Turning SMAX back into strategy game

[bvanevery]

Why do we waste time discussing small inconsistencies in phrases?

Well I think we've worked that out now.  I think we agree to discuss whatever, and not see things as some huge big deal.

[Alpha Centauri Bear]

To compensate for that and to aid AI I increased sensor bonus to 50%.

Interesting to see how that turns out.  I've often stressed about whether I should knock out Sensor Arrays before beginning an assault.  Usually I default to not doing so, because that's going to be my base in a minute.  But if I really feel I've brought a marginal force in, I take out the Sensors.

Yes. Sensors can be destroyed. That's an element of the war. Sometimes you can, sometimes you cannot. Especially if they are deep behind the base. What's you gonna do? That's the game and tactics as it is.

[Alpha Centauri Bear]

I removed penalty for ship caught in port. I don't really understand this. Sea battles are insanely skewed toward attacker even more than land ones.

I guess I should watch some sea battles to verify how this works.  Also I wonder if ranged ship battles, with ship artillery, are different from direct attack ship battles.

The "ship in port" penalty is only when non sea unit attacks ship in base. So ship to ship duel is unaltered.

[Alpha Centauri Bear]

Psi combat equalized.
1,1,     ; Psi combat offense-to-defense ratio (LAND unit defending)

This is a dramatic change.  It means that piles of mindworms heading for your bases, aren't going to bother you much.  A mindworm apocalypse just became a lot more survivable.  It also means, you will have trouble fishing in the fungus for planet pearls.  I wonder if it's going to penalize +PLANET factions when they try to capture wild mindworms?  Usually you just catch or kill the next one with your own healed up mindworm.  If you actually have to worry about losing your mindworm in the exchange, that's a problem.

Your change is also lore destroying.   

The hell with the lore. Playability is the king!

I've already played few of my games with 1:1 psi combat. Feels pretty good. At least I am not blindly poke fungus with every one of my units. I either chose promoted or bump Morale/Planet or hide in bases and build more defenders if I am on Free Market, etc. The nature of the game stays the same but it become more filled with decisions and thinking.
In short, I become more cautious about waking up worms but my colony pods sometimes survive sudden native attack. Just a slight shift in feeling and tactics. To better, I guess.

[bvanevery]

I guess you meant to amend my method #4 by introducing some preference variables. That may work and make it a little better. However, I wouldn't bother with it as neither of these methods is precise anyway. I did mine just because it was easy to do (a matter of minutes) and it gave me feeling of slight improvement over #3. That's it.

Aren't you afraid of GIGO?  How can you make any determination about anything, if you are not modeling the way blind research actually works in the game?

I actually think your approach better if strict succession is imperative. I just don't know if said strict succession is crucial for playability. I would guess not.

What actually tends to happen in the original game, is you get rapid spurts of pointless weapon improvements.  Oh look, it's the strength 10 weapon.  Now the Strength 12.  Now the (ridiculous!) strength 13.  Except they don't actually all come in that order anyways, they sort of jumble around.  So you often get the 13 first.  I guess you could think of it as the "10 to 13 pointless cluster" of weapons.  It clutters my Unit Design Workshop and has no other gameplay value.  I got rid of that.

There's just a lot of goofiness of weapons and armor jumping around in the original game.  And imbalance.  Strength 8 Chaos guns going against strength 3 Plasma armor is very common in the stock game.  Add Fusion reactors and oh goody.

Another philosophy of the stock game that I oppose, is "every faction gets their turn having a reward".  No!  Deirdre gets mindworms, she should not be getting a strength 6 missile launcher from Synthetic Fossil Fuels!  It's stupid!  Morgan's Silksteel Alloys isn't as bad, because at least it's defensive.  He's presumed to be turtling up and building in a Free Market.  The Civ equivalent is building Phalanxes, then Musketeers.

Your imperfect simulation is almost right for the wrong reasons.  I see full evidence of a lot of random crap in the stock game.  There's a pretty big window that the pass filter of random crap falls over, as the tech tree progresses in time.  It approaches the point of being unstructured, which means random stuff tips the balance of power one way or the other.  I prefer a game that is cogent and understandable, because then I can tweak that design.  I believe the tweaking is where design quality actually comes from.  The opposing "design" is "hey here's a bunch of random crap.  Deal with it!" Well you might get a satisfying experience out of that randomness, or you might get a stupid frustrating one.  I'm not really into rolling the dice on whether a player is going to enjoy the game or not.

