Author Topic: Schooled By My Own AI...  (Read 2575 times)

Offline x4000

  • Chris Park, Arcen Games Founder and Lead Designer
  • Arcen Staff
  • Zenith Council Member Mark III
  • *****
  • Posts: 31,284
Schooled By My Own AI...
« on: January 16, 2010, 02:44:18 AM »
Right, so I coded this here AI, for those who don't know me.  That perhaps makes this story all the sadder. :)  Short version: we literally lost in 42 minutes.

I tend to play on difficulty 7, and I tend to win around 80% of the time.  Give or take.  Mostly I play 3 or 4 player co-op with my alpha testers, which are my dad, my uncle, and my uncle's colleague.  We tend to play about twice a week, or less when someone is really busy (in which case just whoever is left, if there are at least 2 or 3 of us, plays).  We've actually been doing this with various RTS games since 1998, all the way back to the very first Age of Empires (I had already been hooked on Warcraft I and II, but then that was my first time getting my dad and uncle into RTS -- but, they'd always liked Risk, and Chess, and so forth).

Anyway, but I really digress.  The point is, they and I have been playing AI War since it was a week into development, back in November 2008, and all the way up through alpha and beta -- we know the game perhaps better than anyone else (not to say we are collectively better at it than everyone else, but we know it inside and out), and usually do pretty well.  Last four or five sessions had been spent on a map where we'd made several key mistakes and had started in a bad position, and so we'd lost three of our four home planets.  We had about 12 out of 80 planets, but were really stuck in many respects.  This week we decided to just call that one a loss, and move on, since there were a lot of new expansion features (minor factions) that we wanted to turn on, anyway (otherwise we would have stuck the hard one out -- we normally love that sort of protracted slugfest where defeat seems certain).

So tonight we started a new game:  Grid, with a three-planet start (my uncle's colleague was away), and Random Non-Technologist AIs on diff 7.  All of the minor factions except marauders and miners.  We tend to like starting planets with relatively few hostile holes on them, so this was an interesting start because I had four hostile wormholes, my uncle had I think three, and my dad had two.  One of the neighboring planets (to all three of us) had a whopping 10 connections in total.  My uncle was dead-set on wanting teleport battle stations, and I chose deflector drones, which were next to him.  My dad wanted the low number of adjacent planets next to him above all else, and so he chose shield boosters, which are helpful anyway, if not overly exciting.

The game starts, and the first thing I do is start scouting while also cranking out little military ships.  I'm the unspoken designated scout.  My dad and uncle are also cranking out ships of various sorts, and we're all unlocking our starting technologies.  So far so good.  Two planets that are adjacent to all three of us, and which would connect my dad to us (and completely buffer him from adjacent enemies), so those seem like natural places to start.  As luck would have it, those are the weakest of the nearby planets, too.  Yes!  They are still somewhat formidable, though, with around 100 ships on one, and around 150 on the other.  And looking at those ships... it's clear that our light brown AI is an experimentalist, given all the experimental ships.

That makes it a bit tougher.  Just doing my due diligence, of course I am still sending out scouts all over the place, and marking priorities as new planets are discovered.  Within minutes, it is plainly obvious what the other AI type is, as well: Alarmist.  There are alarm posts all over the place on dark brown's planets.  Lovely.  That's about the worst AI type we could pull for a grid or hub map, given all the adjacencies.  We all kind of chuckle, thinking this is going to be an interesting game.  Sounds like a fun one.  And it was -- but also short.

Scouting reveals that pretty much any planet we attack also has multiple alarm posts next to them.  Also, my uncle's and my planets are like swiss cheese, with tons of ingress wormholes.  So, the main focus is to protect ourselves as best we can, prevent ourselves from losing our home command stations, and we'll just rebuild the harvesters after the inevitable waves.  We're too poor to use exo-shields, we need those resources to power reactors and keep churning out a bunch of ships.

