Jump to content
Hiemfire

Flying Solo (FFG X-Wing Solo Play article up)

Recommended Posts

16 hours ago, GuacCousteau said:

Oh, one other thing I'd love to see down the line - special, more complex 'hero' AI for certain ships/pilots. I'd love a scenario that involves taking on Boba Fett or Han Solo, for example. And AI that's designed around their abilities and ships wold be awesome. 

 

2 hours ago, GuacCousteau said:

Instead, I added a Lambda shuttle to the mix. It starts in one of the AI side's corners, pointing diagonally across the board to the opposite corner. The first two TIEs were positioned slightly ahead and to the sides in flanking positions. The shuttle used the I1 OGP, so it could be activated in any order with the Academy TIEs, but it could only ever perform the 1 straight and only ever focus. It will shoot anything that ends up in its arcs according to the same rules as the TIEs

This shuttle is carrying a 'VIP' the Rebels must capture. To do this, the shuttle must sustain three ion tokens and become ionised. Once it is, it becomes 'disabled' and also receives a disarm token. It does not perform the ion maneuver next turn - it is stuck in place. The disarm token is not removed. As soon as the shuttle is disabled, a U-Wing is deployed in the Rebel's corner (the one to which the shuttle is heading). The U-Wing must maneuver until it is at range 0 of the shuttle. Next turn, it must perform the stop maneuver. This is the boarding action, and once it has completed this, the VIP is captured and the U-Wing must then leave the battlefield intact. 

 

Turns out I should read the full document before commenting.

There's two paragraphs in the 'Now What' section that describe basically exactly these. 

They literally suggest dropping a shuttle in and having it fly towards the opposite corner. Only their system 'locks' the opposite corner. This is actually pretty cool, it means the shuttle will always fly towards it, but the AI will randomise its movements a bit more. Glad I was thinking along the same lines as FFG though in terms of objectives. Interesting that their version doesn't buff the player in any way. As I found in my game, even adding an extra 'passive' 3 dice arc and 2 dice rear turret from a large base heading down the centre of the board has some balance implications. I probably wouldn't have lost both ships if my shuttle hadn't shredded my X-Wing with a lucky range 1 front arc shot and brought it down to 1 hull. 

 

The 'ace hunt' sounds fun too. It's really rudimentary, but at least it provides some framework for aggressive 'hero' AI behaviour. Hopefully that can be refined going forward. I might have to try that mode out.

 

It's a slow day at work today, so I ran two Jedi Knights vs Scum just now. 

Yeah, I *really* think FFG are going to need to do ship specific charts. Or maybe at least control certain matchups. It's amazing what happens when you put a ship with a bad dial in. Those Z-95s were totally helpless. Especially against something as arc dodgy as Delta 7s. The firepower gap gets more serious here too, as does the lack of evade actions. I killed one Binayre Pirate in a single shot - 4 red dice with one guaranteed hit (because you've always got a Force or a focus) just obliterates them. The other one lost a shield to a range 3 plink and melted to another range 1 CLT shot next round. 

That said, I realised afterwards I've been playing it wrong. I've been giving the AI first player. I really don't think this scenario is balanced around that at all. I probably shouldn't have been able to arc dodge or line up as many CLT bullseye shots as I was. 

Still, the Skull Squadron Fang showed up at round 4, and that's moving last and shooting first no matter what. But it really wasn't the threat it shouldn't have been. Bad approach rolls meant it wasn't getting arc even with its repositions. It rolled an 'average blue straight' - a 3 straight when any sane player in the world would have done the 2 forward and boost or the 2 bank to get arc at the side of one JK. It had done precisely nothing before I killed the last Z-95 and got bored. 

Again, I wonder if it's the specific nature of the Fang's dial that caused this. It didn't start in the best position, but the first two rounds it made good use of its blue hard 2s. But once the tally started appearing in the front and side arc it just couldn't pick the right maneuver. 

 

After three games, I'm definitely in a bit of a dilemma. On the hand, I know FFG want to keep this as simple as possible, and I do enjoy how easy it is to reference the movement with just a couple of dice rolls. But on the other, there's some basic elements of the game that don't feel right here. The biggest things for me, in addition to what I've already described re: stress and obstacles would be a range check and an activation order check.

I1 AI pilots should be able to make more of an effort to block. They should be able to recognise that an enemy ship in front arc at range 2-3 that hasn't activated yet presents the perfect blocking opportunity with a 1 or 2 forward, but that a ship in front arc in range 1 should probably be k-turned or fast straight past. 

I'm aware of how complicated this would start to get, but I do think some way of 'preferencing' if a tally is in two arcs would be good. A ship in front arc presents a slightly different challenge from a ship in front arc and side. Especially if it's mostly in the side arc. There should be more of a bias towards turns in that context, because a few times the straights it prefers when the tally is in front arc just take them right by the tally. I wonder if maybe instead of a rigid sequence check of bullseye, front, side and rear, the rules should allow some correction for the player and say that if the closest ship is 'mostly' in side arc instead of front (judged by eyeballing), the player can choose whether to use the side or front approach based on what would be better for the AI. 

1 hour ago, Azrapse said:

The biggest obstacle against his approach is that it requires personalized maneuver tables per ship type, essentially like HOTAC.
While it doubtlessly leads to smarter AI, I don't think FFG wants to (or even can) commit to keep reference cards for each ship type present and future. 
I am betting this solo system will have like a couple of months of attention by FFG, then it will be abandoned as they did with the Mission Control editor, the app, cinematic play, etc.
So it's best that the AI system is future proof and doesn't need any upkeep from FFG for its survival.

