AI analysis: The first 100 turns

[MartinHarper]

So, if he had an escort it would have been a great move to build that settler as he'd win the race against Monte and pick up a damn good city site. As it stands, it looks like he'll lose out.

The obvious answer would be to force the AI to build the settler escort first, if it only has one city. This also allows the AI to send the escort on a couple of squares ahead, and thus benefit from the settler's faster speed.

 

[jdog5000]

Turns 40-49


De Gaulle

Techs: T46 - finishes Animal Hubandry, selects Iron Working
Paris builds: T43 - finishes Barracks, restarts Granary; T44 - finishes Granary (1 pop rush), selects Settler by method 1; T48 - finishes Settler (2 pop rush), selects attack Archer
Other: On Turn 49, De Gaulle becomes the last AI to send out a settler, heading for the horses on the coast.

De Gaulle elects to burn 1 of 5 pop to finish Granary since he recognizes it will help him grow, plus Paris is working unimproved tiles. Four turns later he spends 2 of 4 pop to rush out a Settler, again with the rationale that it will help him grow and Paris is working a lame tile.




Montezuma

Techs selected: T41 - finishes Pottery, selects Masonry; T44 - finishes Masonry, selects Iron Working
Tenochtitlan builds: T43 - finishes Archer, selects Granary; T44 - abandons Granary, selects Settler using method 1
Teotihuacan builds: T46 - finishes Monument (1 pop rush), selects Granary
Other:

Monte sacrifices 1 of 2 pop in Teotihuacan to finish Monument because it will grant access to new tiles, he's working lame tiles.

In the screen shot below, Monte's next city site is the Cow-Wine site on the coast. We also learn from Monte here that the AI can pick out city sites without ever having explored the land ... cheaters!




Zara Yaqob

Techs selected: T46 - finishes Agriculture, selects Pottery
Aksum builds: T43 - finishes city defense Archer, selects attack Archer; T47 - finishes Archer, selects attack Archer
Gondar builds: T48 - selects city defense Archer
Other: On turn 44, after 23 turns of patiently waiting, Zara's settler heads out West towards the horses. Gondar is founded on turn 48.




Since we're now at the halfway point of the 100 turn test, here's a breakdown of where things stand based on in game measures:

De Gaulle

Power: 37

Total score: 222
Pop score: 43 (2)
Tech score: 53
Land score: 110
Wonder: 16

Montezuma

Power: 38

Total score: 362
Pop score: 87 (4)
Tech score: 65
Land score: 194
Wonder: 16

Zara

Power: 37

Total score: 362
Pop score: 109 (5)
Tech score: 53
Land score: 184
Wonder: 16


In a somewhat surprising turn of events, Zara is actually looking alright after 50 turns. The key to his catch up in score is hooking up the bonuses around his capital, as it's population and land which have him propped up. In my mind though he's still well behind Monte as Monte has his second Settler on the way, Barracks, and more infrastructure. We'll see how that plays out over the next 50 turns.

De Gaulle is a weird one. His score is artificially low at this moment because this a deep local minimum in his population. He has Barracks and a Granary in Paris and okay infrastructure, so he should catch up soon. The lack of a coherent strategy is pretty apparent with him though ... none of his individual decisions have been bad, but they haven't added up to anything really either.

 

[Mattygerst]

May I ask...exactly how are you doing this? What would I have to set up to do this on my own?

 

[jdog5000]

Funny you should ask, I was about to write up a description ... will be added to the Debug Mode thread.

 

[TheMeInTeam]

The granary is a good building if one needs to whip or grow...but early low caps can push its priority back if the capitol isn't a good candidate for whips.

Nobody should be building a barracks this early unless they are specifically building units with which to attack another AI. This is not the case. A barracks should *not* be a high priority build for any of these civs. They'd be better served putting those hammers towards more workers, settlers, or escorts. Barracks become more important later.

 

[jdog5000]

Turns 50 - 59

On turn 54 the first Barbarian city appears, south of French territory.

De Gaulle

Techs:
Paris builds: T51 - switches from attack Archer to attack city Archer; T52 - finishes attack city Archer, selects attack Archer; T54 - switches from attack Archer to city defense Archer; T57 - finishes city defense Archer, selects Worker with method 4; T59 - finishes Worker (2 of 5 pop rush), selects Settler with method 1
Orleans builds: T53 - selects Granary
Other: On turn 53, Orleans is founded.

Growth of Paris again leads to some funny production switching, though fortunately there's no waste this time. De Gaulle spends entire interval researching Iron Working


Montezuma

Techs selected: T53 - finishes Iron Working, selects Fishing; T54 - finishes Fishing, selects Sailing; T58 - finishes Sailing, selects Polytheism
Tenochtitlan builds: T50 - finishes Settler (1 pop rush), selects Granary; T55 - finishes Granary, selects Worker with method 3; T58 - finisher Worker (1 pop rush), selects city counter Axeman as a floating defender
Teotihuacan builds:
Tlatelolco builds: T53 - selects Monument
Other: On turn 53, Tlatelolco is founded.

