PCGamer: Soren Johnson says Civ 3's bargaining table was a 'big mistake' (GDC22)

[The_J]

Soren Johnson has given a talk at this year's "Game Developers Conference", titled
"My Elephant in the Room: An 'Old World' Postmortem". In this talk, he apparently also coverd the Civ3 bargaining table, since PCGamer has an article titled "Soren Johnson says Civ 3's bargaining table was a 'big mistake'".
In this article they cover how Soren thinks that this design as a big mistake, since it basically turns diplomacy into a min/max strategy, without much actual diplomacy.

An excerpt:
"In another part of the talk, Johnson argued that while some players may say they want AI opponents to behave like human opponents, they typically don't. When the AI starts metagaming—for instance, by ganging up on whichever player is winning for no reason other than to stop them from winning—the player can perceive it as unfair or out-of-character. In Johnson's experience, players want AIs to act according to their fictional context rather than picking the optimal strategy like a person would."

"Every time the AI got new technology they would [use the bargaining table to] contact all of their friends, rivals, even enemies to see what they could get in return by trading it away, which cost them nothing to do, but would get them a little something in return," said Johnson. "From the human's perspective, it looked like the AIs were a giant tech cartel and were selling techs to each other at bargain prices, but the AI was simply pursuing the optimal strategy.""

This article is well worth reading, just click here: "PCGamer: Soren Johnson says Civ 3's bargaining table was a 'big mistake'".

 

[Civinator]

The "burden of min-max the bargaining table" in C3C is massively reduced by the Flintlock patch.

 

[Hygro]

Civ3's bargaining table was one of the most fun parts of the game.

 

[Jivilov]

It seems most humans play to win and AIs do the same. What's wrong with that? Probably more to the point would be addressing the lack of Shield or Beaker overflow. That's what bugs most players IMO.

 

[vorlon_mi]

A lot of ideas were new / innovative back in 2001 when Civ3 was released. I remember reading advice in these very forums that the Civ3 AI would trade any and all techs among themselves, so the human should be aware of that and account for it.

Just because Soren wouldn't do that again *now* doesn't mean it wasn't the right call *then*, given the state of the computers that could run Civ3. Check it -- Civ3 could be installed on Windows 98, Windows ME, with only 32MB of RAM. Why shouldn't the AI try to gain an advantage to win the game?

 

[Knightfall]

Soren makes some good points, but what would an alternative to this system look like? The idea of diplomacy being largely governed by semi-random events seems like it would become furstrating quite quickly.

 

[AnthonyBoscia]

In this article they cover how Soren thinks that this design as a big mistake, since it basically turns diplomacy into a min/max strategy, without much actual diplomacy.

I always thought the goal of the game was to make zero sum diplomacy the norm for powerful civs. I'm with vorlon on this. It worked well at the time and was a big improvement over one of Civ 2's few real weaknesses.

Civ3's bargaining table was one of the most fun parts of the game.

That and cultivating your land to get kind redcaps and a Pastha to take down eselred. Wait...

 

[Hygro]

I always thought the goal of the game was to make zero sum diplomacy the norm for powerful civs. I'm with vorlon on this. It worked well at the time and was a big improvement over one of Civ 2's few real weaknesses.



That and cultivating your land to get kind redcaps and a Pastha to take down eselred. Wait...

Spoken like a man who knows the deeper min-max

 

[Quintillus]

It looks like I might be in the minority, but I've been turning off tech trading for a few years pretty much due to this problem - the "cartel" and how tech spreads instantly. Over time, I've come to prefer the EU4 model of nations that haven't researched a tech getting a discount, so tech still spreads to slower-researching civilizations, but in a less artificial manner. I think Civ6 went this direction too, although I haven't played it enough to say for sure if it gives discounts or just doesn't allow shopping around by default.

The downside is the Civ3 AI tends not to invest enough in research to keep up with an experienced human at similar difficulty levels without tech trading; so essentially disabling tech trading means turning up the difficulty level a bit to compensate.

The other half is the nickel and diming. I'm split on this. It is annoying trying to get the optimal deal. I like what EU4 added a few years ago, a button to add as much gold as the AI is willing to give on top of the other things you are looking for; it at least solves the tedium while still allowing negotiation, which is generally enjoyable. We want the Maya to help against the Aztecs; how much gold do they want for it? What if we toss in Ivory or Chemistry?

 

[Fergei]

I am a bit confused by this topic to be honest. When I play Civ3 the following are not uncommon.

- AI in Republic gets a big late game tech lead over more military/warmongering rivals. The rivals don't catch up even if they have gold and are not at war with the tech leader.

- AI on a small island falls behind in tech and advances at a glacial pace with tech even if having some gold. In this instance I often swoop in and give them techs for hardly any coin just to make them more of a buffer state against a rival AI. So what are they doing at this bargaining table?

So even if there is a bargaining table, there appears to be some limitations rather than chaotic benevolence. The AI doesn't appear willing to sell a tech for one gold for example. So I've never had a problem with it. The human has huge tech advantages to compensate. For example, racing for Philosophy, picking a Scientific Civ, greater priority given to contacting other Civs etc etc.