The complexity of its tables could probably be reduced in some clever way. I am thinking on those code wheels from the copy protection systems in PC video games in the 90s. Three or four concentric wheels with punched holes in it. Select the target relative position in the outer wheel, the target current vector in the next wheel. With that alone, you get rid of having to cross-reference maneuver codes between tables.
Then some punched holes labeled with [green result][red result] display selection of maneuvers by target range. Just roll a green and a red die and select the matching hole.
For the actions, add more holes, or just some 1. 2. 3. list like the ones in current FFG proposed system.

To deal with the fact that not all ships have all maneuvers, instead of having a single wheel set for every ship in the game (current FFG proposal), or a wheel set for each ship type (HOTAC, or Tactical Droid), have one wheel set per ship archetype. Define like 4 or 5 ship archetypes and have 4 or 5 wheels. When resolving the AI maneuver and actions for each AI ship, just pick up the wheel that matches that ship's archtype.

Example of archetypes: 

  • Dogfight: X-wing, B-wing, TIE Ln, M3-A, Trade Fed Drone, Mist Hunter, TIE Defender, TIE x1, Kihraxz, TIE sf, Belubab, TIE Phantom...
  • Interception: A-wing, TIE Interceptor, TIE v1, Fang, TIE vn, Aethersprite...
  • Strike: Y-wing, K-wing, Starwing, TIE Bomber, TIE Punisher, ...
  • Transport: Lambda, YV-666, U-wing, Upsilon, ...
  • Turreted: YT-1300, YT-2400, HWK, Decimator, ...

I think you're spot on about the rules needing to remain as general as possible so they can continue to be relevant past support of the game mode. Personally I think it would be best to keep the general rules as they currently are (with some tweaks here and there, the approaches aren't quite optimised yet) but have specific ship cards for some of the more commonly used AI ships. The list they have at the moment would be the obvious starting point, with maybe some of the current player only ships and some famous faces like the Firespray or Falcon included. 

 

That said, I also like what you're suggesting as a middle ground. The archetypes idea is a good one, as it allows you to tailor things a bit more to specific ships, but keep the rules general and 'future proof'.

Based on my experience so far, though, I don't think combat role is quite the way to go. Whether the approach system works or not really does seem to come down to what's on the dial, and I think they should be grouped according to some commonalities. 

Off the top of my head, ships with all three speed turns available should be a different group from those that only have one or two. 

Ships with no turn based advance maneuvers (s-loops, t-rolls) should be in a separate group from those that do. 

Ships with 6+ blues available should be in a different group from those with fewer. 

That alone gets you eight archetypes, unfortunately.

  • 3 turns, no turn based advanced, 6+ blues
  • 3 turns, no turn based advanced, <6 blues
  • 3 turns, advanced turns, 6+ blues
  • 3 turns, advanced turns, <6 blues
  • 2/1 turns, no turn based advanced, 6+ blues
  • 2/1 turns, no turn based advanced, <6 blues
  • 2/1 turns, advanced turns, 6+ blues
  • 2/1 turns, advanced turns, <6 blues

And arguably a fourth distinction could be set between ships whose only 1 speed maneuver is a turn vs those that have a bank or straight. But that would take you up to 16 archetypes, at which point you may as well just go ship specific. 

Then there's who has access to boost, SLAM etc.... 

I'm sure there's a way to boil this down. Maybe you're right with the combat role thing after all. I feel like the way to do it is probably somewhere in the middle. A couple of key distinguishing features. 

Share this post


Link to post
Share on other sites
Posted (edited)
18 minutes ago, GuacCousteau said:

Maybe you're right with the combat role thing after all

I was thinking more on roles or archetypes indeed taking into consideration the ship's abilities (dial, action, and even weapon preference), more than a realistic meaning of the role. The names given were just fancy names, for immersion sake.
That is, the Interception role was basically "ships with lots of turns, green turns, and plenty of repositioning". While the Strike role was "Ships that want to face a target frontally and release heavy attacks, without any fancy flying".

I think a turreted ship, or one with side arcs like, I don't know, the Corellian Corvette, will want to strafe around its target, rather than go full speed towards it. There is no point for an YT-2400 to go directly towards its target once its at firing range. A K-wing loaded with torpedoes will be a Strike craft, while the same K-wing having just its primary turret set sideways will want to strafe targets while keeping them in range.

We need some sorts of "flight plan preference" based on what the ship is supposed to do, otherwise we will end up with 90% of the ships doing a poor job at being dogfighters, for which they are absolutely not equipped.

Edited by Azrapse

Share this post


Link to post
Share on other sites
Posted (edited)
3 hours ago, Azrapse said:

The components issue is irrelevant. Instead of using d12's or d10's, just adapt the rolls to match a green and die roll. The results won't be identical, but just close enough.

Actually, I believe the complete opposite. The components are probably the most important factor.

If you saw the Team Covenant interview with Asmodee USA head Steve Horvath a few months back, you should probably know that there are a lot of shifts happening around Asmodee, and FFG in reaction to the changing tabletop game industry. They are making some big shifts in their thinking and actions about the industry, and what we see happening here is a part of that I think. They are tightening their belts, made obvious by many games getting "no new physical content" announcements, and splitting FFG into two separate companies and breaking apart those remaining game lines. 

One of the shifts in the industry, even before the COVID crisis, was a shift towards solo/co-op playable games. You can look at recent games like Rangers of Shadowdeep which is a solo/co-op game that spawned from competitive Frostgrave (which had added solo rules after launch), The Elder Scrolls: A Call to Arms which is solo/co-op only. FFG announced an AI expansion for Fallout in February. When FFG launched the 2nd Edition of Mansions of Madness, they completely removed the Keeper/competitive aspect of the game and went solo/co-op only. 

