AI worker Inefficiency

[dexters]

I was recently watching the AI play through a game exclusively against themselves in PTW's scenario debug mode. No rule changes. Regent difficulty. Standard map 100X100

Read This Thread for the details and other AI observations.

Skipping ahead, the Chinese whooped the Romans in the late ancient era. After peace was settled, the Romans were essentially in a though spot.

I still haven't figured out if there is a way to access the AI civ's trade/tech/diplomacy screens in debug (Firaxis should really add this in their next patch) but from their city screen, I could see that the Romans were having money problems.

What it did then was it started disbanding workers, outside cities no less, so they don't even recover the shields. What's worse, a few turns later, they would build another worker, then a turn later, they would disband another worker. The workers are also built from their Core cities (pop 12) high production, low corruption. So now, not only are the Romans producing and destroying workers, they are losing out on production and commerce from population drops.

I'm beginning to think that this may be a contributing reason to why Civs that are beat down early on tend to not recover on their own. They run into cash crunches and just do things like this that cost them even more.

Any suggestions for fixes? I'm trying to think of something, but I'd like to hear from the rest of you.

[badams52]

"We're running out of gold! Let's disband a worker."

"We need to make our cities more productive! Let's build more workers."

[dexters]

In case anyone want to see it for themselves, here is the debug game I was watching.

Watch the Romans.

You need PTW to run this.

[dexters]

A potential fix for AI worker inefficiency is to artificially give AI foresight.

worker management is all about knowing how many workers you want in the future and keeping a checkbox in your head checked at all times that because you need there workers for the "future" if they don't have a job to do, the last thing you should do is disband them.

1) Every 5 or so or however many tunrs, the AI asses its empire and decide on its optimum worker count. Once it is decided, this number is FIXED.

If workers are lost due to war, they are replaced after a given fraction of the workers are lost.


Alternatively

2) We can use a worker maximization model with constraints, like a standard linear programming problem. Constraints to the model would include terrain improvements required, ratio of workers to empire size, ratio of workers to military size, existing workers, and gold income.

The gold constraint should be somewhat loose, perhaps with a nice range of say 20 gold.
This way, a sudden bump in the budget of 2 gold from a city improvement doesn't immediately require the disbanding of worker units.

The 2nd alternative sounds more flexible and follow the spirit of the game AI, which was to create a set of basic instructions, without too much specifics. So naturally, in playtestings, there may be tweaks that is required and who knows what else might happen.