to read a magic card and analyzing the card pool alone is already more difficult than solving a game of 40k.

To be simplistic : it doesn't need to analyse anything, it played millions of games in the time you would have troubles to play a hundred.
If you take a player that played MtG for decades, chances are he will beat someone who's been playing for a few months.

And solving is different than training AIs.

And to get back to the "flow chart" over simplification, I would like any of you try to make one just to move one single unit in a battle.

dhallnet wrote:
To be simplistic : it doesn't need to analyse anything, it played millions of games in the time you would have troubles to play a hundred.
If you take a player that played MtG for decades, chances are he will beat someone who's been playing for a few months.

And? Doesn't that just prove the superiority of a well-trained AI?

And solving is different than training AIs.

It is, but you could also build an AI to do that, it's just orders of magnitude harder and not within the scope of most current AI research.

 Canadian 5th wrote:

dhallnet wrote:
To be simplistic : it doesn't need to analyse anything, it played millions of games in the time you would have troubles to play a hundred.
If you take a player that played MtG for decades, chances are he will beat someone who's been playing for a few months.

And? Doesn't that just prove the superiority of a well-trained AI?

Yes but your initial posts tried to correlate the ability of this kind of AI to play/win a game with the simplicity of said game.
Which you can't, reason why you moved on and are trying to make me say something else.

it's just orders of magnitude harder

Yep, so much so that we aren't even sure that it is possible in the case of chess for example.

I don't understand your point;

Is it that Warhammer 40k could be solved by an "advanced enough" AI and so isn't that tactically complex?

Or is it that solving Poker/MtG would need an "advanced enough AI" to solve, and so these things are more tactical complex than 40k?

 Canadian 5th wrote:

TangoTwoBravo wrote:
How many games of 9th Ed have you played?

You may as well ask how many games of chess I've played as I've said the same about chess as I have of 40k.

(this answer means "Zero" kids)

 Canadian 5th wrote:

It might be possible that an AI could look at the pool of cards in any given MtG format, the established meta for said format, and its own ability to read an opponent and reduce MtG to a game with very few variables. However given the pool of played cards - let alone total cards - in any given format is vast I'd argue that learning to read a magic card and analyzing the card pool alone is already more difficult than solving a game of 40k.

So now you're comparing 1 game of 40K against the entire meta of magic? I think to be analogous here you'd need to compare across the entire design space - which if you're including deck construction and solving the meta for MtG you'd need to apply to list building and solving the meta for 40k.

 Yarium wrote:
I don't understand your point;

Is it that Warhammer 40k could be solved by an "advanced enough" AI and so isn't that tactically complex?

Or is it that solving Poker/MtG would need an "advanced enough AI" to solve, and so these things are more tactical complex than 40k?

You cant solve MTG. You can mathematically prove the best combinations and assume which deck has a better chance to win. However the card order in MTG (which is random) decides the winner more often than not. Magic is just as much about list construction as 40k is - if not more. I would say MTG requires a lot more knowledge about what cards are out there in the meta. 40k has a much smaller list of stratagems/gimicks that you need to know about to play against properly. Poker at the top levels is a very advanced game. A computer that mastered poker would probably be able to tell your exact hand just by your facial expressions - kind of like the pro poker players out there right now except it would be even better at it. A computer with the right set of sensors would be literally unbeatable at poker.


Automatically Appended Next Post:

 Mezmorki wrote:

 Canadian 5th wrote:

It might be possible that an AI could look at the pool of cards in any given MtG format, the established meta for said format, and its own ability to read an opponent and reduce MtG to a game with very few variables. However given the pool of played cards - let alone total cards - in any given format is vast I'd argue that learning to read a magic card and analyzing the card pool alone is already more difficult than solving a game of 40k.

So now you're comparing 1 game of 40K against the entire meta of magic? I think to be analogous here you'd need to compare across the entire design space - which if you're including deck construction and solving the meta for MtG you'd need to apply to list building and solving the meta for 40k.

MTG also has sideboards...kinda wish 40k did too.

 Xenomancers wrote:
MTG also has sideboards...kinda wish 40k did too.

Off-topic, but have you tried using Reinforcement Points? Not every army has access to it, but a few are able to bring in a "side board" of sorts. I've been tempted before by the idea of a Daemon army that uses summoning to replace whatever critical part of the army's design you managed to kill. Did you kill part of the army that's really good at killing, but left alone its objective-grabbers and support characters? Bring back the killy part. Did you kill the support characters? Return them to the field! They killed off your objective grabbers? Bring them back too! It could be a fascinating force.

 Xenomancers wrote:
...MTG also has sideboards...kinda wish 40k did too.

