A closer look discussion of tech trading since V1.29

[cracker]

I can add that nationalism seems to be one of those techs that the AI's would rather go to their grave with in V1.29 rather than give it to you in a peace deal.

In a recent game, I had saved Otto from death twice by giving him handfuls of new cities and then I gave him 4 technologies (Gravity, Magnetism, Chemistry, and Metal) all for free just to drive him into the Industrial Age so I might get his free Nationalism. On Otto's tombstone it now says, "Achtung, Nationalism is worth more than Steam Power and Medicine plus my worthless life combined."

 

[Catt]

Just to make sure I wasn't completely off-base, I spent 10 minutes and set up a very small test. Created a scenario - small island of all bonus grassland, 2 civs. Rome (human) has all ancient and middle age techs, plus Steam Power. Egypt (AI) has all ancient ancient and middle age techs, plus Nationalism. Each civ starts in a democracy, with 100,000 gold, one settler and one worker -- Rome also has an explorer. Identical terrain. After loading the scenario and starting the game on Regent level, Rome's explorer moves three or four squares and contacts Egypts settler, opening the diplomacy screen. Neither civ has even built a city yet. Cleopatra is Cautious.

An offer of Steam Power for Nationalism, my foreign advisor warns me, would be insulting. Only an offer of Steam Power and 2824 gold (lump sum) will pry Nationalism from Cleo's clutches.

???

 

[gonzo_for_civ]

I am not sure if it has been mentioned yet, but remember that they will pay you less for a tech they are currently researching just as you can buy a tech you are one turn away from getting for about 1 gold usually

 

[cracker]

We should also not forget the world size factor when we are performing these tests. In V1.29 the factors are 1.6, 2.0, 2.4, 3.2, and 4.0 for the five typical map sizes. A tech that is marked at a 2 factor will cost the human player 31-32 gold on a tiny map on regent. The same tech will cost 40 gold on a small map and 48 gold on a standard map.

Catt, your simple test is a brilliant tool to look at some of these things. It eliminates any potential relative power assessment variables since both trade partners have no territory, no citizens, and no military. Brilliant.

I set up Egypt and Hiawatha next to each other and gave myself the Upper half of the ancient tech tree will I gave the Iroqs the lower half. Both civs have Alphabet and Writing, 100,000 gold in the treasury and are still in despotism with having researched the advanced government that they could reach.

When I contact Hiawatha with my artificial explorer I can test a simple set of trades on Regent on a Tiny map to find:

Hiawatha will trade me:

Pottery(2) for 46 gold
The Wheel (4) for 116 gold
Warrior Code (3) for 82 gold
Ceremonial Burial (2) for 46 gold

and he will give me gold in exchange for:

Bronze Working(3) at 92 gold
Masonry(4) at 97 gold
Philosophy(6) at 128 gold
Literature(10) at 144 gold *
Code of Laws(10) at 212 gold

The numbers in the parathesis are the designed cost multipliers.

Notice how severely and artificially the AI undervalues Literature which should be at least equivalent to Code of Laws. This is a significant variance from expectation. This is some goofy impact of the optional tech flag because even though the Literature tech may be optional to advance to the next era, it would be hard to consider it optional for any game strategy that did not focus on extremely early conquest.

Also notice how the AI has artificially inflated the value of The Wheel and Warrior code based on some other factors other than the designed in cost factor of the tech. If you compare the value of the wheel which is a 4 factor to 2x the value of either of the 2 factor techs you can see that it should be valued at 92 gold if there were not other factors being considered. Instead the value is inflated by 24 gold which is roughly 25%.

For the techs that are not distorted by some hidden factors, the AI seems to offer payment of 137% of what the tech can be researched for; while it demands payment of 150% of the cost you would have to spend to research the tech yourself.

This logic still intrigues me a bit. The programming currently says that you will fully recover you entire research cost for any tech plus make a 35-50% profit if you will sell the tech right away to some other civ. You don't even have to sell the tech to multiple civs to make a profit, you just have to be lucky enough to find one of the Marx brothers that has enough cash.

 

[mydisease]

And a slight continuation on the same theme...

I made the same scenario and can get 144 gold for literature.
I then modified literature so that it was no longer optional. Start again and now I can get 218 gold for literature. So removing the optional tag adds 50% to the value of literature!

 

[Nethog]

I would like to throw in a couple of theories regarding how the AI valuates techs for trading ...

Originally posted by gonzo_for_civ
I am not sure if it has been mentioned yet, but remember that they will pay you less for a tech they are currently researching just as you can buy a tech you are one turn away from getting for about 1 gold usually

Expanding gonzo_for_civ's line of thought, perhaps one of the variables the AI uses for valuating a tech for trading to you is the number of turns it would take you to research the tech yourself (perhaps with Science slider set at 100%?), including all prerequisite techs. Just as in real life, time is money, so it makes sense that the AI should inflate the price of a tech that would take you 10 turns to research yourself, if you could receive it immediately.