I think X-Wing Marketing being able to say, "Play solo/co-op right out of the box!" is a big point that they will want to stick to, and then push cardboard sales later with a campaign expansion. So, yeah, any AI system needs to stick with what is in the box. 

Edited by kris40k

Share this post


Link to post
Share on other sites
Posted (edited)
36 minutes ago, kris40k said:

Actually, I believe the compete opposite. The components are probably the most important factor.

You are probably very right in all of what you say.
But I meant irrelevant in the sense that you can adapt almost any probability table to work with any kind of randomness generators, more or less. You won't get the exact same probability distribution, but you can make do with what you get.

Each of the 12 sides in a d12 has a 8.3% chance to come up. By rolling a green and red die, you get, instead, these chances (rounded):

Blank Blank: 9%
Blank Eye: 9%
Blank Hit: 14%
Blank Crit: 5%
Eye Blank: 6%
Eye Eye: 6%
Eye Hit: 9%
Eye Crit: 3%
Evade Blank: 9%
Evade Eye: 9%
Evade Hit: 14%
Evade Crit: 5%

That is, like 12 possible results with slightly different chances. It's not the same, but surely you can adapt a table that uses d12 rolls (or d10, or whatever), to the kind of probability distribution provided by rolling a green and a red X-wing die.

Geez, we could even throw the hyperspace markers in the core set and interpret the way they fall in some way to give us a different probability distribution if we wanted. We would look like chamans throwing bones, but whatever.

The important parts of the HOTAC or the Tactical Droid approach to AI aren't the dice, but the idea of taking into consideration range, movement vector, and yes, ship type. These three factors are the main differences with the FFG's Solo system, that makes them so much superior as opponents. And we should try to make Solo be more like HOTAC/TD while keeping Solo's own virtues, like ship type-agnosticism and component-light requirements.

Edited by Azrapse

Share this post


Link to post
Share on other sites
Posted (edited)
15 hours ago, That Blasted Samophlange said:

First issue I have with your system is it doesn't seem to go with the design intent for FFG.  

FFG is trying to make their system work with x-wing products.   As it is, I could buy a core set print off the AI and play, requiring nothing else.   No additional dice Needed, using what is in the box.   That is a good idea for FFG.   
 

As far as I can tell, FFG has not really mentioned what is their design intent. All we know is that they have these Alpha solo rules in mind and that they are requesting feedback.

My thought is that the design intent would be a trade-off that could go three ways:

A) They indeed aim to only require a core set and nothing else. The solo rules are made available online and it's up for the player to print/design his/her own reference sheets. This would pose severe have design constraints, the core set was not designed with solo play in mind. This is a good idea if the aim is to have a simple, short-term system out the door. On the long run, how are they going to release content, if any? Free online PDF campaigns? From a business perspective FFG would need to anticipate this free content would drive ship sales to go up.

B) Sell a purpose-built solo pack. This pack would include reference cards and most likely a more sophisticated AI system, not bound/constrained to what already exists in the core game, and with the potential to expand on the long-term. It would also set precedent for high-quality campaign books and content. Personally, I would gobble this up and pay good money for such campaigns (e.g. something of the sorts of Imperial Assault). The boardgame industry is more and more catering solo players, I think FFG is realizing this, beyond covid19 lockdowns. There is a business opportunity here if they can have a grip on it.

C) FFG can design a scalable AI system. This would be a system that has the ability to be very simple if only the core logic is followed. For advanced players, a second layer of logic can be added to have a more sophisticated AI. This option would be outstanding but would require more design/testing time.

I guess there could be arguments made against either A), B) or C). There is no right or wrong answer. It depends on how FFG decides to approach this business opportunity.

15 hours ago, That Blasted Samophlange said:

Second issue; Looking at your AI... it looks complicated.    Say I’m running a dozen AI ships.  Going between all those charts is going to slow the game down. 
 

Running a dozen ships, be it AI or human, takes time. You can't simply say the AI takes time because running ships is the nature of the game. I haven't seen so far any X-Wing AI system (mine included) in which the AI takes on average longer to resolve than a human player.

15 hours ago, That Blasted Samophlange said:

It seems you are proud of your creation, and that is good, but it is not what FFG is likely to do.   They want people to keep buying x-wing products - they are a company, and they aren't likely to start making d10’s and d12’s or whatever your system needs.  They are designing their system with existing materiel, namely their dice.   Providing a complete overhaul and new AI system (that looks very complicated at first glance) is not providing feedback.   
 

There seems to be confusion about my intent on posting on this forum. To be clear:

- I am not asking FFG to copy my system. My system is not perfect and am sure it can be improved/streamlined further.

- I see FFG making some of the mistakes early AI systems did and I am just pointing out the flaws so that they can consider improving it. Namely that ships should not determine maneuvers post the planning phase.

- A lot of people seem to comment with statements such as "first glance", "looks like", etc. It is not possible to get a feel of the nuances of an AI system without playtesting and playtesting some more. My feedback is offered after 100's of hours of playtesting. An opinion is great, an informed opinion is better.

- I am indeed proud of my creation. If I already have a system I love, why then bother posting here in the forum? Because again I think it can be further improved by FFG and also because I enjoy a good discussion. So far the only comments I've received towards the Tactical Droid AI on this thread is from people that have not tried it and think "looks like" its complicated. People who have tried it have loved it as stated on BGG and feedback I've received in person. X-Wing is a complicated game, far more complicated than any AI out there, if you already play it chances are you enjoy this type of thing. 

