44 minutes ago, Azrapse said:

That's why both HOTAC and this came with the concept of waves or reinforcements. Nobody is expecting the AI to actually be good. They are expecting the AI to be overwhelming enough to suppose a challenge.

^^This

If you look at some decent solo/co-op games like Descent, Frostgrave/Rangers of the Shadowdeep, Mansions of Madness, etc., that have their origins as competitive games, its not about attempting to out-think the enemy, anymore. Its about accomplishing some task/objective while being efficient and surviving waves of enemies/attrition. No one should expect the AI opponents to be as smart or as challenging as another player, that's just not going to happen.

This Alpha version of the AI that FFG has released is working on the waves part. The objectives will likely come later.

So iteration 2, same setup, but only 1 child wanted to play so the Partisan for me and Red Vet for the child.

[Addendum] Forgot to mention that we started with 4 Patrol Solo ships this time.[/A].

We called it in round 15. Damage deck had run out(!), 1 Black Sqd on 2 hull left really far out as only Solo ship left, Red Vet had lost one shield, Partisan really badly smashed, down to 1 hull and 1 def die. Had the tough luck of all the reinforcement spawning near the Partisan. Had been hanging on his dear life since suffering a Panicked Pilot, which nearly cost us the game, as at that point the Red Vet alone never would have done it against 4 solos which were alive and in quite good shape then.

Issues:

-what if a player ship blocks one or more positions (especially position nearest player/player edge; or blocking actually 2 pos on the marker) at a Hyperspace spawn point in a round for reinforcement arrival, and a Solo ship rolls that position, what are you supposed to do:

reroll for other position at same spawn or

choose best position at that spawn for AI or

spawn at the other hyperspace marker or

not spawning (like normal deploy rules)(which could be exploited)

-AI managed to run over asteroids uselessly several times (but that does HotAc as well, although to a lesser degree thanks to swerve rules)

-looked again like player keeping in vicinity to board edge makes it difficult for the AI, being an exploitation possibility.

-Again AI several times switching randomly to defensive robbed the AI of good shots. However, I maybe lean towards having slightly weaker reinforcements (lower number or lower Ini) if Solo ships are more or even constantly aggressive, 2 T65 struggle against too many aggressive mode ships (esp if Reinforce ships are high Ini bombers actively using their proton torps against by then damaged players).

----

A neat Reference sheet like the one @GuacCousteau quoted from a Reddit user makes life a lot easier and play so much faster. Although there is one error in it, front evade/blank is 3 bank not 3 turn.

Next iteration we will test @pheaver 's mod.

On 6/5/2020 at 11:12 AM, kris40k said:

^^This

If you look at some decent solo/co-op games like Descent, Frostgrave/Rangers of the Shadowdeep, Mansions of Madness, etc., that have their origins as competitive games, its not about attempting to out-think the enemy, anymore. Its about accomplishing some task/objective while being efficient and surviving waves of enemies/attrition. No one should expect the AI opponents to be as smart or as challenging as another player, that's just not going to happen.

This Alpha version of the AI that FFG has released is working on the waves part. The objectives will likely come later.

Except that waves of dumb/simple AI bots in X-Wing ends up being an overlap-fest. The more waves, the worst this problem becomes. The multiple waves logic does not translate well into X-Wing, unless it is a lot of waves with few AI ships.

There is also inflection point: while it is true that a simple AI takes less bookkeeping, overlapping ships takes longer to resolve (fiddly). You might end up taking longer to resolve a simple/incompetent AI that overlaps a lot.

I would rather have my X-Wing be about good dogfighting than turning it into a Zombicide with space ships! 🧟‍♂️

20 minutes ago, OoALEJOoO said:

There is also inflection point: while it is true that a simple AI takes less bookkeeping, overlapping ships takes longer to resolve (fiddly). You might end up taking longer to resolve a simple/incompetent AI that overlaps a lot.

Nope. Not buying it. Resolving collisions will take less time than resolving the ship's entire action sequence. Whether having a bunch of AI collide is fun or not, that is a function of what you are hoping to get out of this. If you like complex AI that does its best to simulate a competent opponent, then it isn't for you.

The current FFG system is more than Zombicide with TIE Fighters but it is not a particularly competent opponent. Howver, I am enjoying the edits Paul made and adding the swerve rules Ysenhal suggested. I think it is getting there for me and once I'm confident on the the AI, it will be easy to expand on it with missions and scenarios.