Originally posted by billindenver
Is it possible it has anything to do with how many other techs have this tech as a prerequisite? I mean, with Bronze Working, you can only get to Iron Working. With Warrior Code, you can get Horseback Riding and Monarchy

To me it would make sense that another factor that the AI should use for valuating techs for trading to you should be the TOTAL cost for you to research the tech yourself (i.e. the sum of all of the prequisite techs and the tech being traded).

Can anyone share any experiences or test results that support or refute these two possible influences on AI tech valuation for trading techs?

 

[Grey Fox]

Yeah, the AI values the tech somewhat based on what you get for the tech. Resource etc.


And don't forget. THE number of civs that has the tech, effects the price much. And the number that can buy the tech of the Ai when you are selling, increase your possible earnings, as they will bid more when there are competion on who will get the tech first.

EDIT: take a look at my program, the TechCalc btw. you can find it in my sig

 

[Nethog]

I would like to emphasize that my question lies along the lines of which variables the AI uses to valuate a tech for trading to you - not the variables that influence actual tech cost which appear to be well understood (see http://forums.civfanatics.com/showthread.php?s=&threadid=29485).

In this thread I see the following proposed variables:
- optional techs
- techs "valued" by a Civ
- culture
- 1st to discover a tech (NOT the tech cost deflation ratio in the tech cost formula)
- reource influences
- tech type, examples: government, Nationalism
- Human player or AI research cost set

So I have added to this list of variables:
- total tech cost required for human player to research tech.
- number of turns human player would require to research tech themselves.

 

[Cartouche Bee]

Nethog, don't forget AI attitude. You will get better deals from Polite civs than Annoyed civs, the margin might be slim but if you make an equation it will fine tune it in to real results. Establishing embassies does help with better trade deals because usually the AI civ attitude improves unless you've been really bad.

 

[Nethog]

Good point Cartouche Bee ... so the total list so far is:
- optional techs
- techs "valued" by a Civ
- culture
- 1st to discover a tech (NOT the tech cost deflation ratio in the tech cost formula)
- reource influences
- tech type, examples: government, Nationalism
- Human player or AI research cost set
- AI attitude

And my two hypothesis are:
- total tech cost required for human player to research tech.
- number of turns human player would require to research tech themselves.

 

[Dark Sheer]

Based on my latest PTW game AI definite put a much lower value in optional tech. In fact, the priority for optional tech is so low that its becoming ridiculous.

I played with 24 Civs and as me (Arabia) and America both already in Industrial era, the rest of the world are still 50% in ancient and 50% in medival. The weird thing is a number of those Civs in medival has Musketman but NO LITERATURE!

I was getting a bit sick with that and start selling literature to most of them at whatever they can afford. In the process, I find out that some of them is willing to pay much much more for a world map than Literature.

 

[Catt]

Seeing as how this thread came alive again . . . alexman did some real leg work and posted his results over at Apolyton. I am copying the "tech value metrics" that alexman discovered here, but feel free to check out the actual work (and discussion) at Apolyton in the Strategy Fourm in a thread entitled "Ideas for a Killer AI." DaveMcW also recently quoted this same text in the CFC Strategy Forum.

It should be noted that the metrics and values below apply only to how the AI chooses its next research project and not to trading values (and includes some guessing to that point as well) -- indeed even Soren chimed in to point out that the AI values tech researching and tech trading distinctly and that alexman's yeoman's work might not have grasped it all. I post it here however, because I don't think it unreasonable that some (if not all) of the factors listed below as influencing tech research paths might also influence tech pricing when trading.

Orignally posted by alexman at Apolyton

The AI places a value on each technology it can research next. The values depend on the various things allowed by each technology, as well as on the turns needed to complete the research. Here are the values:

SS Component: 261/turns
Government: 259/turns
Conscription: 259/turns
Mobilization: 259/turns
Defender (no resource): 198
Attacker (no resource): 134
Defender (resource reqd): 70
Naval Transport: 34
Resource: 16
Double Worker Speed: 8
Allows Diplomats: 8
MPPs: 8
ROPs: 8
Alliances: 8
Embargoes: 8
Trade Over Ocean: 8
Wonder: 6
Attacker (resource reqd): 6
Irrigation: 4
No Disease: 4
Trade Over Sea: 4
Bridges: 2
Double Wealth: 2
Map Trades: 2
Communications Trade: 2
Small Wonder: 2
Specialist: 2
Unit (no A/D/NT): 2
Improvement: 2
Recycling: 1
Prec. Bombing: 1
Worker Job: 1
Empty Tech Cost: 256/turns
Optional: divide by 1.5

These values are cumulative. For example, Iron Working allows swordsmen and iron, so it has a value of 22 more than if it were just an empty tech. Furthermore, for the Romans who build Legionaries (attacker and defender), Iron Working is worth an additional 70 points!

I made a big assumption to get these results. I assumed that the Science advisor suggests that the human player researches the technology that the AI would research under the same circumstances. There was no random factor whatsoever involved in the advisor, but it's possible that the AI players have such a factor.

 

[Qitai]

Here is the rough estimate to the cost of trading.

