Discussion in 'Civ4 Strategy Articles' started by Manco Capac, Jul 23, 2012.
Yea I mad that it's not possible to eat the damn cake.
Tachywaxon speaks "Civilization IV". For example:
I hear Tachy's message(s) loud and clear in the HoF Challenges and other CivFanatics' places too.
BTW, congratulations on a perfect HoF Challenge score, first place in all ten challenge games! I hope it remains perfect in the remaining < 50 days of the HoF Challenge VIII series.
Sun Tzu Wu
I am drawn here and thought that you were talking about the slavery civic.
Glad to see something much better than I expected, hope you completed it.
Good job Tachy! I am going to have to find the easiest warmonger one that I can lazily do just to troll him. But it was a joke about how Quabecuaians speak butchered French, and English.
I'd love a youtube illustration
Here we develop in details the tricks used to worker in situation of a maintained war slave economy. Not ideal of course, but occasionally necessary.
Woodsman II or Guerilla II Units
Probably the oldest known exploit to snatch workers.
First, let's start about the worker checks if danger is near or not.
Under war state, 1 units scare workers if standing by then and while IBT, they flee. Whereas 2 units scare them from 2 tiles away.
Some promotions allow certain 1 units to behave faster on certain features like forests and hills.
Woodsman II promotion makes the 1 unit moves twice as fast if at least the first half of its only move is used on a forest.
Guerrilla II promotion makes the 1 unit moves twice as fast if at least the first half of its move is used on a hill, forested or not.
In the early game, woodsman II is associated with warriors, spearmen, axemen and swordmen while guerrilla II is tied with archery units.
Mounted units cannot benefit from such promotions.
Special case is the IMPI from Zulu Empire: they won't double their movement ( 2 ) with woodsman II.
Thus, by making them twice as fast while the stats still assume as 1 units, the workers won't flee. Normally, if Firaxis had seen the exploit, they would have tied woodsman II to 2 units, but given this issue was never fixed, the trick is available.
And think it is not a complete exploit given the XP you need to reap via risky barbarian units confrontation and the risk of an AI retaliation once the worker is snatched. Indeed, IIRC, BTS patched the remaining moves of a stolen worker and reduced to zero, so we have to wait one turn before running away with the worker.
Workers have 2 , but the lack of mobility promotions means not all tiles will eat their moves equally. Most knows hills (forested or not) and forests will eat two moves. Comparatively to CivIII where the workers were slow, their 2 permits them to flee quite easily. But a good pattern of the lands where they work on combined with AI idiocy lead them to attempt a fail escape. They are attempting to hide behind the walls, but they are obstructed.
Now, how the worker will react (will it go in the good direction?) is a huge matter of experience and by the time, one will predict the AI automatic reaction to danger. Don't be surprise the workers once in a while do a long stroll in the savage lands amongst barbarian just to avoid you. And surprisingly their survival rates are high thansk to the "omniscience radar" (a check completely unrelated to the concept of sight that influences the AI units behaviour) that all AI units possess.
Successive Worker Steal / Sheep's Clothing Trick
From this point, we enter into terra incognita. The following tricks are usually more obscure and sometimes are simply exploit.
After many worker examples throughout years and observations of AI behaviours, I started to notice few peculiarities.
One of these peculiarities is stealing workers one by one under certain circumstances. I haven't found yet the exact origin in the code what makes the worker behaves that way, but this present trick is a complete non RNG behaviour. Deterministic pattern.
Let's take a situation. Two workers are improving and are adjacent one to the other. You capture one and wait one turn to flee along the worker. Nonetheless, you notice the second worker hasn't obeyed to the caution rule to flee from imminent dangers. The second worker is still busy with the improvement (or chopping). Then, why not rapture him too. Afterwards, you notice another worker in the fog just now with the last capture. Busy with an improvement. Next turn, you flee with 3 workers.