Like how are you even going to achieve what you want out of your "weapons and armor to create strategy" progressions, if you allow so much random noise to enter the calculation?  "This game I got lucky and tore everything up."  Wow, what a game design.  Not saying you've done this, I'm saying this is what "throw it over the fence" designers do.  They put their faith in random numbers, in the players themselves... it's stupid.  It's like crowd sourcing a screenplay or a novel, doesn't work.

How do you verify?  Do you play your mod a few times and think / hope you improved things, letting your own cognitive biases take over?

When you get a game in front of other players, at least they might tell you whether you're meeting your stated design objectives or not.

One of my objectives for instance is improving the perceived competence of the AI's performance.  I hope I achieved that.  A few players said I achieved that, so that encourages my hope in that area.  But even with those encouragements, I've had such a small testing audience, it's not so easy to say.  I know I haven't made a silk purse out of a sow's ear.

[Alpha Centauri Bear]

Resonance Laser,      R-Laser,        9, 1,11, -1, Bioadap,

Bioadaptive Resonance is a relatively early tech.

Look at my estimate. It has index of 32 out of 85 which is about 40% down the line. It's more like middle, not early one. On average you get it even after Fossil. Based on this it is placed right.
I forgot about Ogre equipped with this weapon, though. Well this is a first draft there be a lot of misses. We can think how to deal with it later or just swap it for some earlier technology and lower the strength.

[Alpha Centauri Bear]

Plasma Steel Armor,  Plasma,      5, 2, 5, Chemist,

Pulse 3 Armor,       3-Pulse,     6, 1, 6, AdapDoc,
Resonance 3 Armor,   3-Res,       4, 1, 5, FldMod,

I thought about spreading out the strengths of the "red" armors, but I thought it might mess too much with people's sense of armor value.  Could be a lotta "doh!" from players, asking them to check on attacker's numerical/letter ratings to determine whether to attack or not.  User unfriendly, I figured.

Hm. I always turn on combat odds confirmation dialog and can clearly see all battle parameters starting with weapon/armor and down to all modifiers. Then I decide whether attack or not. Never bothered color guessing armor.

[Alpha Centauri Bear]

Pulse 8 Armor,       8-Pulse,    24, 1,11, Solids,
Resonance 8 Armor,   8-Res,      20, 1,10, SentRes,

You should rename these if they're not actually strength 3, strength 8 anymore.

Good idea. I didn't touch much these special items yet. Can I just rename them in file?

[Alpha Centauri Bear]

I guess you meant to amend my method #4 by introducing some preference variables. That may work and make it a little better. However, I wouldn't bother with it as neither of these methods is precise anyway. I did mine just because it was easy to do (a matter of minutes) and it gave me feeling of slight improvement over #3. That's it.

Aren't you afraid of GIGO?  How can you make any determination about anything, if you are not modeling the way blind research actually works in the game?

Sorry, I didn't understand what I should be afraid of? I can even share my code to show how did I do it. In short, there are some number of technologies available for research. So I randomly pick one and "research" it. That changes list of available technologies and so the cycle continues until all are researched. I run this many times and gather statistics for each technology at which step it was researched. Then I averaged it and this is my average sequence number for given technology.
This is not how blind research actually works in the game, you are right. Again, I chose wrong word. I should say random research assuming any available technology can be researched on next step. I don't think modeling blind research mechanics exactly would do cause any significant change but I am willing to try if anyone give me a procedure how it works.

[bvanevery]

The hell with the lore. Playability is the king!

This is true, in my design universe at least.  However, I do advocate looking for opportunities to repurpose lore, rather than just ignore it.  I've "harvested" various technologies in the tree to use for something else.  "Frictionless Surfaces" for instance, has the Deirdre voice narrator going on about a Morganic researcher who talks about tricking substances into hiding from themselves.  For quite awhile I used this as a basis for Clean Reactors, because I didn't need a Cloaking Device to be living in its own tech.  I put that in with Photon Walls, figuring hey, you're manipulating light, should be able to get a cloak out of that.  So then I changed it to "Single-Sided Surfaces" and pretended that a Clean Reactor is something you get from a Klein bottle of containment.

