AI League 2.0

[Thrasybulos]

Looks like it indeed, thanks.

But damn, I was positive turn order mattered.
I guess those ties simply don't happen that often, and that the very few times they did happen and I paid attention, the RNG produced the same result turn order would have.

After all these years, keep learning stuff about this game.

 

[Genghis the Barbarian]

I’ve had a couple ties when running sets and I think at a certain point it’s just purely random, especially for same victory conditions

BTW I love the awards @Thrasybulos I wonder if I inspired them?

 

[Thrasybulos]

The one case which happens often enough for me to notice and remember is the same AI achieving two different victory conditions on the same turn (Domination / Space or Domination / Culture).
Domination has always taken precedence in those cases.

I also remember a few same turn Space races, but those were decided by one of the ships missing case(s) and failing.
If I witnessed other instances, I can't remember them, so they must have gone (by chance apparently) according to turn order (and thus didn't strike me as odd).

As for the awards, it wasn't a direct inspiration, but since you planted that context association (AI games <=> awards), I'd say the idea to present those stats in that form certainly owes in great part to you.

 

[Thrasybulos]

I also remember a few same turn Space races, but those were decided by one of the ships missing case(s) and failing.

See, that didn't take long to happen again:

Now, when you think about it, it's actually not surprising.
The AI may not be programmed to optimize its research path towards the Space race, but it is aware of where it stands in the race. Which is why for instance it will simply not launch its ship if it's too late to win the race.
So it actually makes sense that when two AI ships arrive at the same time, the odds are that one of them is incomplete: it is more likely to have one AI in the process of completing its ship when another AI launches, than to have two AIs complete their ships on the very same turn.

 

[Thrasybulos]

Series 5 Results

Spoiler Results :

Oh, look who's back on top.

Now, watch him utterly crush the next League Series?

Called it !

Now, HC didn't actually "crush" that series. In fact, he had a pretty bad first half: when he finally joined the top pool in round 5, he wasn't able to win a single game there and dropped out immediately.
Then he won 4 games the next round, in a lower pool, which got him back to the first pool, and this time, he won 3 games, which propelled him to the top of the leaderboard where he stayed to the end.
But his results do not invalidate what I wrote for the previous series: out of his 21 wins, 3 were by Spaceship, only 2 by culture, and 16 were Domination wins!
While I'm not watching the games (just the replays, and getting a glimpse of the situation when my first batch of aiplay turns completes without the game being over), I can tell for instance that he did lose his very last game in the series because he turned the slider on: he was far ahead, but lost to a rather slow Hannibal's spaceship only because he stopped teching.
(Of course, as I write this, HC's first win the next series comes thanks to the slider: he would have lost otherwise. )

Justinian had the complete opposite progression through the Series: he was far ahead of the rest of the field after 5 rounds, with 13 wins. And then, he just stopped winning: only 4 games in the last 5 rounds.

Once again, both Egyptian leaders did well (Hatty overall actually did better than Ramesses, but she faltered in the very last round while Ramesses had an extremely good last round), but we were back to the pattern where the few high peaceweights who made it to the top pools weren't made to feel welcome there.

Sury and Cathy had an appalling performance (somewhat alleviated on the leaderboard by a decent last round result, but you can see that they stayed throughout in the lower half of the pools, with even a prolonged stay in the very bottom pool).

Qin would have actually fared slightly better than Mao this time, had the RNG not decided otherwise (see previous posts). But it's increasingly looking that my original assessment (Qin = Good, Mao = Meh) is indeed wrong, with both of them being just average leaders.

Peter repeated his miracle round from the previous series (4 wins), and did slightly better this time: 4 wins in the remaining 9 rounds instead of just 2.
Amazingly, Freddy also had such a miracle round.

Spoiler Awards :

Spoiler Map Data :

Spoiler Pool 1 (S8 Game 7) :

Spoiler Pool 2 (S7 Playoff 3) :

Spoiler Pool 3 (S3 Playoff 1) :

Spoiler Pool 4 (S1 Game 5) :

Spoiler Pool 5 (S3 Game 4) :

Spoiler Pool 6 (S8 Game 8) :

Spoiler Pool 7 (S7 Game 8) :

Spoiler Pool 8 (S7 Game 4) :

 

[Thrasybulos]

Regarding ties...

Spoiler :

And the winner is...

Joao!

I know it's confirmation bias at play, but I swear most of Joao's wins happen with him being a runt civ and getting an opportunistic and lucky spaceship victory.

Joao's ship landed on the very same turn that Saladin hit the Domination threshold.
Joao had the turn order advantage.

So either this is yet another case where the RNG happened to yield a result aligning with turn order, or turn order prevails when the tie involves a non-spaceship victory?

 

[Thrasybulos]

Here's a stat which makes sense, while being remarkable nevertheless.

With the caveat that kill credit with Sullla's method is a very unreliable data point, the leader who gets credited with the first kill ends up the game winner in 39% of the games.

And just to illustrate that, here is the endgame situation for the latest game I've run:

Spoiler :

Qin, in light yellow, is just shy of the Domination threshold (62%+ for 64% required), but Saladin wins by Space.
The game started with Saladin conquering his neighbour, Pericles, early.
Then it was Qin's turn to spring into action: he conquered Hammurabi, then Napoleon, and then finally Justinian.
But Saladin's early conquest allowed him to develop that extra land earlier, and that gave him a tech edge which carried him to the end.
There's no doubt that if the game went on, the map would turn yellow. But... too late.

 

[Thrasybulos]

This is unrelated to the League, but not worth creating a topic for, and is about the AI anyway.

You know that thing about the AI settling on ressources which has Sullla regularly laugh bemusedly?
It would seem that something in the AI's code makes it deliberately go for those spots: ie, it's not an unfortunate random pick among several possibilities.
I'm currently working in a map for another side project, and as I was testing the map, I noticed these:

Spoiler :

The AI will always settle there.
But if I move the Iron to the hill 1S1W in the first case, then the AI will always settle on that hill instead.
And in the second case, if I move the Iron to the hill above the Crab, and move the Crab 2W, then the AI will pick the new Iron hill as its settling spot.