Wonders being completed ONE TURN before you

poom3619 · Oct 6, 2015

That's cute, I've spend Great Engineer in a wonder and then lose the wonder, Once, But that's hurt enough.

To anyone who looked at AI code. Can you confirmed that "snipe player wonder in one turn" isn't coded in invisible ink? /jk

TBH, Building military should be a better alternate from wonder fail-spamming.

stormtrooper412 · Oct 6, 2015

claudiupb said:
Manpanzee is actually right when saying that the probability of losing by a single turn is higher than losing by a greater number of turns, it's not really debatable because it has a mathematical demonstration behind it. If someone were to make a mod when you get notifications each time an AI loses a wonder by a single turn you'd be surprised how often that happens (and would probably explain why AIs covet wonders you currently own ).

I highly doubt that. It would imply the AI has the "check if anyone else is building wonder X and don't rush it if it's not the top priority but still calculate to finish first anyway" which nobody mentioned yet?

I am actually quite flabbergasted at the fact that actual people looked at actual code and saw actual routines being actually executed but no, it doesn't count because I've seen a few odd examples that may or may not agree with that. That's like saying 2+2 is 5 because I asked 10 people and half of them said it was 5. Really sorry if I sound like a jerk but it had to be said

poom3619 · Oct 6, 2015

^ Well, this.

Can someone link/make a mod that make an observer able to see AI's production as if it was controlled by player. That would be an actual solid argument for "AI stole my wonder" if AI proved so.

Kriogen · Oct 6, 2015

AI does troll players here. Why does it hold WW by 1 turn?

Its not that AI is in the middle of building a WW and Shaka attacks so its a good idea to switch to building army. I could understand that. But it's peace. What is so important to not finish World Wonder when it needs just one more turn?

ShakaKhan · Oct 6, 2015

As to whether or not "there's nothing in the code therefore the AI isn't programmed to screw you out of a wonder by one turn" or "it happens so frequently that there must be intent here," isn't going to reach a consensus. Acken has an inarguable point that there's nothing in the code, and computer programs don't do what they aren't programmed to do and are incapable of malicious intent (though this game tests that.) Hugh also has a point that if it their finish times change with reloads to match yours, then there's something fishy. To me, it seems kind of similar to string theory or Freudian psychosexual stages: there's absolutely no definitive proof that either of these concepts exist in the real world, but there's something too coincidental about how all the non-definitive data compiles so nicely...

I think what to learn from this is how to react when it happens. On King difficulty and below, you slam your fist on a table, tear out a few hairs and continue on; you can still win the game without the bonuses of any wonders. On Emperor or Immortal, you're allowed 1 mistake: You can take a gamble on ToA or Petra, but if you lose you'll need to make up for the lost production turns, the gold you get is not enough compensation so now you need to compensate for it with uber-safe play. Can't afford two mistakes like this. On deity, the gamble is akin to an all-in shove in poker. Attempt it sparingly and only in the circumstances most likely to succeed (if you can't identify these, then don't try.) If you fail, you lost all your chips and are out of the game, so just consider who you'd like to play as next.

docbud · Oct 6, 2015

^^I definitely slam my fist and curse pretty loudly (on Emperor). My upstairs neighbors must think I'm in a lunatic!

But I can make more than one mistake and still win.

ShakaKhan · Oct 6, 2015

I'd highly recommend the following experience to all players who are having incredible frustration, rage-quitting, or potentially losing interest in the game(heaven forbid) due to the AI wonder-sniping by 1 turn.

