Touch up the city governor - suggestions!

This part is a x-post from bug reports:

I guess I've never officially reported this bug, just complained about it extensively elsewhere. The governor thinks we're in vanilla 1.0, and would suck even there. It ignores emphasis commands when building wealth/research now. Take a look:

First, we take a look at a city emphasizing commerce, building something normally:



Looks fine to me. What happens when we build wealth, which supposedly wants to max gold?



We move off the mine to work an unimproved tile that does nothing but eat a citizen slot? What? And before we get any "works as intended" ignorance here, let's see what happens if we just disable the governor before clicking wealth:



More gold. Novel concept.

Alright, next situation. This time, we're telling the governor to EMPHASIZE HAMMERS. In theory, this would work hammer tiles, and when we're not building wealth, it seems to do it like it should:



So, what happens when we build wealth while emphasizing those hammers?



Again, what's up with this? I'll try to keep the forum family friendly here, but believe me it's taking a LOT of restraint. A lot. So what happens if the governor assigned tiles like it SHOULD assign them?



TWO SLIDER NOTCHES there (if this difference is taken across all cities that are building wealth).

Now, I have heard that back in 'nilla building wealth used to actually go through gold commerce multipliers, so maybe this was slightly more passable then (not sure how that worked). Maybe, but I doubt it since we're seeing unimproved tiles chosen that are yielding absolutely nothing to the city, while other tiles are ignored.

Nevertheless, I don't care what happened in history, the governor is ACTIVELY SELECTING TILES TO OPPOSE WHAT YOU ARE SPECIFICALLY INSTRUCTING IT TO DO. But only when building wealth, research, and possibly culture. It is deliberately decreasing the raw output of the output you instruct it to prioritize.

This is a pretty material bug. It would be far superior if the game just ignored that we build wealth and act like we're making a normal building/unit, rather than forcing us to disable to governor to keep it from doing something that is indisputably idiotic.

Suggestions

These could make the governor better, and the AI better when it uses build wealth or research:

1. Make building wealth or research work the same tiles as an ordinary building or unit.
2. Put less emphasis on growing to the cap unless the tile improvements are in place. Working unimproved tiles isn't good for the human and it isn't good for the AI either. Don't do it.
3. Tighten worker improvement assignments based on city emphasis (for better AI, consider having the AI pick an emphasis based on tile criteria and stick with it, should also improve national wonder placement automatically as a result).
4. The build selection governor chooses research usually, even though wealth is *almost always* more powerful, because the research multipliers and even academies are usually available earlier.

Having automated human governors AND the AI governors working the proper tiles would give it a huge boost. The biggest issue here by far is lowering the short (and even long term) output of a city by pulling citizens off of hammer tiles...even with hammer emphasis on...while building wealth/research. Some AIs actually choose to build wealth or research with decent consistency (I know toku is somewhat known for doing this and going early alpha), so this could only help strengthen them.

"Building wealth" doesn't mean that the governor says "do anything to make money" -- it just means "every hammer produced generates 1$ output".

It still goes off and evaluates the worth of food vs production vs commerce.

So when you switch to building wealth, it decides "having the city grow is more important than the $ from the mine". Before you built wealth, it decides "production from the mine is more important than having the city grow".

Next, when you emphasize hammers, you are actually emphasizing production. When you are building wealth, that means that you aren't making any production.

...

1. Make building wealth or research work the same tiles as an ordinary building or unit.

Click to expand...

That is actually off. When building wealth or research, you are making commerce.

Imagine you had a 2 food 3 commerce tile, and a 1 food 3 production tile. When building a unit, going the production may be better. But when you switch to wealth or other commerce-builds, it will usually be the case that switching to the commerce tile is superior -- because you are just turning production into commerce (of some kind).

2. Put less emphasis on growing to the cap unless the tile improvements are in place. Working unimproved tiles isn't good for the human and it isn't good for the AI either. Don't do it.

Click to expand...

Improvements can be built after the fact -- size, meanwhile, takes time to develop. And especially if you have slavery, having excess size is quite valuable, as you can turn it into production.

This does mean that the governor is assuming you have a sufficient supply of workers to improve your tiles. You'll notice that the AI governor auto-build will go off and build workers. Possibly the AI should notice if your worker load is extremely high, and not expect any to show up -- but that could get computationally expensive.