Although recently, I threw Clean Reactor into Bio-Engineering, where it originally came from!  But that tech has moved around a lot over all these iterations, so it's only happy accident that it has landed close to where it started out.  Longevity Vaccine stays at Eudaimonia, which is Tier 6 in my mod and expected as a viable SE choice for 1/2 the game, so it's not exactly the same Bio-Engineering.

Anyways Cloaking Device is now in Single-Sided Surfaces, which I could restore to Frictionless Surfaces.  But I won't, because my version of the tech name has grown on me, and I might prevaricate again.  It's like my little thumb print in the clay of this world.

I also have Phasers.  Renamed Fusion Lasers.  I realized the other day that if you acronymed that, you'd get FLASER, very similar in pronunciation to Phaser.  So I decided to keep Phaser.  And of course the real reason, is I'm paying homage to Star Trek TOS.  Back when weapons were weapons and everybody had one!

I've already played few of my games with 1:1 psi combat. Feels pretty good.

Ok, I'll suspend judgment until I've seen it in action.  As a +PLANET faction.  You might have crippled the Gaians, but what if you've only changed their play style?  Maybe this is the new defensive unit to bring to the enemy's front door.

At least I am not blindly poke fungus with every one of my units.

This is the thing.  You've worried a lot about players doing this.  Do you think it's something lots of players do?  Or is it just something you did a lot of?

I've poked fungus as Chairman Yang, in order to work up the 40 credits necessary to switch to Police State.  Then I stop, because I got what I want.

Usually when I've poked fungus, it's because my mindworms got stuck on an island somewhere, with no Transport to get them anywhere else.  They could be there for 100 years!  In past games I've marched them around for the entire time, to stir up more recruits.  I have not gotten any vast yield for the effort spent.  And for the mouseclicking effort spent, my real life work as a player, I think it's ok if I get something out of it.  I certainly didn't get anything game breaking out of it.

The exploration money is not in mindworm harvesting.  It's in popping supply pods.  I've gone through vast cycles of AAR writing where I did no supply pods at all.  Too much of an easy crutch to victory.

I've never captured a highly effective army of mindworms.  They've managed to harass long distance enemy cities, that's about it.  They tend to die off.  They're much more effective when brought home for defense.  They can really mess up an interloper coming over a patch of fungus.

[Alpha Centauri Bear]

I actually think your approach better if strict succession is imperative. I just don't know if said strict succession is crucial for playability. I would guess not.

What actually tends to happen in the original game, is you get rapid spurts of pointless weapon improvements.  Oh look, it's the strength 10 weapon.  Now the Strength 12.  Now the (ridiculous!) strength 13.  Except they don't actually all come in that order anyways, they sort of jumble around.  So you often get the 13 first.  I guess you could think of it as the "10 to 13 pointless cluster" of weapons.  It clutters my Unit Design Workshop and has no other gameplay value.  I got rid of that.

There's just a lot of goofiness of weapons and armor jumping around in the original game.  And imbalance.  Strength 8 Chaos guns going against strength 3 Plasma armor is very common in the stock game.  Add Fusion reactors and oh goody.

Another philosophy of the stock game that I oppose, is "every faction gets their turn having a reward".  No!  Deirdre gets mindworms, she should not be getting a strength 6 missile launcher from Synthetic Fossil Fuels!  It's stupid!  Morgan's Silksteel Alloys isn't as bad, because at least it's defensive.  He's presumed to be turtling up and building in a Free Market.  The Civ equivalent is building Phalanxes, then Musketeers.

Your imperfect simulation is almost right for the wrong reasons.  I see full evidence of a lot of random crap in the stock game.  There's a pretty big window that the pass filter of random crap falls over, as the tech tree progresses in time.  It approaches the point of being unstructured, which means random stuff tips the balance of power one way or the other.  I prefer a game that is cogent and understandable, because then I can tweak that design.  I believe the tweaking is where design quality actually comes from.  The opposing "design" is "hey here's a bunch of random crap.  Deal with it!" Well you might get a satisfying experience out of that randomness, or you might get a stupid frustrating one.  I'm not really into rolling the dice on whether a player is going to enjoy the game or not.