Try a game with a powerful, unbalanced civ (Babylon, Poland, Korea, or a properly-situated Aztec.) Turn the difficulty down one notch, re-roll until you find what you consider to be a ridiculously good map setup - maybe even download the IGE just to save some time and tailor the best conceivable (but realistically possible - no 9 luxury capitals) map. Now go through the game without building a single world wonder. Your goal is to create the most runaway, powerful civilization that you can manage; put yourself in a position where you're a god among insects and see just how powerful you can become. When a victory condition becomes available, don't complete it, rather continue on to see just how powerful you can get. Can you reach 700 GPT? 1000 GPT? A treasury of 150K? 500K? Can you complete 4 SP trees and have a well-developed Ideology? A military 10 times stronger than the rest of the world combined? Eradicate everyone else down to just you and one other crippled capital? Then remember that you accomplished all this without building a single world wonder.

I did this when I had trouble moving from Prince to Emperor because I had difficulty accepting what some of the veterans were saying. I understood that engaging in wonder construction was a gamble but couldn't imagine I'd gain enough power without the bonuses. And I especially had difficulty accepting that even if the wonder paid off, it wasn't worth the opportunity cost. After doing so, I found that my expansion occurred significantly earlier, helping the capital a bit but having powerhouse secondary core cities. Military was 3 or 4 times as strong as it was before, now I was dictating who declares war on who, and when. Big bonuses were lost, but more options and better options were available.

Then try to same experiment on your current difficulty level. Not only can you further your understanding that wonders and player score don't indicate power, but also you can take note of when wonders are being completed so you can better assess when they are an option.

Sincere apologies if this comes off as condescending or dismissive, but world wonders in this game are akin to training wheels, and the game becomes much more flexible and consequently engaging when you take 'em off.

claudiupb · Oct 6, 2015

stormtrooper412 said:
I highly doubt that. It would imply the AI has the "check if anyone else is building wonder X and don't rush it if it's not the top priority but still calculate to finish first anyway" which nobody mentioned yet?

I was arguing that statistically it is more likely to lose wonders by a turn, not that the AI has a routine that does this deliberately. If two players start producing the same wonder at the same time, it is very likely one of them will lose with one turn left. The difference in hammers can be rather small, take LToP for example. It costs 500 hammers, and if you can complete it in 12 turns it means that your city has between 42 and 45 hammers. If another player start producing it at the same time and has a 39-45 hammer output one of the players will lose by just one turn. There are 8 players on the map, and many of them reach certain techs about the same time, and have similar science outputs meaning similar population and a similar hammer output. Sometime a wonder is attempted by 4-5 players at the same time. Statistically it is much more likely that one player will lose by one turn, than none of the players losing by one turn.

The other statement I made was that sometimes AIs will pre-build wonders and switch to something else unexpectedly, that is something observed if you check each turn the espionage screen. I don't know if someone found this behavior while snooping through code, but it is something the AI often does. Why it does that, it is an interesting question. I highly doubt it's a scheme introduced by the devs to trap players into building wonders because in most cases it works against the AI, and the AI loses a lot of hammers because of it, because it will lose wonders because of this.

Sometimes you are unlucky and you will lose a wonder no matter what, but it really doesn't mean that the AI was cheating, the AI beat you fair and square. It either started at about the same time as you but had a bigger hammer output, or it started way before you and it had the wonder pre-built and didn't' finish it earlier but it does finish it when it notices a threat to losing it. So in fact you never had a chance to build that wonder in the first place unless you somehow hide the fact that you are building the wonder (which is impossible once you give out embassies).

wideframehoss · Oct 6, 2015

ShakaKhan said:
I'd highly recommend the following experience to all players who are having incredible frustration, rage-quitting, or potentially losing interest in the game(heaven forbid) due to the AI wonder-sniping by 1 turn.

Try a game with a powerful, unbalanced civ (Babylon, Poland, Korea, or a properly-situated Aztec.) Turn the difficulty down one notch, re-roll until you find what you consider to be a ridiculously good map setup - maybe even download the IGE just to save some time and tailor the best conceivable (but realistically possible - no 9 luxury capitals) map. Now go through the game without building a single world wonder. Your goal is to create the most runaway, powerful civilization that you can manage; put yourself in a position where you're a god among insects and see just how powerful you can become. When a victory condition becomes available, don't complete it, rather continue on to see just how powerful you can get. Can you reach 700 GPT? 1000 GPT? A treasury of 150K? 500K? Can you complete 4 SP trees and have a well-developed Ideology? A military 10 times stronger than the rest of the world combined? Eradicate everyone else down to just you and one other crippled capital? Then remember that you accomplished all this without building a single world wonder.