By this point I have around 300 ships, and my dad and uncle each have around 150 or so. My deflector drones are cheap, some of theirs (tele-stations in particular) are not so much, so hence the difference.  The three of us do a simultaneous strike on the eastern planet that borders all three of us, and it's over in minutes.  We're perhaps 10 minutes into the game, and this first planet was cake.  Of course, that sets off about two alarms, jumping our threat to around 500 (some of which are Mark III), but so it goes.  I leave some of my forces on that planet, my dad leaves all of his forces there, and the other half of my forces fall back, as do my uncle's. 

My dad has been steadily working on very good turret defenses on his planet, along with doubled Mark I force fields, so he is semi-secure.  I have a few turrets, but still just the one force field, and not really that much a sense of security.  But, I've got 500 or so mobile guys now, and I've recalled a bunch of them home.  The threatening AI ships strike from a variety of different directions, and it pretty much seems like a wash -- we've got this thing, so far.  On my planet the sheer number of my ships take care of those that come in, and on my uncle's planet it is much the same story, overall.

Except.  Three or four mark III bombers from that alarm manage to slip past all the rest of the carnage on his planet, and before we notice they have his puny little forcefield pretty much down.  He notices, and sends the rest of his fleet (which had temporarily not been FRD) to defend.  I'm seeing the potential danger just then as well, since I'd happened to switch to his planet to see how things were going with the 8 ships I saw were remaining.  When I saw those bombers and the force field, I realized I should send some ships to help him guard his command station from the next wave, when it hit.  There had already been a wave or two (in batches of three, of course), but they were not terribly notable yet.  With his forcefield down, though, he might need some extra firepower.

Boooooooooom.... and a flash of white.  My uncle had just been commenting over voice chat about how some little bombers had gotten through and were actually hitting him.  "What was that!?" asks my dad, of course knowing exactly what it was.  "Well," I say.  "Economically-speaking, at least you're better off now, anyway.  With the home station core, and the ability to rebuild a new command station there, you're much richer now."  Even if we were not 1/3 of the way towards death at the 20-minute mark.

So, we rebuild and refortify.  I put in way more turrets, but still not a second force field (those things are pricey).  More waves come in, and two of these are just 18 electric bombers.  Of course, that's nothing to scoff at at all, despite the difference from the 100s of ships we'd seen in the other waves.  Those ships come in, do a fair bit of damage, but we prevail fairly handily.  It's early yet, after all, even if there are alarms everywhere.

I decided it was time for me to take the western planet, before it got too much larger.  So over I go, and it's really not too bad.  Still only a couple of hundred ships there -- so early in the game, still.  I knock all those off with ease with my 500 ships, and there's still a force of around 70 ships of mine on my home planet.  This planet, too, is full of experimental stuff -- speed boosters, decoy drones, experimental engineers, etc.  I wouldn't realize the significance of this until a few minutes later.

I cleared all of the guard posts, ignoring the wormholes because there were a trivial number of ships there.  Then I decided to drop in and kill the command station, despite the neighboring three alarm posts -- because, well, I should have been able to handle it with my 500 guys, and my ability to crank out more.  And if not, my dad and uncle easily had 500 guys between them and could bail me out.  And at the end of this, we'd have one more planet.  I think I discussed what I was doing very vaguely with them, but did not say that I was about to kill the command station -- that was all under control, right?

So I kill the command station, and the alarm goes off, and the guys by the wormholes scatter.  My guys go into FRD mode and hunt most of them down, and everything is happy.  The threat jumps to 600, but I still have 500 guys concentrated on this one planet, and if the enemy comes in batches I'll be just fine.  And if not, there's plenty of warning time for my dad or I -- plus a few dozen turrets each, force fields on both, and smaller mobile forces on each.  My move was not what I'd consider risky, normally, and it's the sort of move we routinely make without much discussion.  Some things require discussion, others do not.

