If the intent was to measure the best AI, lots of changes would be needed. Since the intent is good fun, I think the changes are a positive...its nice when the crazies have their day.
(July 14th, 2020, 13:46)darrelljs Wrote: If the intent was to measure the best AI, lots of changes would be needed. Since the intent is good fun, I think the changes are a positive...its nice when the crazies have their day.
There is no practical way to run a single "fair" competition anyway, because Civ 4 was never designed to be played in One True Way ... things like difficulty level, barbarian activity, map script, landmass size, ecology, starting technologies, and tech trading toggle all matter a great deal in how different personalities, leaders, and civs benefit.
I don't think the AI actually "knows" any of this at the start, either.
I wonder: if you put all the Civ4 AI leaders into an Always Peace game and disabled every victory condition except for Diplomatic and kept the Apostolic Palace disabled, would the game ever actually end?
(July 14th, 2020, 13:46)darrelljs Wrote: If the intent was to measure the best AI, lots of changes would be needed. Since the intent is good fun, I think the changes are a positive...its nice when the crazies have their day.
There is no practical way to run a single "fair" competition anyway, because Civ 4 was never designed to be played in One True Way ... things like difficulty level, barbarian activity, map script, landmass size, ecology, starting technologies, and tech trading toggle all matter a great deal in how different personalities, leaders, and civs benefit.
I don't think the AI actually "knows" any of this at the start, either.
I still think it's not too hard to run an AI contest (although not very view-friendly) by using the debug tools and running a couple thousand games... Keep it on normal barbs (raging is too much RNG), run it on pangaea and archipelago since those are the only two fundamental map types in terms of AI contact, run a game with/without tech, we can ignore the brokering option, give all starting techs (since giving only starting 2 is measuring tech tree rather than AI personality balance, and again, too RNG), test on small, standard, and huge map sizes. Ignore ecology because I don't know anyone who plays other temperatures (other than for gimmicks).
Custom generate 3 maps for each of these options - (script) 2 * (tech trading) 2 * (map size) 3 * 3 maps = 36 different world builder files, run each simulation 1000 times, with each run designed as a run thru one of Sullla's tournament style pairings - top 1 goes thru on small maps, top 2 on standard, top 6 on huge. 36,000 tournaments (about 432,000 games?) would be very feasible to run on some sort of supercomputer where batches of runs can be done in parallel, especially since civ 4 is a 15 year old game that is not memory intensive. Running everything on the same map (for each option setting) would also be fine, although we'd really have to examine start performance as a category in much more detail - using a mirrored start like last season's final could work in that regard. So about 144,000 games I estimate.
Log the results of each save, run some analysis, bada-bing bada-boom, done.
"I know that Kilpatrick is a hell of a damned fool, but I want just that sort of man to command my cavalry on this expedition."
- William Tecumseh Sherman
(July 14th, 2020, 15:22)Dark Savant Wrote: There is no practical way to run a single "fair" competition anyway, because Civ 4 was never designed to be played in One True Way ... things like difficulty level, barbarian activity, map script, landmass size, ecology, starting technologies, and tech trading toggle all matter a great deal in how different personalities, leaders, and civs benefit.
I don't think the AI actually "knows" any of this at the start, either.
I still think it's not too hard to run an AI contest (although not very view-friendly) by using the debug tools and running a couple thousand games... Keep it on normal barbs (raging is too much RNG), run it on pangaea and archipelago since those are the only two fundamental map types in terms of AI contact, run a game with/without tech, we can ignore the brokering option, give all starting techs (since giving only starting 2 is measuring tech tree rather than AI personality balance, and again, too RNG), test on small, standard, and huge map sizes. Ignore ecology because I don't know anyone who plays other temperatures (other than for gimmicks).
That disadvantages "research Archery right away and build tons of archers", which is actually a sound strategy at high-difficulty raging barbarians, and is what Sitting Bull more or less is made for in so many ways (military flavor; builds lots of units; UB; Protective; isn't aggressive).
