(3-NS) Change Deal Price for Human-AI Trades

[azum4roll]

Glossary:
buy price = the maximum value the buyer is willing to offer for the resource
sell price = the minimum value the seller is willing to sell the resource for
deal price = the actual value that the resource is being traded for

The proposal:
AI offers to buy something from human at the average of their buy price and the perceived sell price of the human. This price CANNOT be increased (AI treats this as their actual buy price for deal purposes). Both AI and human gain trade opinion modifier with each other based on the difference of deal price and buy/sell price.

When human wants to buy something from AI, AI should charge them the average of AI's sell price and the perceived buy price of the human. This price CANNOT be decreased (AI treats this as their actual sell price for deal purposes). Both AI and human gain trade opinion modifier with each other based on the difference of deal price and buy/sell price.

Something includes everything that's not gold or gold per turn. Technically it can also include gold which always has the same buy/sell price, but for the sake of implementation, we're not counting it.

Example 1:
Civ A, a human player, is trying to sell Cotton to Civ B, an AI player. Using resource valuation, the AI is willing to pay 72GPT for the Cotton, and it thinks the human wants to receive at least 6GPT from selling it.
CURRENT behaviour: AI offers the deal at 72GPT, and the human happily accepts it.
PROPOSED behaviour: AI offers the deal at 39GPT, the human cannot increase the price but still happily accepts it, and the AI likes the human more now as if the human just gifted it 33GPT.
Note that if this was an AI-AI trade instead, the deal would have gone through at the same price, and both sides gain the positive opinion.

Example 2:
CIv A, a human player, wants to buy Silk from Civ B to start WLTKD. Using resource valuation, the AI is willing to part with the Silk for 4GPT, and it thinks the human is willing to pay at most 32GPT for it.
CURRENT behaviour: AI offers to sell at 4GPT, and the human happily accepts it.
PROPOSED behaviour: AI offers to sell at 18GPT. If the human accepts it, the AI likes the human more as if the human just gifted it 14GPT.
Note that if this was an AI-AI trade instead, the deal would have gone through at the same price, and both sides gain the positive opinion.

Rationale:
1. The human is gaining a lot from luxury deals as sell price and buy price differ too much. This is unsolvable through changing the valuation formula since a duplicated luxury really does worth nothing 99% of the time.
2. This reduces human gold income and may make working merchants and building gold buildings viable again.
3. Both sides of trade gain positive opinion of each other because

i) a positive opinion reduces the incentive to war and end the deal early, which is not beneficial to both sides;​

ii) positive opinion is already gained through trade routes, and trade deals should work the same;​

iii) both sides really do gain the equivalent of GPT from the deal.​

Note:
1. This is a copy of the vetoed proposal last month (because of a bug that has been fixed this version), but now it includes ALL goods that can be included in a trade deal instead of just luxury resources.
2. This proposal does not affect AI's valuation of resources/votes/etc. in any way.

 

[ma_kuh]

This seems objectively better than the current situation. That said, I'm curious how the following example would play out:

Civ A, a human player, wants to buy Silk from Civ B, and does not need it to trigger a WLTKD (they are pre-purchasing it, arguably "just for the happiness"). Using resource valuation, the AI is willing to part with the Silk for 4GPT, and it thinks the human is willing to pay at most 6GPT for it.

So the deal price will be 5GPT, and then the next turn the player's city ends their WLTKD and maybe rolls this new resource, paying off the player for the risk.

So the player (and AI, for that matter), become extremely incentivized to purchase as many extra luxuries as possible, regardless of WLTKD requirements, before "prices skyrocket". I'm sure there's some true-value calculation you could make for when it is worth paying 36GPT for six luxuries versus 32GPT only when you need the luxury, and how that factors into likelihood to cover future WLTKDs, but it seems like a significant change in the luxury meta to enact this. At the very least, we should consider teaching the AI how to leverage this system when they can as well (if they don't already).

Again it seems objectively better than the current situation, but I want to think through possible repercussions in other parts of the trade system, and maybe get a head start on possible ratification tweaks to look out for as this change is play-tested.

 

[azum4roll]

That would be a problem of resource valuation, which AFAIK hasn't been changed since the WLTKD system changed.

To put it VERY simply, a luxury should be worth very little (but not nothing anymore) if no city is asking for it, since it's a gamble whether this purchase will trigger WLTKDs in the duration. The value of a resource-triggered WLTKD should be unchanged, but the price should be - there's now a higher supply of WLTKD. So overall there should be a decrease in deal price for requested resources and an increase in deal price for unrequested resources. There should still be a gap, especially if the buyer's cities are already asking for something else.

Economics don't work like that? Who cares, it just feels right.

There's IMO no way to have a fair pricing system unless we make cities ask for a new resource every turn and resource-triggered WLTKD only lasts one turn.

 

[Recursive]

Proposal failed due to lack of sponsorship.