3. Tighten worker improvement assignments based on city emphasis (for better AI, consider having the AI pick an emphasis based on tile criteria and stick with it, should also improve national wonder placement automatically as a result).

Click to expand...

*nod*, I have some non-working code that tries to do this.

It attempts to run a greedy algorithm and binary search attack what the ideal 'price' of various resources are to generate max production, food and commerce from the tile mix around the city. It still needs to be, well, made to work.

4. The build selection governor chooses research usually, even though wealth is *almost always* more powerful, because the research multipliers and even academies are usually available earlier.

Click to expand...

*nod* -- the trick here is to generate the "marginal commerce multiplier" of each kind of commerce.

Ie, what is the marginal trade off of (more science) vs (more cash) using the slider?

And doing the same for culture and espionage.

Then if the entire civilization wants more commerce of some kind, you can see if it makes sense to build another kind of commerce, and move the slider, rather than build the kind of commerce you want to build.

The AI, sadly, doesn't make that kind of trade-off. It is very ... primitive ... in how it deals with the commerce sliders and city building.

"Building wealth" doesn't mean that the governor says "do anything to make money" -- it just means "every hammer produced generates 1$ output".

It still goes off and evaluates the worth of food vs production vs commerce.

Click to expand...

Yes, but there is still absolutely no reason to make building wealth override the emphasis buttons. I'm saying its evaluation is painfully off, and the screen shot showing the governor selecting unimproved tiles that do little more than eat food (while at the or health cap, too!) is pretty convincing evidence.

That is actually off. When building wealth or research, you are making commerce.

Click to expand...

No, you aren't.

Imagine you had a 2 food 3 commerce tile, and a 1 food 3 production tile. When building a unit, going the production may be better. But when you switch to wealth or other commerce-builds, it will usually be the case that switching to the commerce tile is superior -- because you are just turning production into commerce (of some kind).

Click to expand...

Again...this is just wrong. There is a difference between commerce, research, and wealth. You're missing the conversion here. Back in vanilla 1.00 or something, it worked differently, but for today's game, building wealth or research uses to convert directly into or :research:. Directly generating commerce using hammers is IMPOSSIBLE. Also, when doing this, the only multipliers relevant to the hammers converted are hammer multipliers. The gold can later be applied to commerce, but only a tiny fraction of that gold gets re-applied to commerce in that city. If you're running a high science or gold slider, very little of the opposite can be produced by commerce. So rather than what you said, working the mine is almost always superior to working the crap commerce tile for gold. In fact, to pull an even amount of gold with 3 commerce, you'd need to be at 0% of anything else assuming equal multipliers. Humans tend not to dwell there long, and I strongly doubt the AI does for any reasonable periods...at least not on difficulties above noble.

Improvements can be built after the fact -- size, meanwhile, takes time to develop. And especially if you have slavery, having excess size is quite valuable, as you can turn it into production.

Click to expand...

I know that you need time to grow cities. I also have a reasonable assessment of when slavery is a viable means of production. For example, whipping away special tiles or grassland hill mines is going to be inefficient in terms of hammers produced no matter what, assuming you have cap room...and if you don't have cap room, why work UNIMPROVED tiles to grow? After pop 6 or so (I forget exactly) even plains hill mines are more efficient than the whip.

You're not seriously going to try to tell me that working an unimproved grassland tile is a better investment for long-term production, right?

This does mean that the governor is assuming you have a sufficient supply of workers to improve your tiles. You'll notice that the AI governor auto-build will go off and build workers. Possibly the AI should notice if your worker load is extremely high, and not expect any to show up -- but that could get computationally expensive.

Click to expand...

It doesn't mean that. Look at the screen shots again, especially the last 3. With the food available to that city, what improvement, pray tell, would a sufficient supply of workers build that would increase the city's net production...say...ever, holding the techs I had constant?

Not a one. No, the workers if I dared automate them that early would probably put cottages there (even if I DIDN'T build wealth). I didn't have metal casting, and thus no workshops were available, but even if they were they'd not exactly improve the hammer output without more food. Chain irrigation? Setting aside the fact that any assumption automated workers will do that on a consistent basis is a pipe dream, I didn't have CS, so that was also not possible. There was no improvement that could possibly have increased my hammer output there...so why work marginal tiles only when building wealth?

