A Better AI.

[Tattila the Hun]

Say, was there a way to verify, that Blake's AI was infact in use? Some text in some place in-game?

[Thedrin]

Hold down alt while mousing over your name in the score list.

[Blake]

Originally posted by lockstep
Build Dec 9th - my starting position is 3 (!) tiles away from an AI start. Maybe not a bug, but I haven't seen such a cramped start before.

Thanks for the save!
I can see how this happens now, I'll fix the logic up and for the next build it REALLY should be gone.

[Willem]

Originally posted by Blake

Thanks for the save!
I can see how this happens now, I'll fix the logic up and for the next build it REALLY should be gone.

Woo hoo, that's good to hear! It's so damn annoying when that happens.

[Blake]

Basically reposting some stuff I said on CFC:

Okay good/bad news for Noble players. Next version is going to be significantly tweaked to improve prince and lower performance. For long enough it's been tested at Monarch nearly exclusively.

Also the AI will build wonders considerably less obsessively... which is not to say the super early wonder will never happen it'll just be rarer.


Something which is good news for less-strong players (and it's been true for a couple of builds) is that the default governor is considerably stronger, that means if you leave the city to manage itself you'll be performing better along with the AI. Hopefully this will slightly blunten the impact of the higher AI skill.

[MikeH]

I got my computer back at the weekend and played my first game and a half with the new AI.

Computer rebuild meant I actually played both these games on normal speed because I forgot to change it. Previously I was on Monarch Marathon speed and mostly losing but winning if I rigged the settings to small map, all on one continent and started with a good civ.

1st Game, stepped down to noble because of the great fear instilled in me by this thread. Really lucked out on the random leader and got Hatsheput, found horses early and my chariots tore through the continent eventually making the last couple of civs my vassals and getting a conquest win in about 1600 or so with my highest ever score (but I haven't played . Hatsheput is a leader I'd underestimated (I always play random and hadn't got her before). Spiritual Creative is a good combo, that'd only get better at higher skill levels. Her UU is very handy UB also useful for early Great Prophets.

Anyway, I didn't learn much about the new AI from that game, especially as it seems to be better later in the game. Other than actually Noble can still be a lot of fun. It was noticable that the AI had better workers, they were braver and the cities I captured were better sited and better improved, which is much better for warmongers!

Second game, I moved up to Prince. Again I totally lucked out on the leader and got Augustus. Augustus was the guy who took me to my first Monarch win. Brilliant traits for a warmonger, the games most effective UU and a very good UB too. I'm currently about half way through the game, this time there are 2 main continents (I'm playing Fractal) have conquered my home continent and I've just launched a trans-oceanic invasion on Brennus. My target changed from Shaka (who's about equal with Brennus, but has a vassal) to Brennus because it appears Brennus is going for a cultural victory. His culture scores are through the roof. I just captured 2 of his cities as a home base and I've been harassing his retaliatory stacks with gunships. So, he's poised with a massive stack of Trebuchets (the gunships killed everything else) heading towards the cities I took I'm shipping my artillery round the coast to take more cities and I'm just a few percent short of a domination win but then it was time to make dinner. This game is as good as over. I do seem to have hit that level where I have so many cities that I am just teching so fast the AI, even with superior AI, can't get close. This is why I stopped playing noble, in the time it takes my units to ship across the sea they are obsolete.

I think what I've learnt is that although the AI is now much better at noble and prince, I've improved massively by playing on Monarch. The difference in how they set their economies up, and how they use their armies is pretty significant. Also, I think marathon speed isn't quite the advantage for warmongers I'd thought it was. Ok, you can build units quicker, but so can the AI.

[MikeH]

Sorry, I got carried away with my stories of my game and killed the thread.

[Solver]

Also the AI will build wonders considerably less obsessively... which is not to say the super early wonder will never happen it'll just be rarer.

I presume that is to allow for more early military / expansion, or why else?

[Xmudder]

Originally posted by Solver

Also the AI will build wonders considerably less obsessively... which is not to say the super early wonder will never happen it'll just be rarer.