If buyer has only one source of a tech (considering contact), the tech will cost ~ 2x reserach cost.

If buyer has more than one source of a tech (considering contact), the tech will cost ~ cost of research. This will decrese almost linearly with each additional source of tech.

A source of tech is the number of civ on it's contact list with the tech concern.

If the cost defers greatly from above, civ concern is probably researching the very tech you are selling.

Only exceptions are optional tech where the wonders has been build.

The above is what really affects trade cost. Other factors are minor compared to the above, though I did not completely figure them out.

 

[kb2tvl]

hi guys, just wanted to throw out that a GA when your at peace is better than hitting a GA at war so that said.

Hanging Gardens
Cost: 300
Culture: 4
Requires: Monarchy
Expire: Steam Power
Industrious
Makes 3 unhappy citizens content in its city and one unhappy citizen content in your other cities.

 

[TheNiceOne]

Originally posted by Nethog
I would like to emphasize that my question lies along the lines of which variables the AI uses to valuate a tech for trading to you - not the variables that influence actual tech cost which appear to be well understood (see http://forums.civfanatics.com/showthread.php?s=&threadid=29485).

For an additional variable that impacts the trade cost between AI and human civs, look at the AI - human trade revisited thread.

In short, this thread shows that a combination of AI aggression level and difficulty level affect the cost by up to 20%.

 

[Tassadar]

Excellent work The nice one and cracker, i was a little bit confuse about this trading value. but no more now.

I think it is more complicated than necessary but who i am...

 

[Catt]

Originally posted by Qitai
Here is the rough estimate to the cost of trading.

If buyer has only one source of a tech (considering contact), the tech will cost ~ 2x reserach cost.

If buyer has more than one source of a tech (considering contact), the tech will cost ~ cost of research. This will decrese almost linearly with each additional source of tech.

A source of tech is the number of civ on it's contact list with the tech concern.

If the cost defers greatly from above, civ concern is probably researching the very tech you are selling.

Only exceptions are optional tech where the wonders has been build.

The above is what really affects trade cost. Other factors are minor compared to the above, though I did not completely figure them out.

Qitai - I think there are other factors at play, not all of them minor. For instance, many have pointed out that the trade value of Nationalism far outwieghs its expected trade value based on research cost and dispersal of the knowledge. See my post above (Oct 11) re a quick test on just Nationalism vs. Steam Power. While I haven't done any extensive testing, my in-game experience indicates that government techs also have a significant trade premium associated with them even absent a wonder window.

 

[Beamup]

But that is precisely in line with your post above quoting alexman's numbers for research priorities! Nationalism gives conscription, mobilization, AND a no-resource defender. Thus a HUGE value. And government techs again get a large bonus to their value. I'd suspect that the value for trading purposes uses many of these values the same way. Value would increase with research time instead of decrease, and then there would be the various other factors like friendliness, how many civs know it, etc. These, I'd guess, would be simple multiplicative factors applied to that base value.

 

[TheNiceOne]

I think you're right Beamup and Catt. Research cost is only one factor of the base trade cost of a tech, where things like military units, governments, resources etc. also adds up.

And then when all this is added up, the base trade cost is multiplied by a few factors, where the number of other civs knowing the tech is most important.

 

[Beamup]

This COULD be tested with the editor, but it would be somewhat time-consuming. The requirements would be several special scenarios like the one Cracker used above and a good bit of editing of the tech tree. ALL tests should be done on the second turn (immediately after founding of capitals) to keep other variables from coming into play. The testing program I would envision would go like this:

1. Make several techs, all of the same cost, that provide different things from the list above - one each. Give yourself these techs and see what the other civ is willing to pay, and then the other way around. Make sure the culture ratings are the same at the time of the test, as well as military power (via rules mods not giving the AI any starting units). Check the RATIOS of these costs with the ratios predicted by the values from alexman. If they match, it is confirmed that these values provide the base cost.

2. Now make 3-4 techs that are identical except for having different costs. Again check the prices on the special map. The price will likely be directly proportional to the cost.

3. Now do one particular tech on maps with 3,4,5 civs, only one of which does not have the tech. This will allow measurement of the the "number of civs" factor - one would have to be careful and make sure all civs have contact but no trading has already taken place. Perhaps the human could have contact with everyone and make gifts of contact to the civs NOT to be traded with (to avoid changing that civ's attitude).

4. Then it would be possible to try to investigate such things as culture, attitude, and military strength. These are probably subdominant factors, however.

One caveat - if these various factors are NOT multiplicative, this would break down. This is because, at each step, one would be taking the ratios of different costs and comparing to the prediction. This works if all the other factors cancel out in the process of taking the ratio.

I would be quite willing to do most of the data analysis for this investigation, but unfortunately I'm not in a position to gather the actual data.

At the end of the day, I would envision having a formula that would look something like
price = (sum of values from alexman)*(cost to research)*(number of civs factor [1/n?])*(overall normalization constant depending on many things)
Fixing that overall normalization would be somewhat difficult since we really don't even know exactly what it depends on.