Neither what we use nor what you suggest is good for that, though, and I can't think of any other setting where SB's strategy might actually be good ... so Sitting Bull is meme-level bad. (The wildcard game can be raging -- if Sitting Bull ever made it there, he might do very well for himself.)
(June 2nd, 2020, 00:32)Charriu Wrote: I've converted all the data in the BonusInfo.XML into a nice spreadsheet. For anybody trying to make a very natural map you can take this as a reference.
Thanks for the link, although since I don't have a Google account, I can't access it.
(July 14th, 2020, 10:47)ljubljana Wrote: Hmm, after a season and a half with this change active, I'm increasingly starting to feel like turning off the free Deity starting techs has introduced a lot of variance into the competition. We keep seeing civs fail to research Mysticism/The Wheel/crucial early military techs and then ending up severely hamstrung due in large part to that one decision, and I can think of a number of games, both this season and last season, where this played a pretty major role in the outcome.
Mysticism isn't one the free Deity techs, so zero impact there.
The thing is, the free techs actually penalize civs with good starting techs, since they end up starting with LESS techs than the civs with poor techs (who get the good techs as a freebie, and keep the others as a bonus).
(July 14th, 2020, 16:41)Wyatan Wrote: The thing is, the free techs actually penalize civs with good starting techs, since they end up starting with LESS techs than the civs with poor techs (who get the good techs as a freebie, and keep the others as a bonus).
This. There is no perfect way to compare the Civ4 AIs. I think Sullla has come up with a pretty good set of options which also provide an enjoyable viewing/predicting experience. It is also one which he is willing to spend the time doing. If others want to try some different approaches, it would be great to see the outcomes and compare.
On AIs not teching Mysticism, it is at least partly predictable. The AIs with pure military tech flavor will not touch it until they have nothing else to research, or decide they need something further along that path (Hereditary Rule, maybe?). Any AI with cultural or religious tech flavors will obviously go for it almost immediately. I am not sure about the pure science/gold/commerce flavor AIs like Mansa -- how far down the tree do you have to look to find something of that flavor that needs Mysticism as a prereq? They may hit Writing and libraries to pop borders before they go for Mysticism.
The Mysticism avoidance issue is also a positive factor for Creative AIs. Getting those early border pops is supposed to be one of the early challenges for a civ to overcome, and there are multiple ways to do it. That some AIs do not go for any of those methods is a valid competitive factor.
Stalin
He's in a bit of a rough spot, with enemies on all sides : Darius, Washington, Isabella.
I suppose that his supporters might prefer the term "opportunities".
His best chance would be to focus on Washington and somehow prevail in a timely fashion. It could happen, so Stalin doing well and becoming dominant is a real possibility. But he's also at a high risk of dying, and could even be first to die.
I don't think he's got a good shot at winning, though : he essentially shares enemies with Willem, so Stalin doing well would almost certainly ensure Willem's victory.
Cyrus
He really drew the short stick here.
To the north, icy wastes.
To the south, jungle and Mongolia.
To the east, Dutch culture.
His best shot would be to conquer Dutch territory... but he can't plot at pleased, and since his peace weight is nearly identical to Willem's, his hands are tied.
There's a possibility his first two cities miss his copper, which could end up being snatched by Mongolia. Were that to happen, he could be first to die, and the rest of the field in big trouble.
But most likely, he'll spend his early and mid-game capturing barb cities all over the place, and warring with Mongolia.
We could also see our first "long distance wars" : since he might be starved for targets, he could declare on Isabella or even Washington !
So I don't think he's got a real shot at this one.
He could be a late bloomer and finish second, though.
Darius
The Austria-Hungary of this game.
Between Genghis and Stalin, he's already in a tough neighbourhood.