It shouldn't. It's that simple. I'm not the best player out there, but I'm pretty consistent on immortal and have logged a legit deity win or two. I've had many times where I wanted to use my multipliers to convert my hammer cities to wealth (in fact doing that with mining inc is a very common strategy for the HoF deity players). Not one time in dozens of games...not even one...has there been a case where moving off of hammer efficient grass hill mines was to work a 2f tile has been a good idea. Not once. Every time I build wealth, and I mean EVERY TIME I build wealth, either I have to have 0 marginal tiles in the city, or I have to turn the governor off. That should be a hint ----> the way the game handles building wealth is ridiculously stupid...especially if you only have hammer multipliers in that city and are running a high science slider!

And yes, I can GUARANTEE that having it use the same tile choice algorithm presently used for units/buildings is a superior use for emphasize wealth to the garbage it does now. That goes for us, and it goes for the AI.

It's not a hard concept. It's the same reason worker first is optimal in almost every case ---> growing on unimproved tiles is a BAD IDEA. Growing INTO low-yield tiles is even worse. If you compared the long-term hammer output of the two cities, the one that worked the 2f 0h 0c is going to come out behind on EVERYTHING. You'll whip that pop away, and get a crappy ROI compared to just working the mines...it fails in every dimension. EVERY dimension.

For now, just make it like other builds, and at LEAST not override emphasis assignments.

TheMeInTeam said:

Yes, but there is still absolutely no reason to make building wealth override the emphasis buttons. I'm saying its evaluation is painfully off, and the screen shot showing the governor selecting unimproved tiles that do little more than eat food (while at the or health cap, too!) is pretty convincing evidence.

Click to expand...

None of those cities are at the happy cap. One is at the health cap.

And the unimproved tile does more than eat food -- it supports a population point and creates more growth, while waiting for a worker to show up and improve a tile. These are not, I'll admit, immediate benefits. (Actually, in this case, it supports most of a population point).

Have a worker show up, and the AI would value that tile at 2 food 3 commerce (IIRC, the AI values self-improving resources at the value after 2 upgrades, to avoid under-building them -- yes, that is a hack). The population needed to work the tile costs 3 food (due to health cap), making it a -1 food +3 commerce decision.

And if the city grew, it could generate -2 food +6 commerce this way. So you could understand how working that unimproved tile could be in the long-term interest of that cities output.

In this case, you might be saying "damn the future! I need civil service!" The AI currently has very limited ability to vary it's short vs long term strategy -- most of it is hard coded, with some exceptions around panic behaviour (ie, when the AI is in cash trouble, it will activate some panic code).

Again...this is just wrong. There is a difference between commerce, research, and wealth. You're missing the conversion here. Back in vanilla 1.00 or something, it worked differently, but for today's game, building wealth or research uses to convert directly into or :research:. Directly generating commerce using hammers is IMPOSSIBLE. Also, when doing this, the only multipliers relevant to the hammers converted are hammer multipliers. The gold can later be applied to commerce, but only a tiny fraction of that gold gets re-applied to commerce in that city.[/qutoe]
I was using 'commerce' to refer to 'espionage, culture, science or gold, or anything else that is called commerce in the DLL'. Sorry -- there is more than one meaning for the word 'commerce' -- there is the raw commerce from tiles, and the kinds of commerce that it turns into.

The point is, suppose you have a zero-multiplier city. In a zero-multiplier empire (no city has any multipliers). This is a simple case to illustrate the effect.

And you have a 1 food 3 hammer square, and a 2 food 4 commerce square.

When building a unit, the governor might say "3 hammers > 4 commerce + 1 food", and pick the mine.

If, however, you are building wealth, it is (often) stupid to pick the 1 food + 3 hammer tile. Because those 3 hammers will be converted to a (type of) commerce. Naturally, if you have a serious imbalance between your commerce multipliers, this isn't true -- but the AI currently isn't smart enough to know about that.

Now, if you are building culture, because culture is a local commerce, this rule doesn't apply.

If you're running a high science or gold slider, very little of the opposite can be produced by commerce. So rather than what you said, working the mine is almost always superior to working the crap commerce tile for gold. In fact, to pull an even amount of gold with 3 commerce, you'd need to be at 0% of anything else assuming equal multipliers. Humans tend not to dwell there long, and I strongly doubt the AI does for any reasonable periods...at least not on difficulties above noble.

Click to expand...