Edited by OoALEJOoO

Share this post


Link to post
Share on other sites
1 hour ago, pheaver said:

I made my own table for Solo ship movement choices.  It's similar to the current table, but less random.  Instead of two dice rolls, you use the arc of the Tally that the Solo ship is in.  Hopefully it's pretty self-explanatory.  Feel free to try it out and give me feedback!

 

SoloTestingApproaches.xlsxUnavailable

It says Unavailabe for me.

Share this post


Link to post
Share on other sites
16 minutes ago, Azrapse said:

It says Unavailabe for me.

Hmm, that's annoying.  I can download it, but if I log out, I see the same.  I assume it's some thing to stop people from linking virus-laden files or something.

Oh, I can just paste it from the excel sheet:

    Bullseye Arc Approach
If the solo ship's Tally is in its bullseye arc…
And the Solo ship is in the Tally's:
Bullseye Arc Evade Slowest Straight
  Focus Average Straight
  Blank Fastest Straight
Front Arc Evade Slowest Straight
  Focus Slowest Bank in the direction the Tally is flying
  Blank Average Bank in the direction the Tally is flying
Side Arc Evade Slowest Bank in the direction the Tally is flying
  Focus Average Bank in the direction the Tally is flying
  Blank Fastest Bank in the direction the Tally is flying
Rear Arc Evade Average Straight
  Focus Fastest Straight
  Blank Fastest Straight
     
    Front Arc Approach
If the solo ship's Tally is in its front arc…
And the Solo ship is in the Tally's:
Bullseye Arc Evade Slowest Straight
  Focus Slowest Bank towards the Tally
  Blank Average Bank towards the Tally
Front Arc Evade Slowest Bank towards the Tally
  Focus Average Bank towards the Tally
  Blank Fastest Bank towards the Tally
Side Arc Evade Average Bank in the direction the Tally is flying
  Focus Fastest Bank in the direction the Tally is flying
  Blank Fastest Turn in the direction the Tally is flying
Rear Arc Evade Average Bank towards the Tally
  Focus Fastest Bank towards the Tally
  Blank Fastest Straight
     
    Side Arc Approach
If the solo ship's Tally is in its side arc…
And the Solo ship is in the Tally's:
Bullseye Arc Evade Fastest Turn away from the Tally
  Focus Slowest Turn towards the Tally
  Blank Fastest Turn towards the Tally
Front Arc Evade Slowest Turn towards the Tally
  Focus Fastest Turn towards the Tally
  Blank Fastest Sloop away from Tally, T-roll towards Tally, K-turn, or Turn towards Tally (in that order)
Side Arc Evade Fastest Turn away from the Tally
  Focus Average Turn towards Tally
  Blank Fastest Sloop away from Tally, T-roll towards Tally, K-turn, or Turn towards Tally (in that order)
Rear Arc Evade Slowest Turn towards the Tally
  Focus Average Turn towards Tally
  Blank Fastest Turn towards the Tally
     
    Rear Arc Approach
If the solo ship's Tally is in its rear arc…
And the Solo ship is in the Tally's:
Bullseye Arc Evade Fastest Straight
  Focus Fastest Turn away from the nearest obstacle in your front arc
  Blank Slowest advanced maneuver away from the nearest obstacle in your front arc
Front Arc Evade Fastest Turn away from the Tally
  Focus Fastest advanced maneuver away from the nearest obstacle in your front arc
  Blank Slowest advanced maneuver away from the nearest obstacle in your front arc
Side Arc Evade Fastest turn in the direction the Tally is flying
  Focus Fastest advanced maneuver away from the nearest obstacle in your front arc
  Blank Slowest advanced maneuver away from the nearest obstacle in your front arc
Rear Arc Evade Slowest turn away from the nearest obstacle in your front arc
  Focus Fastest advanced maneuver away from the nearest obstacle in your front arc
  Blank Slowest advanced maneuver away from the nearest obstacle in your front arc

I also added the following rules when choosing a move:

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.

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.

Share this post


Link to post
Share on other sites
Posted (edited)

Paul's revised chart is exactly what I was thinking of after I played a few games of the current v0.01 rules. I've already given this feedback in their form but the issue I found is there are too many options for the AI to totally mess up obvious maneuvers.  When splitting out the options based on more game state information (like you have here where the solo considers not only its tally's relative position but also its relative position to the tally) you can divide the AI maneuvers into smaller, less random options.  No more will the solo ship perform a random k-turn when chasing me or veering off in some random direction.

I'll print this out and give it a go tonight.  I didn't jump in on the forum right away but kudo's to FFG for making a go at an official product that is very relevant at the moment. Also, I really like the ambitious restriction of only requiring the core box material and the new rules document.

 

Edit: what does "Tally, K-turn, or Turn towards Tally (in that order)" mean?

Edit2: also attached your new variant in the original FFG layout.

x-wing-solorules_openalpha_PHeaver_Approach1.pdf x-wing-solorules_openalpha_PHeaver_Approach2.pdf

Edited by spacemonkeymafia
added "PHeaver Approach" edits to FFG's document layout

Share this post


Link to post
Share on other sites

After running Paul's revised AI, I like it but have to ask what the intention of the "in the direction its Tally is flying" option is. I had it come up three times in my little core box skirmish and every time it was a complete disaster for the enemy. I think you'll get more mileage pulling it back down to a simple "towards its Tally". In all three cases, that change would have either denied me a shot or gave the enemy a great line on me. 

I think system is improved over the current FFG model (enemies aren't banking randomly into asteroids or circling obstacles turn after turn). I don't know if the system is easier to pick up or I'm just getting used to the concept in general but it feels faster. 