Note: city counter is a defensive unit AI, so the city counter Axeman is designed to counter certain attack city units.


Zara Yaqob

Techs selected: T50 - acquires The Wheel, continues on to Pottery; T56 - finishes Pottery, selects Masonry
Aksum builds: T50 - finishes attack Archer, selects Barracks; T56 - finishes Barracks, selects Worker by method 4; T58 - finishes worker (2 of 5 pop rush), selects Granary
Gondar builds:
Other:

 

[jdog5000]

Turns 60 - 69


De Gaulle

Techs: T60 - finishes Iron Working, selects Sailing; T62 - finishes Fishing, continues with Sailing; T69 - finishes Sailing, selects Priesthood
Paris builds: T67 - finishes Settler, selects city defense Archer
Orleans builds: T66 - finishes Granary (1 of 2 pop rush), selects Barracks
Other:

A pattern emerges here, as soon as the AI settles on the coast it goes for Fishing and Sailing. Monte did it and now De Gaulle.


Montezuma

Techs selected: T62 - finishes Polytheism, selects Monotheism; T66 abandons Monotheism, selects Priesthood; T69 - finishes Priesthood, selects Monotheism;
Tenochtitlan builds: T61 - finishes city counter Axeman, selects city defense Jaguar; T66 - finishes city defense Jaguar, selects Settler by method 1
Teotihuacan builds: T60 - finishes Granary (1 of 3 pop rush), selects Barracks; T61 - finishes Barracks, selects city defense Archer; T63 - finishes city defense Archer, selects city counter Axeman; T69 - finishes city counter Axeman, selects attack Jaguar
Tlatelolco builds: T60 - finishes Monument (1 of 2 pop rush), selects Granary; T67 - finishes Granary (1 of 2 pop rush), select city defense Archer
Other:

In an unexpected move, Monte seems to really be shoring up his defense early. As a result, he holds a 81 - 48 power advantage over neighboring Zara on turn 70. He is also the first AI to begin considering war, having a 1.5% chance per turn of starting total war plans against Zara.

After Zara gets to Monotheism (technically Zara gets it at the end of turn 55), Monte abandons it because it no longer comes with a religion and picks something else instead.


Zara Yaqob

Techs selected: T60 - finishes Masonry, selects Monotheism; T66 - finishes Monotheism, selects Iron Working
Aksum builds: T60 - finishes Granary (1 of 4 pop rush), selects city defense Archer; T61 - finishes city defense Archer, selects Stonehenge
Gondar builds: T65 - finishes city defense Archer, selects Granary
Other: A religion is founded in Gondar on turn 66. Zara now has two, one in Aksum and one in Gondar.

Zara is the first to pick a wonder, doing so on turn 61. Zara also gets a little revenge on Monte for losing out on the primo city site, narrowly beating him to Monotheism and a religion.

Another pattern emerges, all three AIs went to Iron Working long before Writing. I know I've used this fact to set myself up as a tech broker early in the game before.

 

[TheMeInTeam]

IW is a consistent early AI priority. They also avoid aesthetics for a very long time making that tech a priority trade chip on high levels (almost always nets both IW and alphabet, sometimes more).

I didn't know that about fishing/sailing though certainly the AI always seems to have them :/.

A severe fault I've seen in the AI is that once it gets the converter techs (alphabet and currency), it generally builds the wrong thing...often selecting research despite the empire having more SCIENCE multipliers, making gold stronger. Is there a way to get the AI to prioritize its buildings/tiles worked to make better use of the hammer conversions?

 

[phungus420]

I've gone through quite a few of these playtesting. And the main issues I've found with the AI are:

• failure to defend resources from barbs/minors -The AI really should be more apt to attack with it's defenders, and defend improved tiles outside the inner ring of the city.
• Related to the above, moving workers before units (it should move them after), wastes worker turns when a unit could just let the worker keep doing what it's doing and defend it, or kill the threatening unit
• Poor tech priortization, as noted
• Abandoning production, and loosing hammers, this can be severe for World Wonders, I've noticed the AI get up to 90% complete, get a population and switch to war production.
• Failure to stop useless wars because it thinks it's stronger (Start as Minors, it does no good to be at war with 6 civs you're stronger then, pick one, prefferebly the weakest near you and go after them, and stop your other wars the minute you can engage in diplomacy)

 

[karadoc]

The obvious answer would be to force the AI to build the settler escort first, if it only has one city. This also allows the AI to send the escort on a couple of squares ahead, and thus benefit from the settler's faster speed.