You can generally pull off a pretty decent trade-off between science and gold. It is true that the AI currently doesn't take the 'global trade off' into account, and does a simplifying assumption that it is roughly 1:1.

You're not seriously going to try to tell me that working an unimproved grassland tile is a better investment for long-term production, right?

Click to expand...

The city is going to grow faster by working that tile. If workers show up and improve the tiles asap, you'll probably end up with more total resources out of the city after a few dozen to hundred turns (depending on game speed).

It doesn't mean that. Look at the screen shots again, especially the last 3. With the food available to that city, what improvement, pray tell, would a sufficient supply of workers build that would increase the city's net production...say...ever, holding the techs I had constant?

Click to expand...

The AI doesn't make decisions like that. Nor should it -- assuming tech is held constant that is.

Emphasise production means "generate the produced building/unit as fast as reasonable" under the current model. If you want to produce more commerce (and wealth is producing a type of commerce, as far as the governor is concerned -- I think it takes into account the multipliers as well), pick emphasise commerce. If you want to avoid making food, pick avoid growth.

Try that last city with avoid growth and emphasise commerce.

Not a one. No, the workers if I dared automate them that early would probably put cottages there (even if I DIDN'T build wealth).

Click to expand...

*nod*, they'd build cottages. And you'd get some cash out of them. Long-term, not short term.

I didn't have metal casting, and thus no workshops were available, but even if they were they'd not exactly improve the hammer output without more food.

Click to expand...

There was no improvement that could possibly have increased my hammer output there...so why work marginal tiles only when building wealth?

Click to expand...

As far as the governor is concerned, the bottom city is producing 0 hammers, 11 gold and 2 science with 3 surplus food.

In your manual allocation case, the bottom city is producing 0 hammers, 17 gold, 2 science, with 1 surplus food.

Emphasise hammers increases the 'value' of hammers produced. But 4*0 is the same as 3*0.

The AI in this case considers 3 surplus food to be worth more than 6 surplus gold.

Hitting avoid growth, or emphasise commerce, might change that decision.

And yes, I can GUARANTEE that having it use the same tile choice algorithm presently used for units/buildings is a superior use for emphasize wealth to the garbage it does now. That goes for us, and it goes for the AI.

Click to expand...

Except you are doing a strategy that the AI doesn't understand. It doesn't leave tiles unimproved on purpose. And it doesn't run a 'single commerce city, lots of other cities building wealth, high science slider, leverage concentrate science multipliers' pattern, because it doesn't understand city specialisation at all.

(Well, that isn't true -- there is code in there so that if your city is sufficiently high in your empire's production rankings, it values production there more -- and similar for commerce. But the effect seems quite minimal.)

Click to expand...

I'd like to make a few comments.

Firstly, there will be differences between what tiles are best chosen for an AI player and what tiles are chosen for a human player using the governor. For a human player, I'd agree with TMIT's position that working the high hammer tiles while building wealth is the more beneficial, since you are trying to keep the science slider as high as possible so the raw commerce in the cities with beaker-multipliers channel more of the raw commerce into research.

But for an AI, who barely knows how to use the sliders, avoiding growth for the sake of building wealth may often be a bad idea. Because of the AI's lack of intelligence, it's usually going to have cities performing better overall when they are bigger, so the preferring 2 over 6 doesn't seem out of place. Once the city actually reaches the happy cap, I'd imagine the AI would start working the mines, and then it might be able to run a spy specialist as well!

Question for TMIT: What happens if you instead emphasize commerce in the second last screenshot?

If, however, you are building wealth, it is (often) stupid to pick the 1 food + 3 hammer tile. Because those 3 hammers will be converted to a (type of) commerce. Naturally, if you have a serious imbalance between your commerce multipliers, this isn't true -- but the AI currently isn't smart enough to know about that.

Click to expand...

Then it shouldn't arbitrarily hind itself. You'd need a very low science slider for the 2f 4c tile to beat out the mine in terms of raw gold yield.

Don't forget, building wealth is generally a temporary process, even for the AI, and if it's building it (assuming it were to do so rather than research) it's probably out of other things to build and needs to REACH TECHS that allow it to build other things. This is something we haven't even considered yet, but is yet another factor in favor of working the hammers.

You can generally pull off a pretty decent trade-off between science and gold. It is true that the AI currently doesn't take the 'global trade off' into account, and does a simplifying assumption that it is roughly 1:1.