At some point in the future, I might look at coding the FFG AI into x-wing-ai.com (assuming they will allow me to do so without intellectual property issues becoming problematic), but right now my head is mush because I just finished coding HotAC AI into x-wing-ai See post here ...

Regards,

TC2

So I dusted off my X-Wing collection and gave this solo instruction set a try. Honestly I was very happy to see it work so well. I'm not an advanced player so there may be nuanced issues for more advanced play, but the initial play led to some interesting AI manoeuvres and some challenging dog fighting.

I was expecting it to take longer than a standard game with all the referencing, but no, it was actually much faster than a two player game.

So well done! Now if you can do this for Legion I will be seriously impressed. Getting my starter set arrive today

A few thoughts, after a couple of solo games - both the proposed escalation mode, and full squads.

Yes, the AI is dumb. The AI will self-stress and hit obstacles. Most AI shots are unmodded. Flight patterns are often chaotic.

Some time ago, I did play through the whole HotAC campaign. That AI is much better and a greater challenge. But: I don't really want to play HotAC again solo. Lookup tables for each ship, arcs, distance bands, directions, pre-measuring, measuring again, checking for reposition fitting, avoiding obstacles ... The new FFG AI: roll dice, move/action, repeat. Standard rules, simple tables.

Playing solo in casual mode, I have no intention of spending 90% of the game time on executing the AI turn.

In many recent boardgames, the solo (automa) game mode is similar. They speed up the AI turn, partly by simplifying rules, while accepting silly decisions. OTOH, computer-based AI (FlyCasual) can be programmed as smart as possible since the player doesn't have to work for the AI.

That said, I guess there is some space for improvement without sacrificing the simplicity of the alpha. The forward/bullseye approach is a weak spot. The random attitude rarely fits the game state. Low-ini patrol ships should not evade, just focus? Secondary weapons don't work because of missing locks.

If the AI ships have good dials (blues!) and move after the player, it starts to make sense. The most decent game was against an AI composed of TIE/fo and TIE/sf. (One sf even got a missile off once...)

I did last week iteration 3 with 1 of the kids, same setup, 2 T65 against TIE l/ns and TIE bombers, starting with two solos, this time with @pheaver s mod. Game ended about turn 16, with a win, but both T65 were pretty mangled, down to 2 hull.

We used the @pheaver mod even for a number of TIE l/ns accompanying a Raider which we (2 of the kids and I) with a mixed squad cooperatively attacked (I did the Raider, after having set my dials, that thing is a killer, one-shotting stuff).

This AI flies less erratic than FFGs original, thus being a bit more challenging.

It still suffers from the random switches in aggressiveness, and still has issues with getting Target Locks.

Thanks for trying it out! My goal with my table was to maintain the quickness and simplicity that, for example, Canopus calls out above as a merit of the FFG Solo system, and keeping some of the randomness, while removing some of the weird moves that the original table had. It sounds like it is making good progress towards that goal.

I think I'll make the changes that came up in earlier discussions in this thread and post it to Reddit or something to see if I can get more people to try it out.

Did someone post this here already?

Solo AI vs HotAC AI: https://www.goonhammer.com/x-wing-flying-solo-part-3-machine-vs-machine/

1 hour ago, Odanan said:

Did someone post this here already?

Solo AI vs HotAC AI: https://www.goonhammer.com/x-wing-flying-solo-part-3-machine-vs-machine/

Nope, but thanks as I hadn't seen it. It is not surprising regarding the results. The current iteration of the alpha AI is epitomized by the actions of that blue bomber doing parking lot donuts. It suffers from too many randomized options without enough guidance/feedback from the board state. The writer's suggested fixes are interesting in that they align with Paul's in the sense that they both want to account for additional information regarding the Tally's board state. I like Paul's approach simplifying options based on the Tally's directional facing but can see proximity being taken into account as well. I feel like the other AI systems out there like HOTAC are too much "work" for me to really get into so bloating out the system with more conditions may backfire.

We're coming upon 3 weeks since the initial Alpha release. Hopefully they have had enough time to collect feedback and revise for a second wave of testing. The next release (if any) will indicate which direction the design team wants to take this system.

I would be awesome if FFG did a streaming covering both the new points and hyperspace rotation and info on how and when they are moving forward with solo play.