Warmachine had two-list tournaments (you bring two lists, your opponent brings two lists, and you get to read both your opponent's lists before deciding in secret which one to use). In the rules for Infinity you're technically allowed to build your list knowing what faction your opponent's playing and what the mission is, though I haven't been to any real tournaments so I'm not sure how that gets implemented in practice.

 Mezmorki wrote:
So now you're comparing 1 game of 40K against the entire meta of magic? I think to be analogous here you'd need to compare across the entire design space - which if you're including deck construction and solving the meta for MtG you'd need to apply to list building and solving the meta for 40k.

That should have read any given game of 40k. Also, given the starting premise of this thread, we could constrain an AI to only playing human-built decks/lists as that would be the purer proof of tactics in each game. Given this constraint, which game do you think is hard for an AI to solve for?


Automatically Appended Next Post:

dhallnet wrote:
Yes but your initial posts tried to correlate the ability of this kind of AI to play/win a game with the simplicity of said game.
Which you can't, reason why you moved on and are trying to make me say something else.

The answer lies within the complexity and computing power required to make such an AI. Tic-tac-toe is so easy a child can solve it, Connect 4 is more difficult but still solvable with highschool level math, Chess is a step above that requiring either a carefully curated move library and carefully programmed tactics or an early model self-learning AI, Go is a step above that requiring a modern cutting edge AI to play well. I would argue that 40k is closer to Chess level and that MtG is probably slightly above Go level but without any such AI existing I admit that I may well be missing a factor that makes things simpler or more complex for an AI in either of these games.

Yep, so much so that we aren't even sure that it is possible in the case of chess for example.

Chess is a finite game and thus there should be no uncertainty as to it being solvable, the question is how much computational power (or time) is required to brute force a solution and then how much more power is needed to brute force it within the time allotted for a sanctioned game. In either case, it is a question of hardware and not a logical issue.


Automatically Appended Next Post:

 Yarium wrote:
I don't understand your point;

Is it that Warhammer 40k could be solved by an "advanced enough" AI and so isn't that tactically complex?

Or is it that solving Poker/MtG would need an "advanced enough AI" to solve, and so these things are more tactical complex than 40k?

The point lies in how achievable such an AI is. We made a sufficiently advanced Chess game to beat any given human opponent in the '90s as a PR stunt. We made a Go and Jeopardy playing AI for similar reasons but it was orders of magnitude more difficult to do. I don't think 40k is unwinnable with current AI techniques or processing power, there's just no incentive for a company to do it.

I may have been too hasty to say solvable though. In my mind, I was picturing beat the best players consistently as solved but that isn't exactly solving the game. thus while I still think 40k is theoretically machine solvable it may be sufficiently complex as to be unrealistic in any plausible near future scenario.

 Xenomancers wrote:

 Daedalus81 wrote:

 Canadian 5th wrote:

The list is exactly what I said it was, a counter meta wonder that took a tournament by surprise.

That's likely a rudimentary understanding of the list. He made the incubi eligible for SfD so for all we know razor wings were baiting out Soulburst.

The d-cannon on the wraithseer is a monster considering ILOS. You wouldn't need to take hits with good use of terrain before getting the chance to cripple doomed eradicators or attack bikes.

List is also illegal isn't it? You cant take single razorwing units.

Thats debatable considering two sources were released at the same time with different unit sizes and historically razorwings have been 1-12.

And even if it's illegal, that doesnt invalidate the whole list, its not like these 36 points singlehandedly won him the game.


Automatically Appended Next Post:

 AnomanderRake wrote:

 Xenomancers wrote:
...MTG also has sideboards...kinda wish 40k did too.

Warmachine had two-list tournaments (you bring two lists, your opponent brings two lists, and you get to read both your opponent's lists before deciding in secret which one to use). In the rules for Infinity you're technically allowed to build your list knowing what faction your opponent's playing and what the mission is, though I haven't been to any real tournaments so I'm not sure how that gets implemented in practice.

Infinity tournaments tell you in advance what the missions will be and you get to build 2 lists (from the same faction) and decide once you hit the table which list you will play.

 VladimirHerzog wrote:
Thats debatable considering two sources were released at the same time with different unit sizes and historically razorwings have been 1-12.

Given that one source was for PL only I don't think there's much debate to be had.

And even if it's illegal, that doesnt invalidate the whole list, its not like these 36 points singlehandedly won him the game.

It's not the points or their killing power that matters. If you force those 36 points into a single unit it likely changes the entire gameplan as he now needs to position a more valuable unit to score that same table quarter or out-of-the-way objective. If the 3 model minimum size is enforced I doubt we see this list ever again.