Click to expand...

It's even worse than this. You can have a factory, forge, power plant and the AI will still work coastal tiles if you have a lighthouse...over state property workshops...at 100% science...if you build wealth. Even if those are your only multipliers. If the code isn't good enough to handle different situations, they shouldn't use it. The unit build tile choice are better in almost all context, and would be a simple, if stopgap, fix that would be a drastic improvement.

The city is going to grow faster by working that tile. If workers show up and improve the tiles asap, you'll probably end up with more total resources out of the city after a few dozen to hundred turns (depending on game speed).

Click to expand...

Total resources is a poor goal, when looking for one kind of output and only placing that kind of multiplier there. The reason is that output in this city (gold) can be used to fuel more efficient commerce conversion (library+academy city with good commerce) elsewhere. Once you factor in the time it takes to actually get techs that allow for even more multipliers working those cottages first in a hammer city will probably never break even, not even in the ~ 500 turns before time is up (though in practice high level games end much sooner).

And the point remains, the emphasis is on production, and no amount of slavery can make a 2f tile better for production, and without the multipliers you'd need a LOT more than 3-4 commerce to equate to 3 hammers of production.

The AI in this case considers 3 surplus food to be worth more than 6 surplus gold.

Click to expand...

Obviously it does. It's also a poor choice. It isn't worth more than 6 surplus gold at all. I accept the fact that properly evaluating this is outside the realm of realistically designable AIs in the world today, but that doesn't change the fact that what's done here is almost never optimal.

Question for TMIT: What happens if you instead emphasize commerce in the second last screenshot?

Click to expand...

It works the same tiles. The variance on emphasis is minimal when building wealth.

Once the city actually reaches the happy cap, I'd imagine the AI would start working the mines, and then it might be able to run a spy specialist as well!

Click to expand...

If a lighthouse is in, it will work the coast tiles over the mines, and will on occasion use a specialist to slow down or stop growth instead of working the hammer tiles.

Because of the AI's lack of intelligence, it's usually going to have cities performing better overall when they are bigger, so the preferring 2 over 6 doesn't seem out of place.

Click to expand...

But, it's only doing this while building wealth, and the AI in the present form will build most (if not all) buildings available to the city if it isn't building units. I'm assuming if it's out of things to build, it needs to move up the tech tree or it really needs the money. Now. Not 150 turns later when the human has infantry to its muskets.

If the computer has even a crude algorithm for this, it should not take citizens off of hammer tiles in cities where it values production. And for the humans, DEFINITELY remove any variance when building wealth, because it's so rare that it actually does what someone wants that it would SAVE MM to just keep it in building unit mode.

Don't forget, that cities set to emphasize commerce will still work the cottages even without building wealth, and will still work food tiles to grow. The normal governor is far superior to what it does when building wealth or research, I'd go so far to say it's bugged based on the fact that building wealth/research used to work differently, and the governor's general complete uselessness here.

And yes, I'm saying this change will help the AI. I'm quite confident in that, having run some games using both types of governors and automated workers.

Hmm this is maybe why I've always felt like building wealth was weak. It was probably because of the poor choices made by the governor!

I definitely agree something needs work in the case of the governor working 2 2 tiles over 2 grass mines.

PieceOfMind said:

Hmm this is maybe why I've always felt like building wealth was weak. It was probably because of the poor choices made by the governor!

I definitely agree something needs work in the case of the governor working 2 2 tiles over 2 grass mines.

Click to expand...

I KNOW building wealth has potential, ever since Unconquered Sun used caste/sp workshops to win a 1700's domination victory in BOTM 10...which was a deity game. He had tanks and bombers at 1400 AD...and he's also the person who showed you that screenshot of infantry vs muskets/longbows one deity .

Those are feats i've never matched, using anything.

However, the AI doesn't need to capture deity level play in order to improve. The true strength of wealth/research is that, similar to specialists, it allows for output independent of the slider rate. It is, after all, specifically converting hammers to those things, similar to how a specialist converts food. The raw yield of hammer tiles actually tend to be better than specialists w/o rep, with the tradeoff that you don't get GPP. Even if the AI didn't use this well, if it used it at all it would improve.