I did this when I had trouble moving from Prince to Emperor because I had difficulty accepting what some of the veterans were saying. I understood that engaging in wonder construction was a gamble but couldn't imagine I'd gain enough power without the bonuses. And I especially had difficulty accepting that even if the wonder paid off, it wasn't worth the opportunity cost. After doing so, I found that my expansion occurred significantly earlier, helping the capital a bit but having powerhouse secondary core cities. Military was 3 or 4 times as strong as it was before, now I was dictating who declares war on who, and when. Big bonuses were lost, but more options and better options were available.

Then try to same experiment on your current difficulty level. Not only can you further your understanding that wonders and player score don't indicate power, but also you can take note of when wonders are being completed so you can better assess when they are an option.

Sincere apologies if this comes off as condescending or dismissive, but world wonders in this game are akin to training wheels, and the game becomes much more flexible and consequently engaging when you take 'em off.

This is fantastic advice. Once you can routinely build a strong empire without wonders you can start to pick and choose which wonders you want particularly in the late game.

Acken · Oct 6, 2015

poom3619 said:
^ Well, this.

Can someone link/make a mod that make an observer able to see AI's production as if it was controlled by player. That would be an actual solid argument for "AI stole my wonder" if AI proved so.

There are already built in ways to do that. You can observe the AI with the firetuner on autoplay (maybe you need EUI for that to work somehow I'm not sure) or you can also enable logging to see their decisions.

beetle said:
Would you happen to have the code for how the AI decides to juggle the items in its build queue? From what I have seen from other examples, it will need some explanation...

The explanation is that the AI does something else. The AI choses production based on some conditions and then pushes it in the queue at the top. If it selects a wonder already started it will then go at the top of the queue. But none of it has any check for wonders being built and trying to steal them. To the best of my ability

If someone with better knowledge of it proves me wrong that's fine. It's a complex part of the code and I cannot just copy paste 10 lines on the forum.

When chosing what to do the AI launch a function. In this there is a parameter for interrupting the previously built wonder. But the use of this parameter is based on factor relative to the AI, not the player. A war for example. This can put the previous wonder on hold regardless of how many turns left. And iirc a wonder can be chosen as the effect of a war. Would such a coded AI really have strategies just to troll people ? This AI that at immortal doesn't fill universities or doesn't even build them before T160 on immortal ? That AI that choses sacred sites even if it has no faith buildings ? etc etc.

Then when "emergencies" are dealt with it will simply resume the queue.
If someone gets screwed in the process it is just poor luck.

On top of it a lot of these decisions are based on a random selection among a few objects. I already said in that thread or another one cant remember that the only thing an AI does is that it looks at what other players are building to DECREASE the likelihood of selecting a wonder. Of course it only matters during the selection, so if the AI has already selected a wonder, your selection has no effect on this weight. This is done in this code chunk:

Spoiler :

Finally the actual probability is high. For a simple reason.
It is very important to know that the AI is programmed to try wonder after wonder. What it means is that you have multiple AI trying the same wonder. Once this wonder is completed by someone. All these AIs (and the one that completed it) will start a new one. The probability that you, selecting a wonder exactly at the same time as all these civs to lose it again is very high. And since almost all of them have been started at the same time the probability of losing it in the last turn is high too.
That's just how it is. Don't play like an AI chosing a wonder everyone can build everytime you lose one.