Simple as that! If multiple workers are adjacent to each other AND busy with an improvement, then capturing one of them is similar to be in sheep's clothing. In fact, generally speaking, the sudden loss of a friendly unit next to workers completely absorbed with their chores will not notice the change and will not flee. So all units means a military unit killed too. Still, if the worker hasn't started an improvement, it will flee 100% of the times.
Marathon speed is the best for such trick given the inflated investment into an improvement. Quick speed is the worst.
Helpless Workers Paving the Cities
Haven't you been frustrated your worker killed himself for various reasons under automate worker mode? Well, that happens to the AI...seldom.
Our automate workers sometimes calculate and pave routes regardless of the changes and dangers and that is the problem. How many times our worker started the next stretch of road next to a barb or enemy AI unit without thinking it was a suicidal move. That happens to the AI too. And that happens when they create routes between resources or mainly cities; they calculate regardless of the moving dangers.
Now, how to let the worker offers itself as slave; unfortunately, that is no easy task. Experience will give indicators and possible routes they may pave.
Just keep in mind some basic rules that help to bode their sacrifice to slavery:
When creating a path (between two cities for example), the worker will favor non-move-consuming tiles as bare plains or grasslands. Unless those triggers very intricate routes. For instance, after paving a tile, the worker moves forward and has three possible tiles; two hills and one bare tiles; the worker will minimize in this case move losses and pave that bare tile.
Once the second city is settled, that exploit often starts to appear once in a while.
If choking an AI and you warrior (typical example) is stuck to one place that was previously as path for paving route, don't expect the new and fresh workers from cities to pave exactly like the past and now enslaved workers. The enslaved workers calculated a path between two cities before your arrival, while the new ones will calculate by taking account of the new non-moving danger that is you. Yes, shamelfullly, the AI worker has better code (not seen in the code yet, but experimented).
Don't forget this happens anytime in the game when roads and railroad are involved and consider this event as an uncommon one. Still, expertise makes it more common.
Deity has the particularity to present the earliest cases of this exploit due the second free city. Quite often, the deity initiial worker will work a tile (food first if improvable) or, then will start connecting the two cities. Watch out for this!
No, we are not referring to terraforming from Sid Meier's Alpha Centauri (SMAC) game feature. It is a personal dubbing of mine.
And this one a pure tactic not even revolving around any exploit at all. So much that even multiplayer games may allow it if you are fast enough...and cunning.
Resembling to woodsman II/Guerrilla II tricks, it is to modify temporarily the movement capacity of a unit by speeding it up or cancelling the obstructions.
In most cases, it is based on road feature.
For 1 units:
With the help of worker along your enslaver, the workers prepare the way to the victim. Let's imagine a target two tiles away from the warrior. By now, you know if you stumble by the (to be continued).
Target's Border Next to a Neighbour Border
Excellent addition Tachywaxon! You are extremely focused on your enslaving goal! I don't believe I've ever tried capturing a Worker by building a road to it for a 1 move military unit (usually an ordinary Warrior).
Please Avoid Using the Term Exploit:
I wouldn't label all flaws in AI behavior as an exploit. There are too many players that unreasonbly expect that all exploits be banned, especially when its technically impossible to enforce the ban on such exploits. Banning exploits that aren't enforceable doesn't hurt the cheaters; it does unduly penalize the honest players that voluntarily honor the ban. The term "exploit" should have the meaning of a game breaking bug that gives the win to the player. We all know that the AI is stupid; there's no need to call certain unoptimal moves by the AI as exploits when they don't literally give the player a win on a silver platter (for example The Oracle exploit that allowed one to take an arbitrary number of Technologies = not just one).
Sun Tzu Wu
Worker stealing on deity would be rather the opposite of an exploit, gambling
Maybe with an exception of early "super units" being guaranteed, like Skirms.
You need to steal fast, or deity AIs with their rapid progress become more and more dangerous, and you cannot know many factors yet..like your own land and if you shoot yourself in the foot by not going peaceful, res. they will have and if you will be able to monitor all possible spots.