The first few hundred enemy guys come in, and predictably my FRD ships are mopping them up.  I'm keeping an eye on things to make sure that nothing slips by, and then notice that there are about 70 guys on my home planet.  Huh!?

Turns out one of the adjacent planets with an alarm also bordered my home planet.  Whoops.  Well, no matter-- my home force has grown to 100, and we outnumber the enemy.  Plus turrets and forcefield.  The enemy is mark II to my (largely) mark I, but that's okay.  I've seen worse. 

But.  I failed to account for the experimental ships.  With speed boosters and decoy drones, the AI zips past my defenses before I can really even shoot at them, under the cover of those blasted decoys so that I'm not really hitting them, anyway.  My force is largely untouched because the AI doesn't care; their force is largely untouched because I underestimated the experimentals in this engagement.  They have about 30 bombers who all plaster up against the force field, and... you know... boom.  I guess this must have been around 30ish minutes in.  The end wasn't far away now.

My dad is really surprised, of course, since he'd been focused -- successfully -- on mopping up the 80ish guys that had slipped into his system.  My death was as out of the blue as my uncle's.  Wrap up time on my planet: once my ships caught up to the enemy, I made short work of them.  Deflector drones are pretty awesome as a starting unit, incidentally.  Colony ships are heading from my dad's's planet to my uncle's and my home planets, but they'll never get built.

I haven't been watching the battle on the western planet at all -- I've been too busy supervising my own death.  I come back, and the planet is largely scrubbed of the enemy, except there are a few still swirling around from other wormholes every little bit.  My own death only took perhaps 5 minutes at the most, so the threat hasn't even come down that much -- still maybe in the 300s.  My own ships are down to something like 180 on the western planet.  Whops.  But the planet is pretty clear.  I don't know what happened to my guys, but clearly something stronger came through there while I wasn't looking.  So it goes, when you fight multi-front battles.  You just have to roll with it and move on, it's a non-event, completely normal.

Then the rest of the threat showed up.  Out of two wormholes at once come all 300-some enemy ships, most of them Mark III.  They ride on the speed of speed boosters, lapping up my pathetic defenders like milk.  I mean, it looks almost like a literal twister, with the raging winds of their ships swiping away the detritus of mine.  Seconds, literally, and my ships are gone save for one lone scout who watches the carnage from a safely-cloaked vantage.  He's going to be one mighty lonely dude in a few minutes, let me tell you.

Those Mark III ships swarm into my dad's home planet and it's over in seconds there, too.  As soon as they headed for his wormhole, I told him he didn't have a chance.  My uncle and I were moving the rest of our fleets there to make our last stand -- my uncle's teleporters arriving instantly, the rest of everything far too late to matter.  All those turrets the doubled force fields... twigs against the storm.  Less than a minute later, and the final boom.  Looking at the game clock, I could hardly believe it.  I've never lost so quickly in this game, especially not at difficulty 7.  Shoot.

My uncle had a respectable 300ish% kill-to-loss ratio, whereas my dad and I were both down around 100%.  Still not too shabby for having lost so spectacularly.

So what went wrong?  Obviously, we made some blunders in not defending ourselves well enough.  The importance of those enemy speed boosters also wasn't entirely clear until they had reamed us not once but twice on our own planets.  We didn't have any tractor beams up because of all the ingress points requiring too much crystal for that to be cost-effective.  It would have slowed us down, in other words.  In some senses, therefore, we were victims of our own rush.  But with the AI having so many alarm posts, rushing is a good idea because it keeps the penalties lower.  We were so starved for resources that we needed more planets, fast.

We also probably could have stood to collaborate a bit better.  Usually we're really good at that, but generally we are accustomed to all getting our bearings in a new campaign, taking a planet or two each, and then reassessing and working as a team.  It's worked for us for a long time, but not in this game primarily because of the alarms.  Usually the risk of taking an early planet is supremely low, which is why we don't coordinate on those.  But with alarms everywhere on this kind of map, it was like jumping straight into the middle of a game with inadequate territory, resources, and knowledge.  Ick.  No matter what we did, this was going to be one tough map to crack.  I think we may well have lost in any event, and the speed of this loss just saved us some time.