In the case of Niadros which is 3 over the healthy limit, I would definitely work a mine over an unimproved grassland tile.
(2 food 0 hammer 0 commerce tiles ought to be on the bottom of the list to work when a city is over either the health or happy limit when the city has enough food to feed itself.)

In the case of Burka though; the city is under both the happy & health limits. And you don't have a Lighthouse / Trading Post, so it's setting makes sense to me.

On somewhat of the same topic, I've noticed a bit of odd AI behavior at the end of the game.
When I build "Wealth" : It's first specalist priority is Scientists.
When I build "Reserach" : It's first specalist priority is Merchants.
Seems the AI has the settings backwards to me.

The arguments being used against TheMeInTeam don't pass any common sense test. They are simply irrational constructs. Stop making such stupid arguments please, it makes my brain hurt. Actually it's just irritating. I'm sure if you put your mind to it, making such assinine points as you guys are doing here you could argue your sh!t smells good, but it's stupid, and no one with a lick of sense is going to be persuaded.

In the case of Burka though; the city is under both the happy & health limits. And you don't have a Lighthouse / Trading Post, so it's setting makes sense to me.

Click to expand...

I can't speak for others, but sub-optimal yields do not make sense to me.

On somewhat of the same topic, I've noticed a bit of odd AI behavior at the end of the game.
When I build "Wealth" : It's first specalist priority is Scientists.
When I build "Reserach" : It's first specalist priority is Merchants.
Seems the AI has the settings backwards to me.

Click to expand...

I haven't noticed this pattern, but if it's doing this it could use some touching up there too. If you're trying to produce wealth you're probably trying to bump the slider up, so running opposite specs at city level is indeed questionable.

The arguments being used against TheMeInTeam don't pass any common sense test. They are simply irrational constructs. Stop making such stupid arguments please, it makes my brain hurt. Actually it's just irritating. I'm sure if you put your mind to it, making such assinine points as you guys are doing here you could argue your sh!t smells good, but it's stupid, and no one with a lick of sense is going to be persuaded.

Click to expand...

I definitely wouldn't take this approach in criticizing the arguments here, but I must admit a piece of me thought this way (I'm kind of used to irrational constructs when people have difficulty understanding though, so I'm less offended by it).

Growing to a cap is not good fundamentally, no matter how many times people say so. You have to have a reason to grow to the cap, guys. If the cost of growing is lower than the benefit (true in both city examples), growing is the wrong choice.

Now as for the arguments of "long vs short run", long vs short run is a misleading distinction. The object of the game is to win. Selecting the course of action that optimizes chances of winning IS the best "long run" investment. Making a building that will break even after the game ends is obviously a stupid move, even if it would EVENTUALLY pay off. Compounding this is that unlocking techs earlier allows more potential, so what seems like a short-run investment might increase aggregate beaker output at all periods in time, and therefore be the better "long run" investment also.

Now, I'm not the king of micro or anything, so if you want to see the crazy powerful deity stuff like liberalism at 200 AD or sooner (those guys build wealth too, but also use wonder failure cash and very tight control over GPP for bulbnig), then you'll have to look into BOTM 10 or something. However, if you want an illustration of the power of building wealth over spamming buildings, then here it is:



I spoiler the above because it's from the current immortal university game. This empire already has democracy and is well on its way to 1600's assembly line (yes I won that game, handily). There are maybe 6 libraries (used to culture flip barb cities or more importantly run scientists for GPP) and 1 university (capitol) at this point, the rest is just granary + courthouse. Sounds like suicide, but instead it appears quite sound.

Now, I'm not expecting the AI does this (though it would be strong if it did do it properly). I'm just asking that when the AI does build wealth, it does it correctly.

TheMeInTeam said:

I definitely wouldn't take this approach in criticizing the arguments here, but I must admit a piece of me thought this way (I'm kind of used to irrational constructs when people have difficulty understanding though, so I'm less offended by it).

Click to expand...

You're right of course. But here is the thing. In any common sense analysis you look it 1)is the city earning more Gold (that's what emphasized), or 2) will the AI's decision help it to win. Neither of these are true. In fact they are flagrantly false. And there is zero reason to work a 2 food tile, it does nothing. The only possible argument to make to do this is in a very in depth micro analysis where you are optomizing the whip and overflow for something, and you might do it for a couple turns. But the AI is no where near that level. So like I stated, none of the arguments presented passed any rational analysis. They just lacked common sense, and it irks me when people argue things just to argue. I see no other reason to dispute your findings. By any reasonable measure the AI is making a poor choice of which tiles to work.