Have also seen many situations where they might attack my warri after stealing, or not.
And something easily forgotten: while messing with 1 AI, somebody else might get way more land than they should.
Oh nice! I never thought of dropping a road to catch a worker before, but it ought to work great.
Hey Your Tachyness -
Not that haven't stolen workers before, but I've been following your practices at bit more closely of recent, as well as practicing. Sometimes it works really well, but others I have bad luck in the following areas:
1) I just don't find the workers. I always go the wrong way...ha. Is there a sense you have about where to attack? Is there a method to how you approach that first AI or two.
2) What's the deal with AIs guarding their workers. I can understand when I start attacking some workers and scouts, but I talkin' very early in the game before I've pounced on anyone. More often than not this does not actually happen, but it does happen. It's like the AI senses I'm ready to pounce. This one game, Fred was my first target. Not only did he move an Archer to guard the worker to the resource (at the same time), but actually moved another archer in an adjacent tile next turn. Is there something I'm missing?
It is an honor you start challenging along me. Such high score is also due to the lack of opponents. Game 2 and game 4 are the best for warmongering. Still, an advice, be prepared for game 4, the mapscript is a terrible choice and makes the whole challenge game a pure gamble. Ice Age mapscript is a sort of minor fractal type; some are gonna get a cute pangaea, others multiple continents and the raging maps are those with a single AI (the last one) in a continent not accessible without Astro. My first attempt was a map with missing AI's and I knew it was not competitive compared, let's say, someone who spawned a cute pangaea. After several map looking, I started to guess the pangaea types: those maps where you meet 4 AI's scout early (pre-2500 BC). Of course, that doesn't change the odd to getting a single isolated AI, but it is still a good indicator.
Don't worry, I know you typically like to troll...still...I'm surprised you made apologies. I feel appreciated. And Quebecois (sound just like your "Quebecua" thing) people really have butchered Frenchglish. And a very recognizable accent. Weird enough, I have no accent just like the rest of family. Even more, I can imitate french accent from France.
No critics. Come on, you can do it.
Thanks for the notice. Although the place for making the observation is weird.
It won't be youtube...especially since Google took over the corp.
Same as my Charlie thing on Wat.tv.
Indeed, I will examplify with many instances. And random ones of course.
Indeed, I want to finish a project for once. But that is going to take over 30 hours I think. Glad people are observing new tactics. I was no joking when I stated I'm expert in worker stealing. Also, I observed via HoF games and some S&T games how rarely that tactic/strategy was rarely used. So, I guessed the worker stealing aspect was never studied entirely. It is certainly an old tactic, but I'm not sure as a complete strategy that leads to victory. My challenger-VIII-03 is one perfect example where slave economy was the sole basis for a 1070 AD diplo victory.
Perhaps, I'll change the term from exploit to "blatant flaw". Exploit is also a suggestive term that varies depending the morality of the player. Dhoomstriker is a player who refuses to steal settler while I have no qualm to do it. Duckweed has no qualm to put some trap cities while Rusten feels it is completely wrong and refuses to do it.
Interesting. You are more mild than usual. I was expecting some critics. Enthusiastically. Serious. Brennus gets all the critics and I all the "Nice!".
Still, you pointed something important: deity.
Indeed, I have tried enough of sample to conclude deity is bad and almost worse than quick speed to raise a slave economy. One simply cannot base that economy on warriors but need back up like axes, chariots or beefed up archers.
In Settler to Deity, be sure I'll talk about the particular case in details.
Since the article isn't finished, I haven't included the other tools needed more than ever on deity. On deity, you need all the tools to reduce to a minimum the gamble.
Indeed, given how vast the subject is, I forgot to mention a crippled AI means a stronger AI next door. Often more of a subject on IMM+. It is TMIT that made realized that long ago in a ~2009 A.D. comment.
