I hope the game ships with better AI than we've seen so far

[President Jyrgunkarrd]

Well, that's the thing - how is a player (or in this case, a viewer) supposed to know what's a bug vs what is simply shackled behavior, given that one can so easily resemble another?

It would be very clear cut if there were no shackling and the AI was always doing the best job it could to kill you. Since it isn't, it's impossible to know at face value whether the Hunch (as an example) was doing that because it didn't know what to do or if it did that because the AI saw that the player was in trouble and threw them a bone.

 

[Havamal]

Well, that's the thing - how is a player (or in this case, a viewer) supposed to know what's a bug vs what is simply shackled behavior, given that one can so easily resemble another?

It would be very clear cut if there were no shackling and the AI was always doing the best job it could to kill you. Since it isn't, it's impossible to know at face value whether the Hunch (as an example) was doing that because it didn't know what to do or if it did that because the AI saw that the player was in trouble and threw them a bone.

I think that's easily explained by reading the previous posts of those who already understand how this AI works, one only needs to read them to know that it's often heavily tied to weights and triggers rather than "shackling", and if the weights and triggers are still being adjusted for balance, and if something is off it might be an unintended consequence. If its a map pathing issue due to the individual map, which has certainly occurred before that's also another example. Just by being constructive rather than hyperbolic we can raise the usefulness of dialogue to address such issues rather than just lead to friction.

 

[Havamal]

...because the AI saw that the player was in trouble and threw them a bone.

This would require a rather advanced AI for our purposes to recognize when to give up the game in that manner. I wouldnt' expect that behavior even in the hardest AI.

 

[MagnusEffect]

I find it funny that someone posted the link to the actual variables/weights that were already moddable in the beta several pages ago and yet people are still arguing about whether or not the AI can reserve, whether or not the AI will be moddable, etc. This is why we can't have nice things.

^This.

So yeah. People need to relax. If you are looking for a challenge, the game is capable of it in spades.

 

[Varkin]

It would be very clear cut if there were no shackling and the AI was always doing the best job it could to kill you. Since it isn't, it's impossible to know at face value whether the Hunch (as an example) was doing that because it didn't know what to do or if it did that because the AI saw that the player was in trouble and threw them a bone.

How exactly would you know that the AI was doing it's best job to kill you? Long term extremely skilled vet players of battletech very often disagree on what action should be taken in a turn and the tactics employed in a battle. Just because the AI has a different "opinion" of the next best course of action for success than someone else does doesn't mean it is shackled.

 

[Havamal]

Long term extremely skilled vet players of battletech very often disagree on what action should be taken .

People in this very forum disagree on everything too of course.
One group says PPCs are useless and need a buff. The other takes only PPC based mech and wreck faces.
One group Says LLasers are pointless and the AC10 OP, but then several more swear by a LLaser version of the Panther, Griffin, and Centurion .
It's a house rule IP where no two people agree on what the identity of a single mech is and even getting people to just agree to disagree on the color of the sky is hard work for forum staff.

Everyone should be respecting each other though.

 

[Chaon]

I think that's easily explained by reading the previous posts of those who already understand how this AI works, one only needs to read them to know that it's often heavily tied to weights and triggers rather than "shackling", and if the weights and triggers are still being adjusted for balance, and if something is off it might be an unintended consequence. If its a map pathing issue due to the individual map, which has certainly occurred before that's also another example. Just by being constructive rather than hyperbolic we can raise the usefulness of dialogue to address such issues rather than just lead to friction.

Even our own behaviour when playing is tied into weights and triggers. 'Do I jump my mech to there and alpha strike knowing I might not kill it and have to deal with return fire? Or do I play defensive and let it come towards me so I can move to gang up on it in two more turns?' type thing.

The main difference between a human played and an AI is that the human player can change their weights and triggers at will from turn to turn, and decide to do totally different things in the exact same situation. Which is why an AI will never be as good as a human player.