Going by TMIT's last two screenshots,

Governor behaviour creates:
8,44 over 4 turns followed by
21,133 over 7 turns, plus 1 into growth towards 7pop.

Total:
29,177,1

TMIT's working creates:
22,187,0 over the same 11 turns.

So the comparison we are making here is 7+1 vs. 10.

This is not outrageous as phungus seems to be suggesting, but I agree it is sub-optimal (as much as I hate that phrase, lol).

The difference is less pronounced because it is a low pop city. When it activates this behavior in larger cities, the change is more painful.

Considering I have a library and academy elsewhere, gold is a lot more valuable from that city than beakers.

So an obvious improvement would be to keep track of the slider multiplier trade off for gold vs research.

1 unit of gold produced is worth (# of research gained from +10% slider)/(# of gold lost from -10% slider) units of research, at the least.

You could do the same for science vs gold. And bias cities who are building wealth/research to work on the commerce type that has a lower empire-wide multiplier.

The AI, however, currently doesn't do long-term plans of this kind: they build wealth/research when they are in deep do-do and/or out of other things to build. So such a change wouldn't make the AI better (it would, however, make the governor better for players, even ones not trying to do the above trick).

In the above situation, what happens when you select 'avoid growth' by the way?

TheMeInTeam said:

Considering I have a library and academy elsewhere, gold is a lot more valuable from that city than beakers.

Click to expand...

That may well be, but does the AI know that?

Hint: the answer starts with 'n' and ends with 'o'.

I just want to mention that what you want of the AI is not exactly straight forward to program, at least in the framework that already exists. You want it to understand how to use the slider at more advanced level, to coordinate city governor decisions with future worker actions (not possible at the moment), knowing under what circumstances it is wise to use build wealth and under what circumstances it is not.

These are all achievable goals I guess, but things like this are not as easy to fix as they look.

By the way, concerning your number 2 suggestion, that is not clear for coding because there are in fact tiles that are good to work unimproved. An inland financial lake, or any inland lake with a lighthouse is a good tile to work. As are things like oases, unimproved food resources etc.

I think to further strengthen your case for your suggestions you will need to find a more extreme example of governor stupidity (which I'm sure there are many) to highlight what is going wrong. I can't say Birka seems to indicate a huge problem.

That may well be, but does the AI know that?

Hint: the answer starts with 'n' and ends with 'o'.

I just want to mention that what you want of the AI is not exactly straight forward to program, at least in the framework that already exists. You want it to understand how to use the slider at more advanced level, to coordinate city governor decisions with future worker actions (not possible at the moment), knowing under what circumstances it is wise to use build wealth and under what circumstances it is not.

Click to expand...

UNTIL the AI knows that, it shouldn't be trying. One of my suggested fixes is quite easy:

Have it treat building wealth the same as anything else. Even though that lacks any strategic considerations whatsoever, its effects are still almost always superior to now. Now I'm not a programmer, but I'm assuming using code that already exists isn't a big problem.

And yes, birka is a serious problem because that issue can be dragged across an empire (and in the AIs case, frequently a large one) to account for 60-70 bpt variance or more. That's not immaterial! And birka is easily fixed by getting rid of the game's flawed special treatment of wealth/research.

The AI, however, currently doesn't do long-term plans of this kind: they build wealth/research when they are in deep do-do and/or out of other things to build. So such a change wouldn't make the AI better (it would, however, make the governor better for players, even ones not trying to do the above trick).

Click to expand...

What I'm suggesting would still help the AI, although possibly less than governor users.

In the above situation, what happens when you select 'avoid growth' by the way?

Click to expand...

Although I didn't try it, I can 99% guarantee it WILL work the mines in that case (although note that it would run specialists if it could in some cases). However, that comes with its own set of problems...namely when the city still CAN grow by working its best special tiles and hammer tiles, and would do so just perfectly if not building wealth.

I have a couple problems with other governer selections, that I *thought* were BetterAI related but perhaps they were original.

First look at the attached images. In the first screenshot, I have no emphasize buttons selected, and the governer is maxing out production. In the second I select 'avoid growth', and the governer moves a worker from production (and stagnant growth) to a different tile.