I did some further testing with @pheaver 's version as written earlier in this thread. Squads: 2 TIE/x1 (no upgrades) against 3 TIE/fo and 2 TIE/sf with concussions. Before, I had tried the same match using the FFG AI.

A few rules might need some tuning: what means "obstacle in front arc" if there is no obstacle in the front arc because the ship is near the board edge. And what happens if the solo ship is stressed and has to choose an advanced maneuver? This case is covered only in the FFG rules. Otherwise, I like the approach to choose a different speed if stressed.

The AI moves appear somewhat less erratic. But: the new outcome was even more in the player's favor: zero damage, all enemy ships destroyed. No missile shot by the TIE/sf. The two matches lasted longer - up to 17 rounds. Probably I was just too cautious, but the AI ships also had a tendency to run away. Once a solo ship is on the opposite side of the board, struggling to turn around, it will likely find a side-side constellation of solo-tally, which can result in 'fastest turn away from tally'. With the enemy force hiding in a corner, I could pick off the ships one by one ... of course, this is just two test matches.

Solo AI vs HotAC AI - hilarious! Without using any distance information, the solo AI is doomed. But I still think pre-measuring AI moves should be avoided, to speed up the game. The other issues are too many lost actions (stress/asteroids), and the random attitude not matching the board state - maybe this can be improved?

On 6/15/2020 at 5:14 PM, Canopus said:

A few thoughts, after a couple of solo games - both the proposed escalation mode, and full squads.

Yes, the AI is dumb. The AI will self-stress and hit obstacles. Most AI shots are unmodded. Flight patterns are often chaotic.

Some time ago, I did play through the whole HotAC campaign. That AI is much better and a greater challenge. But: I don't really want to play HotAC again solo. Lookup tables for each ship, arcs, distance bands, directions, pre-measuring, measuring again, checking for reposition fitting, avoiding obstacles ... The new FFG AI: roll dice, move/action, repeat. Standard rules, simple tables.

Playing solo in casual mode, I have no intention of spending 90% of the game time on executing the AI turn.

In many recent boardgames, the solo (automa) game mode is similar. They speed up the AI turn, partly by simplifying rules, while accepting silly decisions. OTOH, computer-based AI (FlyCasual) can be programmed as smart as possible since the player doesn't have to work for the AI.

That said, I guess there is some space for improvement without sacrificing the simplicity of the alpha. The forward/bullseye approach is a weak spot. The random attitude rarely fits the game state. Low-ini patrol ships should not evade, just focus? Secondary weapons don't work because of missing locks.

If the AI ships have good dials (blues!) and move after the player, it starts to make sense. The most decent game was against an AI composed of TIE/fo and TIE/sf. (One sf even got a missile off once...)

Speaking of other games with AI, something like the cards that enemies have in Gloomhaven would be welcome in a card expansion but the system seems okay as is. I'd expect these to be somewhat small decks that reference the threat build cards (or maybe have new threat builds included for this purpose specifically). Certain decks would encourage things like taking locks (like a TIE bomber equipped with ion missiles) before taking focuses, and some that encourage flying in a way that bombs might work better (I've got no idea how, but that would be a goal). Of course, some ships would have simpler decks that really are just "move towards them, focus, attack if possible (such as an Academy Pilot).

But the system they've presented does a decent job of letting you work without needing much else than what you already have. I think we all have some small modifications we'd do to prevent some of the sillier behaviours for the AI.

21 hours ago, Canopus said:

A few rules might need some tuning: what means "obstacle in front arc" if there is no obstacle in the front arc because the ship is near the board edge. And what happens if the solo ship is stressed and has to choose an advanced maneuver? This case is covered only in the FFG rules. Otherwise, I like the approach to choose a different speed if stressed.

The AI moves appear somewhat less erratic. But: the new outcome was even more in the player's favor: zero damage, all enemy ships destroyed. No missile shot by the TIE/sf. The two matches lasted longer - up to 17 rounds. Probably I was just too cautious, but the AI ships also had a tendency to run away. Once a solo ship is on the opposite side of the board, struggling to turn around, it will likely find a side-side constellation of solo-tally, which can result in 'fastest turn away from tally'. With the enemy force hiding in a corner, I could pick off the ships one by one ... of course, this is just two test matches.

Thanks for testing it out!

Hmm, good point about the obstacle thing. Would adding "for all nearest obstacle in front arc checks, the board edge counts as an obstacle" help?