It gets worse : if somebody can beat Willem to the second religion, that's him. And if he doesn't, he shares a river with Isabella, so he'll end up Christian. And owing to peaceweight considerations, he'll beat her as Willem's worst enemy.
So he's surrounded by three enemies... and thus a favourite, if not the favourite, for first to die.
But here's the thing : Willem will also have to deal with Isabella, Stalin with Washington, and Genghis with Cyrus. And Darius has access to early metals and Immortals, so he won't be a push-over.
So if he survives, he could do extremely well, to the point that I think he's the second most likely to win the game.
I don't think that he and Willem can survive both to the end.
Genghis Khan
He'll most probably spend his early game alternating ineffective wars between Darius and Cyrus, progressively growing into irrelevance, and die sometime during the mid-game.
He could even be first to die if the other contenders to that title do better than expected.
But...
As noted earlier, there's a possibility that Cyrus proves extremely weak in the beginning. Should Genghis take advantage, we could see a dangerous snowball forming. And contrary to Stalin, I think a dominant Genghis would stomp out remorlessly a Ducth cultural attempt.
So I'd say he has a better shot than Stalin at winning the game, but a much lower chance of finishing second.
Isabella
I don't think that a repeat of last year's performance is in play here.
For starters, I'm almost positive she'll pull a Justinian and found her Holy City in the tundra.
She'll be squeezed in the west by Willem, in the south by Stalin.
So she'll probably stay on 4-5 cities, locked in wonder-building mode (that she'll lose to Willem), chasing after religions (that she'll lose to Willem).
She'll be fighting Willem a lot, and lose in the end.
Her best chance would be to take part in a dogpile against Stalin, to gain more territory.
Along with Darius, she's my top pick for first to die.
And I'd say she's the least likely of the lot to survive to the end of the game, let alone do well.
And even if she unexpectedly did well, her allies (Washington most certainly, Darius if he doesn't self-found a religion) would be likely to do better than her.
Washington
He has a sheltered start, with decent land (if a bit on the dry side), access to early metals, and room to grow to 8-9 cities.
His game plan is simple : kill Stalin, grow from there.
I think their inevitable confrontation is an even match : Stalin has ultimately access to better land (although some of it is jungly) and is a better AI, but Washington will have allies (Darius and Isabella).
Should he prevail, he'd be in good shape to finish second.
I don't really see him as winning though : too prone to signing peace.
He might be first to die if he bungles his early expansion and succumbs to Stalin early, but that wouldn't be my first pick.
One of Stalin or Washington is going to die, though, that's for sure.
Willem
This guy is a weaker version of Huyna Capac : no Industrious trait to make for his wonder-building addiction, and while Huyna Capac might ignore military techs too long, this guy almost certainly will. He might even stop teching before rifling !
But this game has some aspects that should work very well for him.
As already mentionned, he has a perfect neighbour with Cyrus.
He also has the best second city spot, with an early Stone ressource.
He should found the second religion. If not, he'll almost certainly go for Monotheism.
An early masonry would be good for him : Stonehenge would kickstart one of his three legendary cities and provide prophets for his numerous shrines, Great Wall would shelter him from barbarian invasions to which he'd be particularly exposed otherwise.
His enemies are going to be Isabella (who should be weaker, and facing Stalin too) and Darius (who'll have to contend with Genghis and Stalin as well).
All in all, those should be expansion opportunities more than actual threats.
He should be an early game leader, and stay so until at least the mid-game. Then he'll pull the trigger on a cultural win attempt...
I think he'll prevail in three cases :
- He absorbs the most of Isabella and Darius and becomes the dominant AI.
- He gets to 9-12 cities while the rest of the worlds remains unresolved, with several AIs still competing.
- One AI becomes dominant but fails to attack him in time.
He's only got room for 5-6 decent or half-decent cities, though, so he'll have to gain territory somehow.
And while he'd better research rifling before pulling the trigger, he should stay well clear of Assembly Line : his land lacks health ressources, he runs the risk of starving his cities to the ground !