I am guessing that the AI is coded to grow the food bar to max under 'avoid growth' conditions to permit quick expansion when it is deselected, but the actual performance in this case is poor. Hopefully a fairly easy fix.

A harder issue has to do with prioritization early in the game. I frequently see the governer using a lot of low production tiles in an effort to grow - leading to 'growth in 20/production in 20' scenarios. When I am in a situation where I need to build a work boat to get resources, for example, I'd rather stagnant or even negative food rather than trying to grow 2F at a time.

stagnate said:

A harder issue has to do with prioritization early in the game. I frequently see the governer using a lot of low production tiles in an effort to grow - leading to 'growth in 20/production in 20' scenarios. When I am in a situation where I need to build a work boat to get resources, for example, I'd rather stagnant or even negative food rather than trying to grow 2F at a time.

Click to expand...

The game does figure these things out sometimes. Building a granary, for instance, automatically shifts work tiles to finish the granary ASAP.

That's a completely internal decision though. That affects something the governor can see immediately, i.e. the turns to completion, and what tiles will improve this percentage. Even so, it's just been told to do this by an arbitrary script. It doesn't know anything. It especially doesn't know why.

In the words of a friend (who might not consider me the same, but is still glad to see me when he does), "AI doesn't think, it functions". You can give it arbitrary rules, but it's never going to adapt to circumstances unless it's got a very detailed rulebook for doing so. Even then, if there's data out of reach of the system, i.e. a special resource that needs a work boat, then it's just not going to know that building a work boat will get it more food faster. It can't know, since it's never been given a list that says so and a means to see the variables involved.


Taking a direct hand in tile placement is something you're just going to have to find a comfortable medium with. It'll never be as good as it could be. The AI will never guess your strategy, and will be working to its own, coded by the good ol' boys in Firaxis who don't think outside the box because they BUILT the damn box. The AI's strategy works in sync with its other elements, and this governor meshes best with an AI player's grand strategy.

Sooner or later, you just have to figure out which cities you want to control and which you don't, and for how long. If you leave it to the AI, you accept it'll make so-called "mistakes", but you accept it because you can't micromanage everything. All you can do is watch the AI make its "mistakes" and correct them as you see them, directing it away from its own strategy towards your own. That's all.

There's some things you can fix, and some you can't. I can't say the root issue for this thread is a fix-it item, since as others have argued the AI is working to a valid strategy (a great person, for instance). It's just not an optimal strategy...but who the hell wants to fight an AI that always uses an optimal strategy? I don't play Deity...I stick with Noble, where I know the playing field is level. I don't game the system, and neither does the AI. And you know what? The AI never should. It should never be optimal. Otherwise, the human player could never catch up on a level playing field, much less an un-level one.

I can't say this change would make the game more interesting. Just less annoying to you. There's already a way to fix that though...without recoding the game. It's still annoying, but at least you get to make the choice. Besides, we should all be playing a bit less with the governor.


And one other thing: Me, be nice; and Phungus, be nicer. Me may have rebuked you, Phungus, and I totally respect him for that, but he's still being a nasty guy about this, and you both need to be nicer here. Be tolerant of the fact that other people will see a logic to this that you won't. Sometimes, just because you can't see the logic, doesn't mean it isn't there. They just have a different perspective than yours. It's not a wrong perspective, it's just one that sees something that you don't (and which they might not be explicitly stating). Some people can mistake that for illogic...

Before we get into any more heated arguments, let's consider that this is a system that'll always be imperfect. There are some improvements that are obvious and some that aren't...and sometimes, a fix for one person will screw up ten other people. Let's be careful with our pet issues here. We're here to talk about BetterAI, which is ultimately a mod for PEOPLE, not machines. Make the AI a more interesting opponent, not just a more efficient one.

And in case you're wondering where my "you are wrong" argument is in this wall of text, well, too bad, there isn't one. That's because I can't say you aren't. I have a hunch, and some arguments to support it, but that's it. Hell, I expect I won't see this thread again. I tend to avoid threads with a lot of negativity. Not unless my own bad forum habits have changed.

And don't think that I haven't seen this "start out nasty, become reasonable near the end to put the other guy off his guard" trick to win arguments before. It makes the other guy hesitant to call you on it, since you're being so reasonable...but it's so obvious when you've seen it a few times before. Hell, I bet you don't even realize you're doing it...the first step is admitting you have a problem...