Afterward, we started up another map, this time on Vines, and are so far having a much easier time of it. 50ish minutes in and we've captured 3 planets, netutered/gate-raided a fourth, and are about to assault a fifth.  After the wide open spaces of the Grid, the narrow passages and constricting twists and turns of Vines are a welcome relief.  Of course, everything is further away in Vines, so I expect we will see many challenges on that map over the coming weeks.  But so far so good. :)
Have ideas or bug reports for one of our games?  Mantis for Suggestions and Bug Reports. Thanks for helping to make our games better!

Offline HellishFiend

  • Hero Member Mark II
  • *****
  • Posts: 758
Re: Schooled By My Own AI...
« Reply #1 on: January 16, 2010, 02:56:49 AM »
Wow! It sounds like you definitely made a few grand strategy errors, for sure.  ;)  My friend and I play very defensively, and while we're not used to playing at such a huge disadvantage, in that case we would have certainly opted for raid ships (perhaps both getting mk 1 and 2 with our starting knowledge), and focused on defense along with raiding alarm posts. I have no idea if we would have been successful in the end, but we definitely would have made it a long, drawn out struggle.  Like you said though, maybe that setup is simply inevitable defeat.

Great story.  :D
Time to roll out another ball of death.

Offline x4000

  • Chris Park, Arcen Games Founder and Lead Designer
  • Arcen Staff
  • Zenith Council Member Mark III
  • *****
  • Posts: 31,284
Re: Schooled By My Own AI...
« Reply #2 on: January 16, 2010, 03:09:54 AM »
Yeah, Raid starships would have been a good idea.  This is another example of why it is a great idea to scout before you unlock you early techs, too.  Raid starships are not part of our base build pattern, and so we were not able to afford the K once we realized what was up.

But yes, errors abounded.  That's the nice thing about AI War, at least -- if you lose, it's usually your fault.  As in Chess, if you lose, it's because you made some mistake.  Not because the AI was faster or simply invisibly did something that caused you to lose for reasons that you can't fathom.  It's because you didn't play the scenario well enough, and there are usually a string of visible errors (or gambles that turned out wrong, as the case may be).  I think the scenario could have been winnable, especially with raiding by Raid Starships (I had not even thought of that until you mentioned it), but it would have been really hard.  And using out general small-ships-focused tactics, that might well have been impossible.  But... raid starships to the rescue, could have been.
Have ideas or bug reports for one of our games?  Mantis for Suggestions and Bug Reports. Thanks for helping to make our games better!

Offline RCIX

  • Core Member Mark II
  • *****
  • Posts: 2,808
  • Avatar credit goes to Spookypatrol on League forum
Re: Schooled By My Own AI...
« Reply #3 on: January 16, 2010, 03:40:26 AM »
if you *ever*, and i do mean *ever* get a Zenith Trader to pass by, do at least pick up a Shield Inhibitor and Shield booster. With those online i was able to swallow a CPA of almost 2k ships!
Avid League player and apparently back from the dead!

If we weren't going for your money, you wouldn't have gotten as much value for it!

Oh, wait... *causation loop detonates*

Offline raptor331

  • Sr. Member
  • ****
  • Posts: 313
Re: Schooled By My Own AI...
« Reply #4 on: January 16, 2010, 12:40:37 PM »
if you *ever*, and i do mean *ever* get a Zenith Trader to pass by, do at least pick up a Shield Inhibitor and Shield booster. With those online i was able to swallow a CPA of almost 2k ships!