Ive tried HOTAC but even after being shown the game, it still felt like a lot of book keeping. The game did feel like I was playing against a passable opponent and the ships were being used to fit their skills rather than a one-size-fits-all. I have not tried OoALEJOoO's system but I'll take a look as well. 

Share this post


Link to post
Share on other sites

Played a test game of this mode. 2 x-wings vs empire like the test scenario in the rules.

First impressions: One tie fighter turned straight into an asteroid the first turn, and due to the random manouvers the ai's moves where either stupid or brilliant. It felt like the front arc/side arc/rear arc approach didn't affect the manouvers more than the random dice roll IMO. The blank blank result in 2 straight was just a handicap for the ai. 

I had fun and even lost one of my x-wings when one enemy tie decided that instead of moving towards it's tally it did a sharp turn and placed itself in the path of an asteroid, but at range 1 with focus on my second x-wing :P. The first turns where a bit uneventful where the enemy ships just flew around at random, but when they got their reinforcments then they started to get some attacks in, however the random manouvers where they suddenly did their fastest manouvers right past me or they got forced by their action list to barrel roll into a bad position or take an evade instead of barrel rolling into arc.

Honestly it is a decent attempt at a solo play rule wich will help witch the craving I have for gaming during the pandemic time, however there needs to be some uppdates on the flowcharts cause alot of the  time I was tempted to just take over and fly the enemy ships myself to make better choices for them, but at that point there is no need for a solo rule set.

I would give it 6/10. Needs work, but will work for a few games or as an introduction for a new player that like a co:op learning game.

Share this post


Link to post
Share on other sites
8 hours ago, spacemonkeymafia said:

After running Paul's revised AI, I like it but have to ask what the intention of the "in the direction its Tally is flying" option is. I had it come up three times in my little core box skirmish and every time it was a complete disaster for the enemy. I think you'll get more mileage pulling it back down to a simple "towards its Tally". In all three cases, that change would have either denied me a shot or gave the enemy a great line on me. 

I think system is improved over the current FFG model (enemies aren't banking randomly into asteroids or circling obstacles turn after turn). I don't know if the system is easier to pick up or I'm just getting used to the concept in general but it feels faster. 

Ive tried HOTAC but even after being shown the game, it still felt like a lot of book keeping. The game did feel like I was playing against a passable opponent and the ships were being used to fit their skills rather than a one-size-fits-all. I have not tried OoALEJOoO's system but I'll take a look as well. 

Thanks for trying it out!

The "Fastest Sloop away from Tally, T-roll towards Tally, K-turn, or Turn towards Tally (in that order)" (and similar statements) are basically saying "turn around and point at the Tally, but I don't know what turn around options you have."  So do a Sloop if you have that, T-roll if you have that, K-turn if you have that, and finally, if you have none of those, just turn.

The "in direction the Tally is flying" was, in my head, dealing with the Tally being in front of you, on say the left side of the front, and flying towards your right side.  If you bank left, the Tally might just fly past you and now you have to turn around.  Banking right might block the Tally, or line up a shot if the Tally flies past.  But I see how banking right in that case more often will just let the Tally line up a perfect shot.  You're probably right that it should just be "towards the Tally", at least for the Front-Side section of the table.  For the Bullseye table, you can't really do banks "towards the Tally" since it's in your bullseye and not on a side of the front arc.  Do you think "in the direction the Tally is flying" works here, or should I go with "away from the closest obstacle in your front arc"?

Thanks again for testing it out!

Share this post


Link to post
Share on other sites

Do you guys think it would help the AI if the expression "towards the Tally" was redefined to mean:

  • If the Tally has already activated this round, its current position.
  • If the Tally has not yet activated this round, its current position plus 2 forward template length.

?

This way, the AI would try to "lead its target". Otherwise, low initiative AI will try their best to aim towards positions where their Tallies will no longer be at the start of the engagement phase. So what's the point?
By making low initiative ships try to go not where the Tally is, but where it is going to be, perhaps makes them more effective at blocking or shooting.

What do you think?

Share this post


Link to post
Share on other sites
1 hour ago, pheaver said:

Thanks for trying it out!

The "Fastest Sloop away from Tally, T-roll towards Tally, K-turn, or Turn towards Tally (in that order)" (and similar statements) are basically saying "turn around and point at the Tally, but I don't know what turn around options you have."  So do a Sloop if you have that, T-roll if you have that, K-turn if you have that, and finally, if you have none of those, just turn.

The "in direction the Tally is flying" was, in my head, dealing with the Tally being in front of you, on say the left side of the front, and flying towards your right side.  If you bank left, the Tally might just fly past you and now you have to turn around.  Banking right might block the Tally, or line up a shot if the Tally flies past.  But I see how banking right in that case more often will just let the Tally line up a perfect shot.  You're probably right that it should just be "towards the Tally", at least for the Front-Side section of the table.  For the Bullseye table, you can't really do banks "towards the Tally" since it's in your bullseye and not on a side of the front arc.  Do you think "in the direction the Tally is flying" works here, or should I go with "away from the closest obstacle in your front arc"?

Thanks again for testing it out!

Oh I see it now, the spacing of the excel image above made me think the entire command was just "Tally, K-turn, or Turn towards Tally".  I see that the "Fastest Sloop" was part of that whole command. 

When I was messing around after the game with the "in the direction of" commands, I thought that was what you were going for- an option for the solo ship to pursue a ship cutting across its front. It seems like it was a rare instance that that would work out in the solo ship's favor.  I agree with changing it to "towards the Tally" and will run it accordingly.