In conclusion, this should be a very stressful and frustrating game !
If you have Willem as your favourite, expect to be raging at your screen as he gets torn apart by a civ two tech eras behind, because he's carefully avoided researching anything that could give him a tech edge.
Or be ready to facepalm hard enough to leave an imprint for days as he techs to Mass Media... and builds the UN when his opponents are friendly towards one another !
If you don't have Willem as your pick... expect to stare in bewilderment as your civ of choice is running around with massive armies of cavs, while Willem cruises to a cultural win with longbows lining his city walls.
tldr;
One of Darius or Willem will do well and win the game. Advantage Willem, though.
One of Stalin or Washington should do well and finish second. Maybe.
(July 14th, 2020, 20:44)Wyatan Wrote: Game 8 predictions :
Stalin
He's in a bit of a rough spot, with enemies on all sides : Darius, Washington, Isabella.
I suppose that his supporters might prefer the term "opportunities".
His best chance would be to focus on Washington and somehow prevail in a timely fashion. It could happen, so Stalin doing well and becoming dominant is a real possibility. But he's also at a high risk of dying, and could even be first to die.
I don't think he's got a good shot at winning, though : he essentially shares enemies with Willem, so Stalin doing well would almost certainly ensure Willem's victory.
Cyrus
He really drew the short stick here.
To the north, icy wastes.
To the south, jungle and Mongolia.
To the east, Dutch culture.
His best shot would be to conquer Dutch territory... but he can't plot at pleased, and since his peace weight is nearly identical to Willem's, his hands are tied.
There's a possibility his first two cities miss his copper, which could end up being snatched by Mongolia. Were that to happen, he could be first to die, and the rest of the field in big trouble.
But most likely, he'll spend his early and mid-game capturing barb cities all over the place, and warring with Mongolia.
We could also see our first "long distance wars" : since he might be starved for targets, he could declare on Isabella or even Washington !
So I don't think he's got a real shot at this one.
He could be a late bloomer and finish second, though.
Darius
The Austria-Hungary of this game.
Between Genghis and Stalin, he's already in a tough neighbourhood.
It gets worse : if somebody can beat Willem to the second religion, that's him. And if he doesn't, he shares a river with Isabella, so he'll end up Christian. And owing to peaceweight considerations, he'll beat her as Willem's worst enemy.
So he's surrounded by three enemies... and thus a favourite, if not the favourite, for first to die.
But here's the thing : Willem will also have to deal with Isabella, Stalin with Washington, and Genghis with Cyrus. And Darius has access to early metals and Immortals, so he won't be a push-over.
So if he survives, he could do extremely well, to the point that I think he's the second most likely to win the game.
I don't think that he and Willem can survive both to the end.
Genghis Khan
He'll most probably spend his early game alternating ineffective wars between Darius and Cyrus, progressively growing into irrelevance, and die sometime during the mid-game.
He could even be first to die if the other contenders to that title do better than expected.
But...
As noted earlier, there's a possibility that Cyrus proves extremely weak in the beginning. Should Genghis take advantage, we could see a dangerous snowball forming. And contrary to Stalin, I think a dominant Genghis would stomp out remorlessly a Ducth cultural attempt.
So I'd say he has a better shot than Stalin at winning the game, but a much lower chance of finishing second.
Isabella
I don't think that a repeat of last year's performance is in play here.
For starters, I'm almost positive she'll pull a Justinian and found her Holy City in the tundra.
She'll be squeezed in the west by Willem, in the south by Stalin.
So she'll probably stay on 4-5 cities, locked in wonder-building mode (that she'll lose to Willem), chasing after religions (that she'll lose to Willem).
She'll be fighting Willem a lot, and lose in the end.
Her best chance would be to take part in a dogpile against Stalin, to gain more territory.
Along with Darius, she's my top pick for first to die.