And that is what some players are forgetting or not realising. That the AI will always be bad when compared to a human player. It doesn't matter how much they work on it and improve it, the AI will always be at a disadvantage to a human player.

But that's not to say that it can't be a challenge to the average player. Because we will make mistakes and the AI will take advantage of them.

 

[Havamal]

Even our own behaviour when playing is tied into weights and triggers. 'Do I jump my mech to there and alpha strike knowing I might not kill it and have to deal with return fire? Or do I play defensive and let it come towards me so I can move to gang up on it in two more turns?' type thing.

The main difference between a human played and an AI is that the human player can change their weights and triggers at will from turn to turn, and decide to do totally different things in the exact same situation. Which is why an AI will never be as good as a human player.

And that is what some players are forgetting or not realising. That the AI will always be bad when compared to a human player. It doesn't matter how much they work on it and improve it, the AI will always be at a disadvantage to a human player.

But that's not to say that it can't be a challenge to the average player. Because we will make mistakes and the AI will take advantage of them.

:clapping:
Yes, this.

Sigh. It's a complex game. The AI was always going to be the biggest challenge. Expectations being the second biggest.

My hat really goes off to them for what they were able to do so far.

 

[Zythen1975]

One thing that should be a very easy change is turning off that shoddy / ramshackle armor. When we have to do 2 3 times the damage for the same result. That alone should make things more challenging even if the ai is only mediocre.

 

[TaurianMerc]

Bake a pie, eat a pie.

no, no, you've got that wrong - Bake a pie, give TaurianMerc a pie

Have people taken Havamal's advice and decided to agree to disagree on what an "adequate AI" should be?
As he finally done the impossible?

 

[President Jyrgunkarrd]

How exactly would you know that the AI was doing it's best job to kill you? Long term extremely skilled vet players of battletech very often disagree on what action should be taken in a turn and the tactics employed in a battle. Just because the AI has a different "opinion" of the next best course of action for success than someone else does doesn't mean it is shackled.

Note that just because there are differences of opinion about what is optimal or what is the winning strategy for a high pressure situation, that doesn't mean there aren't actions that are just bad. Turning your rear armor to someone and passing your turn is just bad; it is very unlikely anyone come come-up with a good argument for why you should do that if your objective is to win the game. Likewise with never shooting a PPC you put into your mech while sitting at 0 heat, or passing a turn on your Hunchback after walking into firing range. There aren't contentious decisions with a balanced pro / con tree.

If you're sitting in front on an enemy mech near the end of a match and you're both down to breached armor locations, and you can alpha with your remaining weapons but it will shut you down (and might even kill you) OR you can melee but the melee strike is much less likely to outright CT your opponent compared to the alpha, that's a real tough choice to make and something players could argue about all day long with nobody necessarily being right.

Parking your Cat to die is not in the same ballpark of contentiousness, and pretending it is to excuse the AI is pretty dishonest. :|

 

[Tankqull]

Parking your Cat to die is not in the same ballpark of contentiousness, and pretending it is to excuse the AI is pretty dishonest. :|

this.

 

[FlavourBeans]

I just started hanging around this board after having just stared longingly at this game from afar, and I was beginning to worry that it wasn't actually a strategy game at all. But now I'm confident it must be, because it ain't a proper strategy game without the traditional pre-release AI fight thread.

Somewhere on the internet, there's a thread about a tic-tac-toe game's AI, where one person dislikes it because the AI isn't good enough to beat them and the game always ends in a draw, while someone else is mad that the AI is too good and doesn't make enough 'human' errors.

 

[President Jyrgunkarrd]

I just started hanging around this board after having just stared longingly at this game from afar, and I was beginning to worry that it wasn't actually a strategy game at all. But now I'm confident it must be, because it ain't a proper strategy game without the traditional pre-release AI fight thread.