I can add "If stressed and instructed to perform an advanced maneuver, change the green die result to an evade." The evade doesn't do advanced moves ever, so that'll help that.

You have a good point about side-side. There are 3 "turn away from the Tally" in my table. Side-Bullseye, Side-Side, and Rear-Front. I can change the side ones to slowest turn towards the Tally. The rear one might need to be fastest turn away from the nearest front obstacle.

Thanks for the suggestions.

Thanks, Paul, suddenly this has become playable!

On 6/18/2020 at 9:59 PM, pheaver said:

Thanks for testing it out!

Hmm, good point about the obstacle thing. Would adding "for all nearest obstacle in front arc checks, the board edge counts as an obstacle" help?

I can add "If stressed and instructed to perform an advanced maneuver, change the green die result to an evade." The evade doesn't do advanced moves ever, so that'll help that.

You have a good point about side-side. There are 3 "turn away from the Tally" in my table. Side-Bullseye, Side-Side, and Rear-Front. I can change the side ones to slowest turn towards the Tally. The rear one might need to be fastest turn away from the nearest front obstacle.

Thanks for the suggestions.

In a match with those revisions, I actually lost (generic TIE/x1 vs TIE/fo+TIE/sf alpha-scenario as above, about 75 vs. 170pt) with one solo ship left. In another incomplete match the AI also was decent. Disclaimer: I'm not a good player. (But I guess the FFG Solo mode is targeted at casual gaming anyway.) And this is a very small data set.

The revised AI has a tendency to close in on player ships. It avoids the board edge if left alone, the rule 'count edge as obstacle' makes a huge difference. Sometimes it manages to form kill-boxes if the player is not careful. It does like to hit obstacles, but I don't see this as a fundamental problem. Some moves overshoot, but then the ship will try to turn around. If unstressed, a solo ship can target-lock and shoot missiles, because it doesn't suddenly look elsewhere.

There is still an issue with the attitude roll and action selection. There are too many evades which just weaken AI attacks. Maybe rolling a red die instead of a green die for choosing the attitude would be an improvement?

Probably this won't win against HotAC AI, but it is quick and easy. I can resolve most AI moves and actions within seconds without pre-measuring.

About the stress rules:

Quote

If stressed, and the chosen move isn't blue, the Solo ship will check a speed one slower, and then one faster, for a blue move to choose. If it doesn't find one, it does the original move.

would it make sense to also try a blue bank of the same speed, if a white turn is chosen? Some dials don't have blue turns, so stress will stick. And if I apply this rule:

Quote

Unless performing an advanced maneuver, if told to perform a red move, the solo ship will check a speed one slower, and then one faster, for a non-red move to choose. If it doesn't find one, it does the original move.

some ships will never use their red 1-hard, is this intended?

I hope that FFG can adopt some of this for a future campaign or whatever. Comparing this vs. FFG AI, eyeballing two arcs instead of one is as easy and straightforward as rolling another die.

Another iteration, @pheaver ' s mod.

One of my children and I, but we deliberately took Sabine in the Attack Shuttle and Graz (downgraded to Ini 3, going before AI) together against TIE l/n and TIE Bombers. Easy win, even if it took to turn 17 to bring the last reinforcement down. It is still only one sample (and I might be biased), but corroborates a bit my suspicion I voiced some pages ago (were it was suggested to give the player Starvipers, Fangs, Silencers against bread and butter Solos). If the players has ships with too many repositions, high action ecenomy or strong attack values you have to ramp up the AI forces.

10 hours ago, Canopus said:

but it is quick and easy.

Agreed, that is another advantage of your mod, rolling only 1 die makes the AI and tge overall game so much faster.

44 minutes ago, Managarmr said:

Another iteration, @pheaver ' s mod.

One of my children and I, but we deliberately took Sabine in the Attack Shuttle and Graz (downgraded to Ini 3, going before AI) together against TIE l/n and TIE Bombers. Easy win, even if it took to turn 17 to bring the last reinforcement down. It is still only one sample (and I might be biased), but corroborates a bit my suspicion I voiced some pages ago (were it was suggested to give the player Starvipers, Fangs, Silencers against bread and butter Solos). If the players has ships with too many repositions, high action ecenomy or strong attack values you have to ramp up the AI forces.

Agreed, that is another advantage of your mod, rolling only 1 die makes the AI and tge overall game so much faster.

Did you use Paul's last revision of the mod?