And most definately a Zenith power plant, those monsters can power a decent fleet by themselves(but they take forever to build and are extremely expensive)

Offline Oewyn

  • Sr. Member
  • ****
  • Posts: 330
Re: Schooled By My Own AI...
« Reply #5 on: January 18, 2010, 12:47:47 AM »
The Z Power gen take over 3 hours to pay for themselves once they are built.  I wouldn't say they are a godsend, if you have a large influx of resources and nothing better to spend them on, then I guess you could build one.  It's a very long penalty (of 80 metal/crystal per second for 300+ minutes) for a moderate benefit (300k energy for 10 metal/crystal per second).  It's like building a nuke in Supreme command (Forged alliance where you couldn't assist the construction of the missiles)  you /can/ build them, but by the time they are finished, do they really help you?

Offline x4000

  • Chris Park, Arcen Games Founder and Lead Designer
  • Arcen Staff
  • Zenith Council Member Mark III
  • *****
  • Posts: 31,284
Re: Schooled By My Own AI...
« Reply #6 on: January 18, 2010, 12:55:14 AM »
The Z Power gen take over 3 hours to pay for themselves once they are built.  I wouldn't say they are a godsend, if you have a large influx of resources and nothing better to spend them on, then I guess you could build one.  It's a very long penalty (of 80 metal/crystal per second for 300+ minutes) for a moderate benefit (300k energy for 10 metal/crystal per second).  It's like building a nuke in Supreme command (Forged alliance where you couldn't assist the construction of the missiles)  you /can/ build them, but by the time they are finished, do they really help you?

This is true of a lot of the Zenith Trader stuff, and by design -- it's stuff that is for very long-term benefits.  You could (for instance) use these to eventually break a stalemate or a siege.  But, more often, it's better just to capture a Z Power gen if you can -- getting one simply at a cost of 20 AI Progress from some planet seems rather cheap by comparison, right? :)
Have ideas or bug reports for one of our games?  Mantis for Suggestions and Bug Reports. Thanks for helping to make our games better!

Offline RCIX

  • Core Member Mark II
  • *****
  • Posts: 2,808
  • Avatar credit goes to Spookypatrol on League forum
Re: Schooled By My Own AI...
« Reply #7 on: January 18, 2010, 01:18:03 AM »
The Z Power gen take over 3 hours to pay for themselves once they are built.  I wouldn't say they are a godsend, if you have a large influx of resources and nothing better to spend them on, then I guess you could build one.  It's a very long penalty (of 80 metal/crystal per second for 300+ minutes) for a moderate benefit (300k energy for 10 metal/crystal per second).  It's like building a nuke in Supreme command (Forged alliance where you couldn't assist the construction of the missiles)  you /can/ build them, but by the time they are finished, do they really help you?

This is true of a lot of the Zenith Trader stuff, and by design -- it's stuff that is for very long-term benefits.  You could (for instance) use these to eventually break a stalemate or a siege.  But, more often, it's better just to capture a Z Power gen if you can -- getting one simply at a cost of 20 AI Progress from some planet seems rather cheap by comparison, right? :)

It really was a help for me though because i can basically run a maxed out fleet, which lets me steamroll planets with more ease.
Avid League player and apparently back from the dead!

If we weren't going for your money, you wouldn't have gotten as much value for it!

Oh, wait... *causation loop detonates*

Offline x4000

  • Chris Park, Arcen Games Founder and Lead Designer
  • Arcen Staff
  • Zenith Council Member Mark III
  • *****
  • Posts: 31,284
Re: Schooled By My Own AI...
« Reply #8 on: January 18, 2010, 01:31:34 AM »
Oh sure, if you get one early and are playing for at least a certain stretch of hours after getting it, it's very much a win.
Have ideas or bug reports for one of our games?  Mantis for Suggestions and Bug Reports. Thanks for helping to make our games better!

Offline raptor331

  • Sr. Member
  • ****
  • Posts: 313
Re: Schooled By My Own AI...
« Reply #9 on: January 18, 2010, 08:23:00 AM »
I'm thinking that the ship cap of the Z power generator should be 5, or maybe inf.....Its not like the player is going to be able to build alot of them.