Somewhere on the internet, there's a thread about a tic-tac-toe game's AI, where one person dislikes it because the AI isn't good enough to beat them and the game always ends in a draw, while someone else is mad that the AI is too good and doesn't make enough 'human' errors.

Yeah, sure. Except that at least in the hypothetical tic-tac-toe game's case, people are discussing one coherent thing (whether or not the AI should be shackled), whereas the BattleTech AI is being offered every excuse. It's buggy... no, no, it's intentionally being shackled because otherwise the game wouldn't be fun... no, no, the decisions it is making are legit & correct in their own respect.

So it's a disingenuous argument because people are just tossing whatever excuse suits their fancy into the mix instead of actually defending a consistent position.


It's also fallacious to claim that because argument has competing parties, each party must have an equal share of merit.

 

[Cyttorak001]

Yeah, sure. Except that at least in the hypothetical tic-tac-toe game's case, people are discussing one coherent thing (whether or not the AI should be shackled), whereas the BattleTech AI is being offered every excuse. It's buggy... no, no, it's intentionally being shackled because otherwise the game wouldn't be fun... no, no, the decisions it is making are legit & correct in their own respect.

So it's a disingenuous argument because people are just tossing whatever excuse suits their fancy into the mix instead of actually defending a consistent position.


It's also fallacious to claim that because argument has competing parties, each party must have an equal share of merit.

It's also not especially brave to adopt a "pox on both your houses" stance as a way of avoiding conflict. There are real issues at stake here, and they matter greatly.
With all the multiplayer options that have been cut from this game, the single-player experience is now more important than ever, and a huge portion of that depends on the quality of the AI. If the AI is bad, it will impact the game reviews and then the sales.

You are correct.

 

[Tangman]

Paradox Interactive stream just said they are streaming Release Candidate 2, which "should?" be newer than the stuff others played? Tune in and check it!

 

[Exemplar Voss]

How exactly would you know that the AI was doing it's best job to kill you? Long term extremely skilled vet players of battletech very often disagree on what action should be taken in a turn and the tactics employed in a battle. Just because the AI has a different "opinion" of the next best course of action for success than someone else does doesn't mean it is shackled.

Well, one solid clue (to trying to kill) is enemy heavies shooting at the player rather than turning their backs and doing nothing.

It isn't an opinion that we've seen some demonstrably wrong behavior from the AI. Not 'shackled' but flat out @wrong@

 

[Tovarisc]

Paradox Interactive stream just said they are streaming Release Candidate 2, which "should?" be newer than the stuff others played? Tune in and check it!

Yep, they said it's more polished build than the one streamers have access to.

Tutorial seems to be just as none existing as before, shame.

 

[me987654]

The AI will likely take some tuning post release. An AI for a game like this is very complex and difficult to implement.

I reviewed some of the variables that were moddable in the beta and not only can the AI weighting be modded, but the pathing choices as well. We are honestly really lucky that HBS is exposing this stuff as many games do not and you are stuck with whatever hardcoded AI they give you.

I'm not surprised to see the AI sometimes doing strange things; the tactical layer of this game is very deep and complex (as an example - it's much, much deeper than XCOM or XCOM 2) AND to make it every more difficult the AI has to be able to use the same units as the player (it's generally much easier to build an AI for unit types that only the AI uses). There's also going to be some bugs (newsflash - there are bugs in just about every single piece of software ever written )

The only thing that would irritate me personally would be if HBS doesn't continue to tweak and refine the AI after release (particularly if the game is as successful as it looks like it might be).

As an aside - having terrible AI doesn't seem to impact sales that much. As an example, many total war games (in particular empire total war) have mostly or completely broken AIs yet they sell tons of copies (and this goes for quite a few popular strategy games).

 

[kvetcha]

As an aside - having terrible AI doesn't seem to impact sales that much. As an example, many total war games (in particular empire total war) have mostly or completely broken AIs yet they sell tons of copies (and this goes for quite a few popular strategy games).

Civ-il-i-zation.

(I adore Civ.)