I think you are right. To help the AI, I gave it TIE/fo (TL, shield) and TIE/sf (missile turret), and TIE/x1 for player ships (need TL for 3rd die). I expect that player ships start weak and level up during a campaign, while the AI gets more HP and 'boss' ships with tailored AI rules.

The solo play rules have been updated.

https://images-cdn.fantasyflightgames.com/filer_public/2d/6d/2d6d1669-5a0d-4a05-b022-5cf09c937655/x-wing-solorules_openalpha-compressed_v2.pdf

Sweet! I hope they made the AI more aggressive. When I played the first version, I was flying rings around the AI ships while they were constantly running into obstacles. It was so sad.

1 hour ago, Tervlon said:

The solo play rules have been updated.

https://images-cdn.fantasyflightgames.com/filer_public/2d/6d/2d6d1669-5a0d-4a05-b022-5cf09c937655/x-wing-solorules_openalpha-compressed_v2.pdf

Updated the OP and thread title. Thanks for posting the link and giving us a heads up.

Just played a game with the new AI with ARCs against droids. The shifting attitude by game state vastly improved gameplay. I ended up winning with 2 hull left on an ARC, game definitely could have gone either way. They were consistently pointing at me (although that might have to do with all of the droids having blue hard turns). They seemed to be doing too many advanced maneuvers, although again that might be a limitation from the droid chassis.

I think that it's going to be hard to make an AI system that works across the board regardless of ship type and dial, but it was definitely better.

I, too, saw the solo ships pull quite a few advanced maneuvers. One TIE/fo got caught in a Sloop-loop: he bumped into the edge of the board while performing the red maneuver, I modified the approach roll so he cleared stress, but then he inexplicably performed another Segnor's loop! Even though there was nothing behind him but the vastness of space. So he cleared stress, and then, I kid you not, he did ANOTHER Segnor's Loop, colliding with the edge of the board AGAIN!

But overall, a vast improvement. Switching attitudes has quite a learning curve, but by the end of the match (11 rounds and 1.5 hours) I quite enjoyed changing the solo ships' attitudes and marking them with little tokens. Looking forward to giving it another go!

P.S. The Resistance was routed. The Rookie made it to hyperspace successfully, but his mentor fell short of the jump point (map edge) and took a Concussion Missile in the engine nacelle. TIE/sfs...they'll getcha!

I don't think I fully understand modifying the approach roll (p. 8). If a solo ship is in Offensive attitude, but it's stressed, I modify the roll so the evade die shows a focus to ensure that the solo ship executes a blue maneuver and clears stress. But does the attitude change to Balanced now that I've modified the roll to a focus? Or do I ignore the modification for the purposes of attitude and keep the original roll?

There is an asterisk at the bottom of the page that says:

Quote

*Note that the modified approach roll is used for this. For example, a ship in Recovery always modifies the defense die to a [focus] result, so its attitude will not shift as a result of the Approach Roll.

However, there is no asterisk in the text of the page, so I have no idea what "this" is referring to. Let me know if I'm missing something.

17 hours ago, Parakitor said:

I don't think I fully understand modifying the approach roll (p. 8). If a solo ship is in Offensive attitude, but it's stressed, I modify the roll so the evade die shows a focus to ensure that the solo ship executes a blue maneuver and clears stress. But does the attitude change to Balanced now that I've modified the roll to a focus? Or do I ignore the modification for the purposes of attitude and keep the original roll?

There is an asterisk at the bottom of the page that says:

"*Note that the modified approach roll is used for this. For example, a ship in Recovery always modifies the defense die to a [focus] result, so its attitude will not shift as a result of the Approach Roll."

However, there is no asterisk in the text of the page, so I have no idea what "this" is referring to. Let me know if I'm missing something.

I believe the note at the bottom is referring to the change in attitude as a result of the roll. So e.g. if a ship is Offensive and rolls evade, it changes to blank and stays Offensive; it doesn't shift towards Defensive.

To be honest I'm not sure how much the change in attitude as a result of approach role really adds to be honest, it seems like the "did I get shot at/take damage" check would be sufficient, but I want to try it again and keep a closer eye on it. I guess going into Recovery if you modify the roll after a stressy maneuver reduces the chance of constant advanced maneuvers.

I agree that overall this AI is a lot stronger, though - if nothing else it seems a lot more likely to actually point ships in your general direction, instead of having them flop about in the middle of nowhere.