I struggle with the "towards" in those instances that you point out.  I worry that the "Direction" option will still lead to issues of ambiguity when the Tally is headed straight on or directly away.  I think the FFG designers did a clever thing using obstacles as the guiding point and trying out the "away from closest obstacle in front arc" will give the user a clear instruction.  I'll test some more tonight as things seem very close.  I'm only testing with TIE fighters at the moment but I'll expand into more complicated ships later on. Ideally, I'd like to get Epic ships in on the action but those might need individual AI systems and I'm ok with that since they would be more or less "Boss fights". 

Share this post


Link to post
Share on other sites
On 5/31/2020 at 7:57 PM, ImperialAce95 said:

The enemy Squad-building table seemed a bit off - thought I might change / add some stuff:

Faction

Patrol Ship

Reinforcements

Player Ship

Galactic Republic

Naboo N-1 / V-19 Torrent

ARC-170 Starfighter / Athersprite / BTL-B Y-Wing

ARC-170 Starfighter / Athersprite

Separatist Alliance

Vulture Droid

Hyena Droid / HMP Gunship

Belbullab-22 fighter / Nantex Starfighter

Galactic Empire

TIE/ag Aggressor / TIE/ln Fighter

TIE/in Interceptor / TIE/sa Bomber /
TIE/sk Striker / TIE/v1 Adv.Prototype

Assault Gunboat / TIE/d Defender /
TIE/ph Phantom / TIE/x1 Advanced

Rebel Alliance

HWK-290 | Z-95 Headhunter

RZ-1 A-Wing | BTL-A4 Y-Wing

B-Wing | E-Wing | T-65 X-Wing

Scum & Villainy

M3-A Interceptor / Mining Guild TIE/ln / 
Z-95 Headhunter

G-1A Starfighter / Khiraxz Fighter / BTL-A4 Y-Wing

Fang Fighter / M12-L Kimogila / StarViper Platform

Resistance

Fireball / Transport Pod

RZ-2 A-Wing / Resistance Transport

T-70 X-Wing

First Order

TIE/fo Fighter

TIE/sf Fighter

TIE/ba Interceptor / TIE/vn Silencer

Add upgrades as you see fit, I guess

IMHO you have picked far too "fancy" ships as the player ships. I think there is a reason why FFG proposed bread and butter forward arc fighters  like T65, Kihraxz, TIE Adv, T70, Belbullab with limited repositions and action economy. The only exceptions in their original table is Republic and FO, as these factions have no true bread and butter ships, thus proposing as alternatives a gunship-or-Aethersprite and the forward/rearward SF, respectively. Your table introduces  some very fancy extreme repositioners like Phantom, Starviper, Silencer, TIE ba and specials like TIE/D. I do not think an AI like FFG's simple model can handle them and is rather chanceless against such ships as player ships.

See also @GuacCousteau experience with Aethersprites against Z-95s.

 

-------

On 6/1/2020 at 8:26 PM, OoALEJOoO said:

Namely that ships should not determine maneuvers post the planning phase.

 

You keep repeating this. Why should they not do it?

On 6/1/2020 at 8:26 PM, OoALEJOoO said:

also because I enjoy a good discussion

:)

 

Do we want to

a) have a solo system which emulates the 2 person play by setting a pseudo dial already in the planning phase, thus being close to the original, but needing to either have a complicated AI with a lot of cross-referencing or being simple but pretty brainless system AI

b) accept that the AI does its maneuver determination for its better ships first after your ship(s) moved, thus deviating from a flow which would happen in a 2 person game, but generating a much more challening game?

The argument "the AI can be gamed as it is predictable" applies to your system as well, as the tables end up in not so different maneuvers as well.
 

Share this post


Link to post
Share on other sites
1 hour ago, Managarmr said:

You keep repeating this. Why should they not do it?

:)

Do we want to

a) have a solo system which emulates the 2 person play by setting a pseudo dial already in the planning phase, thus being close to the original, but needing to either have a complicated AI with a lot of cross-referencing or being simple but pretty brainless system AI

b) accept that the AI does its maneuver determination for its better ships first after your ship(s) moved, thus deviating from a flow which would happen in a 2 person game, but generating a much more challening game?

The argument "the AI can be gamed as it is predictable" applies to your system as well, as the tables end up in not so different maneuvers as well.
 

I've found that having the AI run through its maneuver logic during the planning phase offers the most elegant solution. Consider this:

  • During the planning phase, none of the ships have moved. The AI could pick a target, then run a single logic to try to address its target.
  • If the AI maneuver logic is run during the activation phase, then some of the enemy ships might have moved already and some not yet, this is particularly the case of a mid-initiative ship. Then the AI faces a dilemma, should it target a ship that has already moved and therefore its easier to aim? This would mean it would always give priority to lower initiative enemies. But what about that other high-initiative enemy ship that has not moved yet? Depending on the situation, it might be best for the AI to target these. How do you track an enemy that has a known position vs. another one which is yet to move? You are left with potentially several maneuver logic sets for the AI to run through in order to be effective. I've tried this method and it required a lot more bookkeeping and still didn't achieve a comparable AI performance.

Bookkeeping overhead is a big con, I think we can universally agree with that. A fundamental rule of AI system design is to place bookkeeping where it gives you more "bang for your buck", i.e. you get maximum AI results for a given bookkeeping amount. In my experience, it is best to invest bookkeeping on nailing the AI maneuver logic. Nailing the maneuver logic during the planning phase gives you this bang for your buck imho. If you get this right, then the following Actions logic is simpler (less bookkeeping). The AI will have a better chance to be in a good spot during the attack phase thus simplifying the Attack logic as well. It is a cascading effect of progressively managing bookkeeping: invest early, enjoy later.