Like how are you even going to achieve what you want out of your "weapons and armor to create strategy" progressions, if you allow so much random noise to enter the calculation?  "This game I got lucky and tore everything up."  Wow, what a game design.  Not saying you've done this, I'm saying this is what "throw it over the fence" designers do.  They put their faith in random numbers, in the players themselves... it's stupid.  It's like crowd sourcing a screenplay or a novel, doesn't work.

How do you verify?  Do you play your mod a few times and think / hope you improved things, letting your own cognitive biases take over?

When you get a game in front of other players, at least they might tell you whether you're meeting your stated design objectives or not.

One of my objectives for instance is improving the perceived competence of the AI's performance.  I hope I achieved that.  A few players said I achieved that, so that encourages my hope in that area.  But even with those encouragements, I've had such a small testing audience, it's not so easy to say.  I know I haven't made a silk purse out of a sow's ear.

Totally agree with your approach. Very well put.
Let me reiterate that I do not insist on having it my way. I can as well work it out of your tree easily. May got even better results. Keep in mind that I did not focus on enhancing tree per se. I just wanted to illustrate the technique of how items can be a little better distributed on any tree. So certainly, combining this approach with well thought tree may bring even better results.

[Alpha Centauri Bear]

I've already played few of my games with 1:1 psi combat. Feels pretty good.

Ok, I'll suspend judgment until I've seen it in action.  As a +PLANET faction.  You might have crippled the Gaians, but what if you've only changed their play style?  Maybe this is the new defensive unit to bring to the enemy's front door.

I think it is quite opposite. Check out the discussion about SMAC combat outcome skew. The closer you are to 50% winning probability the more important are small changes. When you are at 3:2 odds you wins near 100% of the cases regardless of other small modifiers. So it doesn't matter whether you are +1 or -1 on PLANET. When you are at 1:1 odds, any strength modification quintuples!!! So shifting initial balance of psi combat to 1:1 actually improves Gaian's winning chances by 50% comparing to factions without this benefit. That was my whole point.

[bvanevery]

Sorry, I didn't understand what I should be afraid of?

GIGO = Garbage In, Garbage Out.  An old engineering acronym, I think.

So I randomly pick one and "research" it. That changes list of available technologies and so the cycle continues until all are researched.

This is not how either blind research or directed research works in the game.  Research probability is weighted by dot product (Conquer Discover Build Explore) * (power tech wealth growth) for the given tech.  Choosing Conquer is going to make a huge difference for gaining weapons, for instance.  This is part of why Santiago is going to kill Morgan in the stock game.

I don't think modeling blind research mechanics exactly would do cause any significant change

I predict your conclusions on research frequencies are currently very wrong, as a matter of what factions will actually gain.  And that the only way you can actually verify whether your approximation is sufficient, is to do the exact calculation and compare it.

but I am willing to try if anyone give me a procedure how it works.

Dot product to get the weights.  Add up all the weights of available techs, that's your denominator.  (weight / total weights) = probability of this particular tech.  RNG in the range 0.0 to 1.0, map it to the weights somehow.  Most straightforward is probably is to loop through the available techs, accumulating a sum, until the sum is >= the RNG.  Potential endpoint error of 0th or Nth tech left as exercise to the programmer.

Directed research is just find the highest valued weight.  In case of a tie, choose.  Random choice might be better than building in any kind of list order bias.

This stuff is exactly what Induktio and I had a big fight about a few months back BTW.  I've spent a lot of time working on these weights.

If you only care about how the 14 stock factions behave, then you only need to look at the unique EDBC weights offered by those factions, plus whatever a real human player is likely to do.

Explore - Gaians
Discover - University
Build - Morganites, Free Drones
Explore, Discover - Peacekeepers, Cybernetic Consciousness
Explore, Conquer - Believers, Cult of Planet, Pirates
Discover, Build - Data Angels
Discover, Conquer - Spartans
Build, Conquer - Hive, Usurpers
Explore, Discover, Conquer - Caretakers

Interestingly, nobody uses a pure Conquer strategy in the stock game.  The Spartans and Usurpers do in my mod.  Personally I think that choice is important because a real human player will probably make it at some point.  Nobody happens to choose Explore, Build either.  Humans will do it though, it's typical to pick either 1 or 2 research foci in practice.  I think it would be reasonable to ignore the Caretaker way of doing things; they're doing directed research anyways.  So for blind research, that's 10 cases to analyze.