 

[vorlon_mi]

The "AI on an island falls behind" happened in my most recent game. Three different tribes (Japan, France, and Sumeria) all started alone on medium sized islands -- big enough for 5 AI widely spaced cities. Each was more than 8 techs behind me and the leaders. It was a continents map; the AI on the other large landmass seemed to trade freely with each other. Perhaps there's a bias for neighbors that share a land border, or a bias against tribes where there is no land border?

It may also make a difference how early the AI's meet each other. Each of the 3 I mentioned were separated from the large land masses by ocean tiles. They probably didn't make contact in the Ancient Age. And if a more advanced AI did contact them, and shop around their techs, perhaps they didn't feel it was a good enough deal to sell techs to the backward tribes, even for just 1 gold.

 

[Spoonwood]

Civ3's bargaining table was one of the most fun parts of the game.

For me it is and it isn't. It's fun to wheel and deal, broker n-fers, sell tech for gold per turn, catch up in tech with trading, and so on. Finding something stupid that the AI will do that I can get an edge like the "leave or declare" trick after using gpt to purchase some tech or gold with a furious AI that is equal or greater to me in power, also can be fun. What I haven't found so fun is the min-maxing aspect of it, and I would say it's not as fun as some other forms of min-maxing, like trying to figure out good worker moves. That the optimal strategy involves ringing up every civ every turn (or during several consecutive turns during key phases of the game) and checking in the default game and then figuring out deals isn't so happy in my opinion. I've used MapStat for years. But, if I'm looking for an n-fer without any utility program I not only need to ring up every civ, I also need to keep track of each tech that they have. Using MapStat it can be tricky to spot an n-fer sometimes, especially when the AIs have several more techs than you, I think in part, because of how the techs get organized on the screen. If I had to remember what every civ had to spot any possible early n-fer exists, I think I just wouldn't bother. If they didn't end up looking like a cartel, getting an n-fer likely would be easier, and cases where the n-fer was worth it would be more common.

The AIs might not optimize every gold in every deal like some of us do sometimes. But, they will always do all the calculations to know whether or not a deal will exist with someone. Many players won't. Those who will, can get into a situation where they end up with information overload, or need to sort out what is different.

 

[Aiken_Drumn]

The "burden of min-max the bargaining table" in C3C is massively reduced by the Flintlock patch.

What does this do?

 

[Aiken_Drumn]

Civ3's bargaining table was one of the most fun parts of the game.

It especially is when you're really screwing the AI and then making even more by trading it on to others!

 

[Spoonwood]

What does this do?

Flintlock in the update notes says:

Ask/offer gold popup autofills best amount

I can't say I understand what this means. It does NOT mean that if you put technologies on the requested table from an AI, and then say "what may I offer you for..." that they will give you the best deal. Playing "guess the number" is still required if one wants to optimize the deal, while the AI will offer you more than it will accept on the first deal.

I'm skeptical of Civinator's claim that the burden of min-maxing is massively reduced (IF one chooses to try to min-max... why do that?). Flintlock wanted to do that, but it sounds pretty difficult.

 

[Spoonwood]

I don't know why Johnson demolished the trade table and went with prescripted answers that one selects multiple-choice style.

So what if one picks the best choice, when there's already some probability of doing such randomly?

 

[Civinator]

What does this do?

If you pay in gold, a suggested sum is inserted for the human player, that in many cases is the optimal amount. To be sure, if this is the optimal amount of gold the human player has to pay, the human player only has to add one gold less to see, if the offer now is denied. So in most cases the starting amount in gold shortens the procedure to find the proper amount the human player has to pay. Even in combined offers, this function can be helpful, so it is all but optimal.

Flintlock today posted an answer about it.

 

[Spoonwood]

If you pay in gold, a suggested sum is inserted for the human player, that in many cases is the optimal amount. To be sure, if this is the optimal amount of gold the human player has to pay, the human player only has to add one gold less to see, if the offer now is denied. So in most cases the starting amount in gold shortens the procedure to find the proper amount the human player has to pay. Even in combined offers, this function can be helpful, so it is all but optimal.

Flintlock today posted an answer about it.

I'm not clear on the conditions. I'm trying with early diplomacy on Sid level, and not seeing anything suggested. I still get "0" as the option initially when I offer purely gold, and don't have enough gold to purchase the techs.

Edit: I checked in the same situation as I did in the patch thread using Alphabet and then inputting gold. I did get a suggested amount, and it was the number exactly. I'm inclined now to say that massive time savings is correct, once you understand the conditions for how it works. Clarifying those conditions if you understand them, might be helpful for some.

 

[Spoonwood]

If you pay in gold, a suggested sum is inserted for the human player, that in many cases is the optimal amount.

It's also when you suggest that the AI pay in gold from what I can tell.

In my example, I ring up the Maya and offer them Philosophy for Mathematics and The Wheel. Smoke will probably like it, so I then click lump sum and I get an autofill of 29. That's the exact optimum number for that deal.

It also appears to work for gpt, though gpt deals have some inefficiencies in many cases.

 

[Aiken_Drumn]

@Flintlock talking about your mod.