Aren't you the one who didn't how to queue things while making great feats on HoF.
As mentionned in the intro, slave economy is not a must all game. One must take a different path if slave economy is not practical. Along the criterion, your initial warrior needs to find a neighbour and a close one. No, I can't guess the right direction and my only tool now is the culture button on globe view; a good tool to spot an AI and to avoid missing it by taking the wrong path at the wrong moment.
Archer guarding workers is a case I count with my hand's fingers, but that is probably due to the lack of a big sample. I think I have 100 complete games in my whole experience. And perhaps fewer. I have seen some cases of archers protecting their workers and this is an example of a condition you pass slave economy option. Thanks Sun Tzu Wu and Lymond to point this out.
I have not the exact idea why the AI does that, but IIRC, Firaxis attempted with BTS to make the AI more protective of their initial workers. I think it was a known issue and they attempted to fix it. They attempted, that's the word. And I think that happens when the AI goes for an extra early archer (like before the first worker). The RNG is low but that happens, especially when the AI starts on a PH (I've observed the AI going for an archer first more often on a PH).
The AI cannot predict your coming, but Firaxis attempted to thwart the early rushers.
Here is another example I think. ( In fact, I don't remember if it was Firaxis' work or the community).
On the question of very early Archer escorts for AI Workers: I agree 100% with Tachywaxon that this happens when the AI via RNG decides to build an Archer first in its capital. It often uses such a first Archer to escort a Worker. In my experience, once an AI starts to escort its Worker, it will escort all its Workers, assuming all its cities have at least 2 Archer MPs plus a free military unit to escort each Worker.
Whenever I see an escorted Worker, my enslaver units find another AI and enslave its workers instead.
Regarding my congratulations for your current #1 showing in the HoF Challenge VIII here, rather than the HoF Challenge VIII thread:
You deserve a wider audience than just those subscribed to the HoF Challenge VIII thread!
Sun Tzu Wu
I made a little study on that subject: Worker escorts.
First, no one will gainsay the fact as the game progresses, you are going to observe more and more escorts to the workers and even multiple escorts. Unit spammers on the first rank. Often mounted units but that's just a random effect of RNG I think.
People are used to early vulnerable workers because the initial archers (EMP+) coming along the initial units are set to UNITAI_CITY_DEFENSE script that forces thoses archers to almost never attack and to rush 100% of the times to the closest city to garrison. They may attack, but rarely do and only against non dangerous units like a solitary warrior (I clearly remember I read this behaviour was implemented to destroy a bad behaviour from the AI in CivIII where the AI rushes somebody nearby, they send all the military units, leaving nothing or leaving units in inertia; thus making pillaging overly easy.) or a worker (or even barbs because of the benefitting modifiers AI gets). A big army around and a new script is acting, forcing all archers, whatever the type, to garrison. Anyways, that's a different story.
What you pointed out is a rare case caused by RNG. I may say it happens <20% of the times.
The first build of an AI depends solely of RNG. Sometimes, RNG pushes the AI (independently of XML I think) to produce another archer set to UNITAI_ATTACK.
Archers are special units that possesses many possible scripts. They are supposed to be versatile I suppose. UNITAI_ATTACK archers are dangerous ones because they roam around their first cities and are more prone to attack in number. And this is the typical archer script for the second city...at least one of them.
What I observed as a new component I didn't know is escorting doesn't happen all the time.
Escorting is a combination of two RNG elements that makes the real cases even rarer.
UNITAI_ATTACK will escort a worker only under condition the said archer is one tile adjacent to the busy worker. The the archer will fortify along the worker and follow it unless the worker goes away way to fast for the archer (help by roads and mainly the 2 ) and the little squad is automatically destroyed.
If the UNITAI_ATTACK archer has 1 tile separation from the worker, then it will continue to roam as if nothing happened.
In conclusion, escorting early is a combination of two rare RNG components:
Archer first and archer+worker crosspathing.