And I'd say she's the least likely of the lot to survive to the end of the game, let alone do well.
And even if she unexpectedly did well, her allies (Washington most certainly, Darius if he doesn't self-found a religion) would be likely to do better than her.
Washington
He has a sheltered start, with decent land (if a bit on the dry side), access to early metals, and room to grow to 8-9 cities.
His game plan is simple : kill Stalin, grow from there.
I think their inevitable confrontation is an even match : Stalin has ultimately access to better land (although some of it is jungly) and is a better AI, but Washington will have allies (Darius and Isabella).
Should he prevail, he'd be in good shape to finish second.
I don't really see him as winning though : too prone to signing peace.
He might be first to die if he bungles his early expansion and succumbs to Stalin early, but that wouldn't be my first pick.
One of Stalin or Washington is going to die, though, that's for sure.
Willem
This guy is a weaker version of Huyna Capac : no Industrious trait to make for his wonder-building addiction, and while Huyna Capac might ignore military techs too long, this guy almost certainly will. He might even stop teching before rifling !
But this game has some aspects that should work very well for him.
As already mentionned, he has a perfect neighbour with Cyrus.
He also has the best second city spot, with an early Stone ressource.
He should found the second religion. If not, he'll almost certainly go for Monotheism.
An early masonry would be good for him : Stonehenge would kickstart one of his three legendary cities and provide prophets for his numerous shrines, Great Wall would shelter him from barbarian invasions to which he'd be particularly exposed otherwise.
His enemies are going to be Isabella (who should be weaker, and facing Stalin too) and Darius (who'll have to contend with Genghis and Stalin as well).
All in all, those should be expansion opportunities more than actual threats.
He should be an early game leader, and stay so until at least the mid-game. Then he'll pull the trigger on a cultural win attempt...
I think he'll prevail in three cases :
- He absorbs the most of Isabella and Darius and becomes the dominant AI.
- He gets to 9-12 cities while the rest of the worlds remains unresolved, with several AIs still competing.
- One AI becomes dominant but fails to attack him in time.
He's only got room for 5-6 decent or half-decent cities, though, so he'll have to gain territory somehow.
And while he'd better research rifling before pulling the trigger, he should stay well clear of Assembly Line : his land lacks health ressources, he runs the risk of starving his cities to the ground !
In conclusion, this should be a very stressful and frustrating game !
If you have Willem as your favourite, expect to be raging at your screen as he gets torn apart by a civ two tech eras behind, because he's carefully avoided researching anything that could give him a tech edge.
Or be ready to facepalm hard enough to leave an imprint for days as he techs to Mass Media... and builds the UN when his opponents are friendly towards one another !
If you don't have Willem as your pick... expect to stare in bewilderment as your civ of choice is running around with massive armies of cavs, while Willem cruises to a cultural win with longbows lining his city walls.
tldr;
One of Darius or Willem will do well and win the game. Advantage Willem, though.
One of Stalin or Washington should do well and finish second. Maybe.
Super detailed. I'm curious if you're one of those people consistently scoring like 20 points a game or if you just wanted to share your thoughts. I like to talk about civ4 and post my thoughts and they're consistently horribly wrong and I score I think an average of 7 points per game.
My gut reading of the game is Stalin has got to win with a seeming pushover in Washington as his huuuuge backyard. And because that looks good to me, it almost assuredly won't happen.
(July 14th, 2020, 22:39)Fluffball Wrote: Super detailed. I'm curious if you're one of those people consistently scoring like 20 points a game or if you just wanted to share your thoughts.
Nobody's averaging 20 points a game unless you count one or two people who only guessed one game and presumably got lucky. But if you're having trouble finding Wyatan on the "Contest" page of the Current Bracket and Standings that Sullla links from his site, you haven't scrolled up high enough. Someone else might take the spot on Friday, but as of this week, in total points for the whole contest this season, Wyatan's at the top.