This is the main reason I have stressed this point over and over again and I haven't really touched any other AI improvements (vectors, range, priority of actions, etc) because I think this is the starting point for a good AI, the rest comes after this. Another big plus of this approach is that it makes the AI play more by the same rules as the player, lending itself to be more expandable: it better follows normal rules, therefore human/AI tend to be more interchangeable which could be useful during campaign design, would require less patching to adopt content, etc.

As far as the Tactical Droid system being predictable. The issue is not much with predictability but with "exploitability". The TD system indeed has a degree of predictability, as you say the maneuver tables have some common patterns. It is less predictable than FFG's solo system though, and far less exploitable. The tables are built so that the AI will track a ship reasonably well, so that even if you do end up anticipating its move, at least the AI would have higher chances to shoot and probably won't be caught off-guard too badly. Note that even a human player will play along some patterns, thus predictability cannot be at a 100% even for a human player. Having said this, I believe that the holy grail of AI design would be not a system that is totally unpredictable, but a system that has the same level of predictability of a human player. Predictability in and of itself is not a bad thing, it only becomes a bad thing if you can exploit it (another thing I've mentioned a couple of times on previous posts). For a given decision spot, among all the different possibilities, there is a narrower set of competent possibilities than of incompetent ones. Since the competent is a narrower set, then it comes that competence comes with a degree of predictability. The trick is to have the AI pick among the competent set of moves, instead of among the incompetent. There is no use if the AI were completely unpredictable and, in the process, become erratic and incompetent.

Share this post


Link to post
Share on other sites
Posted (edited)

Forgot to mention. Another alternative that I tried was to have all the AI ships move after the player. In this case the AI also runs a single maneuver logic and this requires less bookkeeping. The AI can be much more aggressive, efficient and difficult to beat since you've basically upgraded all AI ships to maximum initiative during activation. However, I did find this to be less fun. I got the feeling the game became too deterministic, i.e. things happen a bit too methodical which detracted from the flying in my perspective.

Still, it could be an avenue for exploration because it does indeed result in lower bookkeeping. You would still need an AI maneuver logic more sophisticated and less exploitable than FFGs solo system for it to shine though.

Edited by OoALEJOoO

Share this post


Link to post
Share on other sites

I was thinking about fun scenarios to try with this:

Poe vs the First Order:

Turn 1: you spawn two TIE F/Os, one on each hyperspace point.  Each following turn, spawn one TIE F/O (you might need to proxy models) on the hyperspace point furthest from Poe.

Your ship is a fully loaded Poe.  Whatever upgrades you can legally fit on the ship, points don't matter.

How many rounds can you survive?  How many TIE fighters can you shoot down?

You can do this "Fox vs Hounds" type scenario with a lot of different ships, and it's very thematic.  Anakin vs Droids, Vader vs Phoenix Squadron, Han vs TIE Fighters, etc.

Share this post


Link to post
Share on other sites

I wonder if the simplest solution to the slamming-into-rocks problem might be a general rule along these lines:

If a solo ship's maneuver would cause it to overlap or move through an obstacle, roll one defence dice:

  • [Blank]: If you're unstressed, change your dial to your slowest bank away from the obstacle. If you're stressed, change your dial to the slowest blue bank away from the obstacle.
  • [Focus]: Avoiding rocks is for COWARDS.
  • [Evade]: If you are unstressed, change your dial to your slowest turn away from the obstacle. If you're stressed, change your dial to the fastest blue bank away from the obstacle.

I think this should prevent them hitting things a decent amount of the time, and if they're stressed they'll at least clear the stress even if they still clip the rock.

Share this post


Link to post
Share on other sites
2 hours ago, Ysenhal said:

I wonder if the simplest solution to the slamming-into-rocks problem might be a general rule along these lines:

If a solo ship's maneuver would cause it to overlap or move through an obstacle, roll one defence dice:

  • [Blank]: If you're unstressed, change your dial to your slowest bank away from the obstacle. If you're stressed, change your dial to the slowest blue bank away from the obstacle.
  • [Focus]: Avoiding rocks is for COWARDS.
  • [Evade]: If you are unstressed, change your dial to your slowest turn away from the obstacle. If you're stressed, change your dial to the fastest blue bank away from the obstacle.

I think this should prevent them hitting things a decent amount of the time, and if they're stressed they'll at least clear the stress even if they still clip the rock.

well the Blank is the supposedly the most aggressive attitude for solo ships so that should the "no fear" option.  After that swap with the Focus roll, I could see it being useful.  I don't think you even need to roll again as you already have their attitude locked in by the time you're throwing the maneuver down.  Using the current FFG AI, I hit rocks multiple times (at least 5 times).  Using PHeaver's modified AI, I only did it twice.  I'll be playing a few more games to see if the PHeaver method needs this but it sounds like a good back condition if obstacles are a problem.

Share this post


Link to post
Share on other sites
Posted (edited)

Tested it with 2 of my children.

2 T65 (each flown by one) against 2 Academy TIE l/N 2 Black Sqd TIE l/n, 2 Gamma Sqd Bomber with Proton torps. I did the AI.

The issues we encountered:

-Random switch between Defensive/Balanced/Aggressive was bad as expected. E.g. enemy ship finally managed to get behind in a good position, to then suddenly go to defensive and throwing away its opportunity to press the advantage home. Overall my fickle green dice which never can evade normally did soooo many evades for the maneuver/attitude selection, lopsiding the game.

-Gamma Sqd bombers had a bit of a problem acquiring and using the locks for the prot torps. Often too defensive to get a lock in the first place, or when finally having a lock and steering towards a good attack position, suddenly going defensive and steering away instead of advancing to smashthe player with a torp.