In a competition game (mainly SGOTM and XOTM), this should not happen often as only the settling date (human player) changes the build order and choice. So, people going to non-SIP may feel different behaviour.
The onset of AI Worker escorting seems to be greatly affected by game speed. Faster speeds will complete the initial Archer much earlier in terms of turn #.
For example, in my current game, my initial Warrior located the first AI and a ring 2 food plot around turn 4-6. My Warrior waits a few turns for the Worker to show up. It shows up a turn later escorted by an Archer. My Warrior starts looking for another Worker supplier (more "trusting" AI).
In Deity level, it seems that the AI often builds an Archer first that often fortifies in a city increasing the defender total to 3 and later in the game to 5 and in case of war even higher. The other option is either wandering or Worker escorting which Tachywaxon already covered well. When there are either 3 or more Archers in a city or a lone Archer in range of counter attack, the chance of that counter attack seems to be close to 100%.
Sun Tzu Wu
Last point I forgot to add in the last post is how PH affects; yes I just confirmed my experience; more AI's are prone to start an archer first. Interestingly, I put new seed after reloading and I even settle my own initial city at different times, and the AI's that chose archer first always chose archer first. In fact, the first build were the same whatever the RNG. As if it is tied to some parameters. Interesting and good for competition games.
EDIT: Ok, I think I have a leak of info from the game. I think I spotted the deterministic reason behind the plain hill archer first. Still RNG I guess but tend to be deterministic. A human player on a PH will see his initial warrior syncs with the second city size, that is 8 turns. For the AI, archer first is also another kind of sync with size 2. The archer takes 5 turns and the next pop takes 6 turns with a 3 tiles and that is the main reason for archer first. PH+3 tile. It doesn't work like that all the time, but that is an interesting assertion. And the AI won't go for worker first once the archer is finished, it will wait 1 more turn (on another archer for instance) just to get more production from the second population. The AI has minimum intelligence I see.
Gamespeed aspect was covered already. The quicker the speed is the less good slave economy is and the more experience you need.
I did a study about counter attack from stack of archers in city. But looks like I forgot to input the critical approximative number of archers for retaliation. Anyways, I clearly remember it was RNG-based too. A very important factor is the AI unit courage. Frederick has a low one and IIRC, around 6 archers, he starts to retaliate. But that could be RNG.
You can't do a thing against escort and waiting for the little squad to disrupt is not efficient. By the time it disrupts, other UNITAI_ATTACK archers are built, bringing up the chance of meeting and future squads and by the time, you have plenty of time to poke another AI.
Being a settler-level player ... the term "worker steal" is a something of an oxymoron. I can attest to your observation of what I would call the "clockwork orange" movements of a second warrior by the AI. But rather than steal only the worker, I grab both the city and worker(s).
That is normal; on settler to even prince levels, taking both settler and worker is best. All you need is a view to the city tile (without DOW preferably) and look regularly the espionage tables. Then you might predict the worker coming out and once next to the city tile (diagonal attack), the worker has an imperative to stay in the city. Experience can make up the espionage table, but the espionage table is so cool.
Gandhi (or Asoka) can do some funky things at settler marathon--huge map. By funky I mean like capturing an AI capitol 15 tiles away by turn 10 with a sequence like: pop worker from hut, said worker speeds through forests snagging writing, sign OB with an AI, pass through AI land and pop warrior from hut which captures said AI capitol. Queue worker in captured city. This is much faster than worker steal or conquer city and worker.
Indeed, huts+marathon+FW+No barbs is an enormous beast. Given the lack of scaling, by T10 (nothing on marathon), human units can start to pop warriors (impossible before). On quick speed, by T10, the warrior is almost insufficient and the RNG will not necessarily gives us that warrior. Not taking account the travel time.
But that article is set along the not huts and no events thing plus more geared towards Monarch and more. Indeed, on Prince and lower, getting both is far superior than just take the worker and make peace.
Separate names with a comma.