Additional corner case: if all weights after dot product are 0, then I'm supposing one of the next available techs is chosen at random.  Ironically, this corner case is what you're currently modeling.  Although, available techs might not be considered equally valuable even then.  Could add up the (power tech wealth growth) weights.  That's equivalent to dot product with Explore=1, Discover=1, Build=1, Conquer=1.  EDBC choice is just a mask over the weights, that's another way of looking at it.

It is really an open question to me, whether all this machinery for the Blind Research game mechanic, is "worth it".  It certainly is esoteric from a mod maintenance standpoint.  How many people are going to get that deep into the weeds?  For a new game, I would think of something else.  If you do this statistical analysis, I'd be mildly interested in how same or different these tech trees actually look, under the various weights.  Maybe it matters, or maybe it's a variation of the thousand bowls of oatmeal problem.  Parametric distinctions without enough of a difference to justify them. 

[bvanevery]

So shifting initial balance of psi combat to 1:1 actually improves Gaian's winning chances by 50% comparing to factions without this benefit. That was my whole point.

That doesn't make sense.  2:1 favors offense with psi, not Gaians.  Gaians will be on offense or defense in any given situation.  If they are trying to trash other factions, you've taken away their teeth.  Trashing someone is an important way to defend yourself.  The Romans were famous for just gradually wearing down opponents that couldn't do anything about them.  I've done that to all kinds of AI factions myself.  Hold, hold, hold, until you sense that they're not managing to supply enough units anymore.  They've spent themselves on your walls, fruitlessly.  Now push.

Well, you may have an overall preference for "WW I trench warfare stalemate" as the way Civs should develop.  We'll see.  It'll all depend on how it actually plays.

One of my major concerns, is whether the AI is written with the embedded assumption that mindworms are strong on offense, weak on defense.  I don't know that it is, but I wouldn't be surprised if it is.  This sort of issue can derail your plans.  All you can do is test and find out.

[Alpha Centauri Bear]

but I am willing to try if anyone give me a procedure how it works.

Dot product to get the weights.  Add up all the weights of available techs, that's your denominator.  (weight / total weights) = probability of this particular tech.  RNG in the range 0.0 to 1.0, map it to the weights somehow.  Most straightforward is probably is to loop through the available techs, accumulating a sum, until the sum is >= the RNG.  Potential endpoint error of 0th or Nth tech left as exercise to the programmer.

Directed research is just find the highest valued weight.  In case of a tie, choose.  Random choice might be better than building in any kind of list order bias.

If you only care about how the 14 stock factions behave, then you only need to look at the unique EDBC weights offered by those factions, plus whatever a real human player is likely to do.

Explore - Gaians
Discover - University
Build - Morganites, Free Drones
Explore, Discover - Peacekeepers, Cybernetic Consciousness
Explore, Conquer - Believers, Cult of Planet, Pirates
Discover, Build - Data Angels
Discover, Conquer - Spartans
Build, Conquer - Hive, Usurpers
Explore, Discover, Conquer - Caretakers

Interesting. Sure I can try that - piece of cake. Just didn't follow your explanation exactly to the letter. Is it published somewhere on the Net?

I can keep adding factors but there still be much more factors to it. Like below
There is a special formula controlling whether a technology ready to be researched is actually appearing in the list of available technologies. About 1/3 of the time they do not.
Factions also quite often acquire technologies other way: given at start, pops, trade, steal, extort. This messes up with the above formula and research priorities.
Technology cost changes non trivial with advancement as well as with relative number of technologies other factions discovered. Faction that fell behind in research get them to research a little cheaper.
I don't even mention the actual speed of research. I.e. mapping research sequence to actual game turns. That is pretty impossible to model but this is what we ultimately need.

Again, I would love to try modeling the way they are researched in the game. This will be fun to see how it changes. However, with all above factors it will still be quite not exact. As I said, I didn't want it to be exact. I just factored in a little bit more logic hoping it would be a little bit more close to reality. The process can continue indefinitely.