-SOLO ships diddling around at the board edge, flying into it several times. This might be exploitable, by leading the enemy to the edge, giving them disarms and knowing where they will show up.

-SOLO ship has suffered a critical hit, does repairing the crit take precedence over the normal action selection? We played it so.

 

Gonna test once more with exact same sqd compositions with the FFG AI, and then with Paul's proposal.

 

 

Addendum: Minor point: While fancy to look at and eye candy, overdesigned rules documents are not a good idea. I wont print an early alpha wasting that much ink, with many pics. So i use a tablet (and not a cheap one). But when hopping between pages, there almost always is a slight lag while the pages of the pdf are build up. Quite annoying when playing a longer time.

Edited by Managarmr
Addendum

Share this post


Link to post
Share on other sites
1 hour ago, Managarmr said:

Addendum: Minor point: While fancy to look at and eye candy, overdesigned rules documents are not a good idea. I wont print an early alpha wasting that much ink, with many pics. So i use a tablet (and not a cheap one). But when hopping between pages, there almost always is a slight lag while the pages of the pdf are build up. Quite annoying when playing a longer time.

Indeed. It should have been 15 pages for explaining the rules in detail, then 1 extra page with all charts, rolls, listings, and reminders in concise form, so that you only need to use that as reference.

About the attitude, it totally feels like they had something more advanced in mind, but for this first version they decided to scrap it and just make it random.
Attitude should not be random, but follow some logic. 

Either a ship is always aggressive (with orders to destroy an objective at all cost), always defensive (objective ships trying to escape the board), or it must depend on the context (escorts are passive until attackers get close to their escortee; patrols are aggressive until they get damaged, then turn defensive; attackers are aggressive, unless they have no-one in arc, and they are in some enemy's arc, then the behave defensively).
Changing attitude just because a roll makes no sense, unless the goal was to have a schizophrenic AI.

Share this post


Link to post
Share on other sites

The test rules being in a nice polished layout is a sales thing.  The design team normally wouldn't send out non-printer friendly early draft documents for the reasons described above.  If you check other companies though, rarely would they put out a beta rule set that was just a wall of text without layout. It looks terrible and can turn off people from actually trying out the rule set.  If you want to create a more printer friendly document, the entire PDF is created with selectable text.  You can copy and paste it into a word processor and clean it up for printing quite easily.  I actually liked it in a layout and didn't mind the one-off printing.

Regarding ship attitude, I can't really use the original scheme as it is too random as many have described but I have had better success with Paul's scheme.  I haven't tried complex ships yet so can't comment on their ability use target lock attacks but ships try to stay in arc and maintain an attack.  The constant switching of "offensive, defensive, or balanced" attitude isn't so much a strategic choice on the part of the pilot but a tactical one.  In a normal game, you aren't dedicating your own ships as solely aggressive and attacking at all cost, you are adjusting your approach depending on the layout of the board.  You may have a high-level strategy of sacrificing a certain ship to do maximum damage or holding a key piece back but you aren't going to do that blindly all game even when optimal choices dictate that you should start playing more conservatively.  The Solo Ship attitude is trying to simulate that.  It doesn't have a board state to judge the best timing of that but it also means you can't just bet that the ship coming straight at you is always going to play aggressive and charge in. The uncertainty is really all AI can offer unless you want to deal with chart up chart of action/re-action tables (or program an app to play them for you).

After removing the "turn in the direction the Tally ship is flying" component of Paul's scheme, I don't have ships randomly veering off an attack any more and find the attitude mechanic quite useful and elegant.

Share this post


Link to post
Share on other sites
17 minutes ago, spacemonkeymafia said:

The constant switching of "offensive, defensive, or balanced" attitude isn't so much a strategic choice on the part of the pilot but a tactical one.  In a normal game, you aren't dedicating your own ships as solely aggressive and attacking at all cost, you are adjusting your approach depending on the layout of the board.

Thing is that playing versus AI, especially with the limitations that "blind" AI like this has, isn't the same kind of game as playing against a human. It can't be.

In simple terms, playing against a human is playing chess, while playing versus blind AIs is playing horde survival: a game of attrition versus a horde of semi-mindless drones, where the main skill is not to outguess your opponent and counter their play, but to manage your resources (hitpoints, charges), and balancing your risk/reward ratio to survive as long as you can, or until the mission objectives are complete, if any.
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.

So I do think AI ships should have one particular attitude that the player needs to react to and adapt, often having to plan ahead how to act depending on which attitude has which AI ship.
Of course, you can play chess solo, and you can also attempt to have that here, making a supercomplicated AI system that is tied to the same rules and follows the same reasonings as a player would. But that would be definitely much more complex and slower to simulate, and I am not sure it would really be the right direction for this.

Share this post


Link to post
Share on other sites
11 minutes ago, Azrapse said:

So I do think AI ships should have one particular attitude that the player needs to react to and adapt, often having to plan ahead how to act depending on which attitude has which AI ship.
Of course, you can play chess solo, and you can also attempt to have that here, making a supercomplicated AI system that is tied to the same rules and follows the same reasonings as a player would. But that would be definitely much more complex and slower to simulate, and I am not sure it would really be the right direction for this.

Well the beauty of a solo rule set is you can play it anyway you want so I can see running it as you describe: Ship X is always offensive and the others are defensive (or whatever).  I'd be afraid things would be too predictable but I haven't put it to the table so can't say for certain.  It may be that the "horde" idea is the best this system can offer if you are looking for a challenge.  I'll keep trying Paul's method to see if/how it can be further refined.

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


×
×
  • Create New...