ShakaKhan · Oct 6, 2015

claudiupb said:
The other statement I made was that sometimes AIs will pre-build wonders and switch to something else unexpectedly, that is something observed if you check each turn the espionage screen. I don't know if someone found this behavior while snooping through code, but it is something the AI often does. Why it does that, it is an interesting question.

I'd imagine there's something affecting the status of that civ when it decides to change build que and then revert back. I'm not one of the code-reader people, but I imagine, or guess, that the AI utilize some sort of trouble-shooting algorithms when their status is disturbed. So in the events where they change from wonder to units, was there a war recently declared on them or did they lose several units in an ongoing war? Did they change to markets if there economy suddenly tanked? Additionally in the instance of economy, did the player recently make a number of lump-sum resource trades reducing their economy from a few hundred to zero?

Just seems very likely that there's something more benign that triggers the change that's more along the lines of, "compensating for distress" rather than, "that'll piss him off good."

docbud · Oct 6, 2015

Thank you for the explanations, Acken.

Sclb · Oct 7, 2015

Actually I frequently build Machu Picchu or Oracle on deity and doesn't care about losing it or not where it's at a point where there isn't many better things to build waiting for Universities. I don't see why people thinks they frequently lose by 1 turn. And you can easily see AI don't cheat by spying on AI that's attempting to wonder whore. You will also see that sometimes the AI is even using unemployed citizens to maximize production. If you lose, you simply don't have enough production.

beetle · Oct 7, 2015

Acken said:
Would such a coded AI really have strategies just to troll people?

I do not believe that the AIs were specifically programmed to beat out the player by a single turn. The explanations on this thread cover 90%+ of the phenomenon.

But it does not account for experimenting that some people have done where they reload back several turns to shave one turn from their build time to get a WW one turn sooner -- but then still lose it by a single turn! What is best explanation for that? RNG?

docbud · Oct 7, 2015

^^That's what I'm wondering as well. I've gone back several turns and shaved off a few turns to get the wonder, but I still lost by one turn a few turns earlier. It felt that it was predestined that the AI get it (which I know must be impossible).

reddishrecue · Oct 7, 2015

You could change that if you have more than 2-3 turns left for the Ai to complete the wonder by purchasing a great engineer before the Ai could finish. With enough faith and tradition completed, purchase a great engineer, hurry production of the wonder and then get the wonder completed 2 turns later. Purchase, hurry and then next turn its complete.

Acken · Oct 7, 2015

beetle said:
I do not believe that the AIs were specifically programmed to beat out the player by a single turn. The explanations on this thread cover 90%+ of the phenomenon.

But it does not account for experimenting that some people have done where they reload back several turns to shave one turn from their build time to get a WW one turn sooner -- but then still lose it by a single turn! What is best explanation for that? RNG?

Yes.

It doesn't happen every single time. I have done my fair share of reloading in the past and got some, missed some (that I would have missed otherwise).

Iberian · Oct 7, 2015

Does anyone have a set of saves showing this? Say one save 10 turns prior to losing by 1 turn and the one before the loss?

I have been playing this game since the beta and I have had my share of reloads and I have never had a failure when shaving a turn off.

Sclb · Oct 7, 2015

Sometimes when you reload, due to RNG, the AI changes priority and shift their queues. It can either delay or speed up their wonder. That's how people can still lose wonders. Also AI can trigger golden ages to suddenly shave off few turns.

docbud · Oct 7, 2015

Iberian said:
Does anyone have a set of saves showing this? Say one save 10 turns prior to losing by 1 turn and the one before the loss?

I have been playing this game since the beta and I have had my share of reloads and I have never had a failure when shaving a turn off.

Sorry--but I don't keep saves for every game and for years and years.

Wonders being completed ONE TURN before you

Ping Pang Poom!

Peacemongering Turtlesaur

Ping Pang Poom!

Warlord

King

Emperor

King

King

Chieftain

Deity

King

Emperor

Emperor

Deity

Emperor

Some dude on civfans

Deity

Prince

Emperor

Emperor

Similar threads