That's what I usually do. As you said, sending the escort ahead means that the faster speed of the settler is not wasted (and the settler is usually just as safe, because the 'escort' can keep watch for any nearby threats and join up with the settler if need be) - but also, building the escort first means that the city can grow a bit before building the settler.

If you want to get fancy, you can even have a kind of escort relay - where the settler overtakes the first escort to meet up with the second while both the escorts keep a lookout for danger.

I haven't actually tried better AI.. so maybe I should just shut up - I just know that one of the weakness I've noticed in the normal AI is that it always keeps its settler right on top of the escort, and thus moves slowly.

 

[MartinHarper]

Two of the civs researching IW don't seem to have hooked up Copper or Horses. Going for Iron in the hope of getting access to axes and spears seems like a sensible plan. Monty has Copper, but gets a unique unit from Iron Working. Writing doesn't seem to offer much at this stage.

 

[TheMeInTeam]

Two of the civs researching IW don't seem to have hooked up Copper or Horses. Going for Iron in the hope of getting access to axes and spears seems like a sensible plan. Monty has Copper, but gets a unique unit from Iron Working. Writing doesn't seem to offer much at this stage.

Writing is the single strongest early tech for research rate in the game...

 

[karadoc]

Writing is the single strongest early tech for research rate in the game...

Fair enough, but do you think the AI should _always_ go for writing? In this game we see 3 civs, each with decent reason for picking IW work (a boost in research rate is cold comfort when you're being attacked by axemen and horse archers, and can only defend with archers) - so maybe it would be nice if one or two went for writing instead, then again maybe in some other games you get three civs all going for writing instead.

 

[jdog5000]

Turns 70-79


De Gaulle

Techs: T71 - finishes Mysticism, continues on to Priesthood
Paris builds: T72 - finishes city defense Archer, pushes Great Wall using wonder build 3
Orleans builds: T72 - finishes Barracks, selects Monument; T79 - finishes Monument, selects Lighthouse
Other:

It's a very quiet time in France. The settler that De Gaulle finished on turn 67 still doesn't have an escort by turn 80 even though he built an Archer for it. The Archer decided that defending Paris was very important. It's possible that there were Barbarians in the area which affected that decision, but I think it's just a crippling miscommunication between the city build decision AI and the city defense unit AI.




Montezuma

Techs selected: T70 - finishes Monotheism, selects Monarchy
Tenochtitlan builds: T72 - finishes Settler (1 of 5 pop rush), selects city defense Archer; T73 - finishes city defense Archer, selects Missionary; T78 - finishes Missionary, selects Worker by method 3
Teotihuacan builds: T74 - finishes attack Jaguar, selects Missionary; T79 - finishes Missionary, selects Settler by method 1
Tlatelolco builds: T72 - finishes city defense Archer, selects Lighthouse; T77 - finishes Lighthouse (1 of 3 pop rush), selects Missionary
Texcoco builds: T76 - selects monument
Other: Turn 74, Monte's third settler departs, headed for the coastal spot north east of his capital. Texcoco is founded on turn 76.

On turn 74 Monte is building two missionaries, but only has one city without his state religion and no ability to open borders yet. Sure, he's about to settle another city but his build decision logic doesn't actually know that ... then on turn 77, he has three cities producing missionaries to send to Texcoco. Not a good use of resources.




Zara Yaqob

Techs selected: T76 - finishes Iron Working, selects Sailing; T77 - finishes Fishing, continues on to Sailing
Aksum builds: T73 - finishes Stonehenge, selects Missionary; T76 - finishes Missionary, selects Missionary; T79 - finishes Missionary, selects Settler using method 1
Gondar builds: T70 - abandon Granary, select Worker with method 4; T76 - finishes Worker, selects Granary again
Other:

When Zara selects Sailing, he has no coastal cities and no settlers on the way. Makes me wonder if that's being evaluated correctly or what benefit Zara is expecting. Another missionary issue occurs here as, just like Monte a few turns earlier, Zara pushes a second Missionary which he will not be able to use any time soon.

 

[Mattygerst]

Maybe this is just a brain-lapse by me, or I use different lingo: But when you are saying:

Chooses worker/settler using method 1,2,3,4, what are you saying?

 

[MartinHarper]

Monty should be building up an attack stack now, not wasting hammers on missionaries.

A boost in research rate is cold comfort when you're being attacked by axemen and horse archers, and can only defend with archers.

My thoughts too. One of the AI weaknesses, versus humans, is its failure to defend against an early rush. Researching Iron Working to try to get at Axes is a reasonable plan for Zara, who otherwise risks trying to defend against Jaguars with Archers and Chariots. An alternative would be Horse Riding.
De Gaulle is more out of the way, so I can see the case there for leaving Iron Working and hoping for the best.