I presume that is to allow for more early military / expansion, or why else?

Actually, the AI was building wonders before building workers. Not a problem on Monarch+, but definitely a problem on Prince-

I suppose a good question would be - should the AI build a settler before it starts working on wonders, or should it build Stonehenge before if founds a second city? I prefer the latter myself - to have a second city up and running before I start building wonders.

[DrSpike]

Originally posted by Xmudder


Actually, the AI was building wonders before building workers. Not a problem on Monarch+, but definitely a problem on Prince-

On the contrary, that would be more of a problem Monarch+.

[lange17]

What's the earliest you guys have seen the new AI building Stonehenge? I finally bought Warlords, and in my first game I was trying to build it as the Celts with stone in my fat cross. Before I even got the stone hooked up, Ramses had built Stonehenge with no stone in 2635 BC! Was that unusually fast or do I need to adjust my expectations of what the enemy is capable of?

[Risa]

And there are couple of (monarch) games in which it was built in 3120 BC.
I think as long as AI starts with bonus worker, such kind of action won't hurt them much.

BTW, youindeed had better adjust the expectation. I've already lost a game to an ultra-early dagger, a six- archer stack invading at around 2400 BC

[Saurus]

Originally posted by Risa
And there are couple of (monarch) games in which it was built in 3120 BC.
I think as long as AI starts with bonus worker, such kind of action won't hurt them much.

In fact, should the A.I use its starting bonuses effectively, then it should use its several starting archers to kill the single human warrior/scout whitin 5 turns and such wipe out the entire human controlled civilization!

Though, what would be the point in playing such a game? I think It is a bit of the same with very, very early A.I wonders.

[Iustus]

Download the new build 12/12 on sourceforge.

release notes:

Starting Point Changes:
- Rewrite of minimum distance stuff to make it more flexible.
- Players should not start very close together.
- Players should not be jammed into unusual spots.
- Changed the food bonus normalizer to be fairer on very dry starts.

AI Production Changes:
- The AI will train workers much more readily on Prince and lower difficulties.
- The AI will attempt to grow to a sensible size before stalling growth for workers/settlers.
- The AI will be less likely to build useless lighthouses.
- AI wonder building madness toned down - the AI will still build wonders early but not so obsessively.
- The AI will be more sensible when it comes to whipping (or not whipping) wonders.
- the AI will build less cultural items in the top city, saving it for the
weaker cities when pursing cultural victory

AI Research Selection Changes:
- Worker techs valued much more sensibly, the AI takes food into account.
- The AI recognizes bonuses which will soon be included within their borders.
- Fishing value is more context-sensitive, much less valuable inland.
- Founding a religion is valued more if it's the next tech rather than 2 techs away.
- The value boost from wonders/buildings is somewhat capped to reduce "Masonry Obsession"
- Techs which give a civic may be valued more (particulary Bronze Working and Monarchy)
- The AI will pursue strategic resources (ie weighting Iron Working more if it fails to get Copper or Horses).

War Changes:
- AIs will prefer going to war with neighbors, they will be very unlikely
to go to war with a target that is very far away

Strategic Changes:
- AIs with favorite civic preventing religion spread will be less likely to
pursue cultural victory

Misc Changes:
- Autoworker now improves food resources with higher priority.
- Autoworker should do a better job of roading things up.
- AI are more flexible and will attack with smaller stacks in the earlier eras.
- Fixed a bug with Coal trade value.

-Iustus

[Solver]

- The AI will pursue strategic resources (ie weighting Iron Working more if it fails to get Copper or Horses).

Wow, now how hard was that to implement?

- AIs will prefer going to war with neighbors, they will be very unlikely
to go to war with a target that is very far away

Hmm... I hope it doesn't mean that wars will only happen between civs who share borders, as that would make it too easy to predict which AIs might invade. Also, it would be great to, eventually, see the AI be better capable of cross-ocean invasions, which can be pretty far away .