 

[jdog5000]

Maybe this is just a brain-lapse by me, or I use different lingo: But when you are saying:

Chooses worker/settler using method 1,2,3,4, what are you saying?

The function CvCityAI::AI_chooseProduction is really long and contains a lot of clauses for building many with different priorities. There are six worker build clauses, several wonder builds, etc. So, these comments (and the similar lines in the log file) indicate which of those clauses was used for some important build decisions. Mainly they're a note to myself so I can more easily make adjustments to this long function later.

 

[TheMeInTeam]

Monty should be building up an attack stack now, not wasting hammers on missionaries.

Missionaries are not wasted hammers. They outsource production from better hammer cities to ones without many to get a border pop w/o a monument. If under OR, they are also an excellent production boost to set up infrastructure (hammers pay back easily).

Now while not a waste, an offensive offers potentially higher returns, but we're still within the constraints of a role-play AI and monty chose a religious opening. Those missionaries will help with his culture and eventually buildings or military units so they're not all that bad...they are way better than a FAILED rush for example.

 

[scu98rkr]

Now while not a waste, an offensive offers potentially higher returns, but we're still within the constraints of a role-play AI and monty chose a religious opening. Those missionaries will help with his culture and eventually buildings or military units so they're not all that bad...they are way better than a FAILED rush for example.

This is a very good point. There was a joke on the forums sometime back about what the AI would say. One of the answers was

AI : "The human player is always restarting its not fair"

We have to remember that sometimes the human player will restart if the starting area is not good or if a gambit does not pay off. The AI does not have this luxury.

Most times it may be better for the AI to go for the safer option rather than the high risk / high reward option.

Ie an AI should only be going for a early rush if its sure its going to get some pay back long term other wise its a waste of hammers.

Unless of course you head back to the situation where by the AI's gang up on the human and one launches a suicide rush just to slow the human player down.

 

[jdog5000]

Turns 80-89


De Gaulle

Techs: T81 - finishes Priesthood, selects Monotheism
Paris builds: T83 - finishes Great Wall, selects city defense Archer; T86 - finishes city defense Archer, selects Oracle
Orleans builds: T86 - finishes Lighthouse (1 of 2 pop rush), selects explore Work boar
Other: On turn 87, De Gaulle sends his second settler east to the space between his empire and Montezuma's. On turn 89 the first trade of the game happens as De Gaulle signs an open borders treaty with Montezuma. Monte has missionaries on standby - we could argue about whether that was a good strategy on Monte's part if it had been a strategy. Instead, it's a fortunate accident.


Montezuma

Techs selected: T80 - acquires Monarchy, selects Construction; T84 - finishes Writing, continues on to Construction
Tenochtitlan builds: T80 - finishes Worker (2 of 6 pop rush), selects city defense Jaguar; T81 - finishes city defense Jaguar, selects reserve Chariot with floating def 2; T83 - finishes reserve Chariot, selects attack Jaguar; T87 - finishes attack Jaguar, selects Library with economy flags 3
Teotihuacan builds: T84 - finishes Settler (2 of 5 pop rush), selects city defense Archer; T85 - finishes city defense Archer, selects Worker by method 6
Tlatelolco builds: T80 - abandons Missionary, selects Settler with method 1
Texcoco builds: T80 - abandons Monument, selects Granary
Other: On turn 80 Montezuma adopts Hereditary Rule, apparently triggering a mass wave of reconsidering city production decisions. On turn 86, Monte's fourth settler departs west (although it'll be beaten to the spot by De Gaulle).

Texcoco abandons its Monument when adopting a new civic causes production re-evaluation because it now has culture coming from the Aztec state religion. Perhaps AI cities in empires which have a strong religion presence should figure they'll get religion soon and avoid monuments early?


Zara Yaqob

Techs selected: T82 - finishes Sailing, selects Priesthood
Aksum builds: T81 - finishes Settler (3 of 7 pop rush), selects city defense Sword; T82 - abandons Sword, pushes reserve Chariot with floating def 1; T83 - finishes reserve Chariot, selects city defense Sword; T84 - finishes city defense Sword, selects state religion Missionary; T87 - finishes Missionary, selects Settler by method 1
Gondar builds: T84 - finishes Granary (1 of 5 pop rush), selects Barracks; T85 - finishes Barracks, selects state religion Missionary
Lalibela builds: T84 - selects city defense Sword; T89 - abandons city defense Sword, selects Lighthouse
Other: On turn 83, Zara sends a settler to the coastal spot to the east. Lalibela is founded on turn 84.

When Zara rushes 3 of 7 pop in Aksum, he does so because Aksum has an unhappy citizen and the rush clears that up at least temporarily. Again we see the AI not counting its existing Missionaries, as Zara has one waiting but starts building two more before the waiting one can get to Lalibela.