New Hotfix Version (12-15)

[CrabHelmet]

In Deity, the AI seem to be getting (slightly) too much Science and significantly too much Culture. I had one game as Tradition Morocco with Uluru in my borders and Incense as my Lux and was 2 policies behind by turn 81.

I think the trouble with Policies is that a lot of Policies are self-enhancing - lots of Policies increase your Cultural income which means you then get more Policies. They snowball, and significantly. This is much less true for Science - most Techs don't unlock anything that gives you Science.

Try playing a Deity match with Ethiopia in. It's not winnable.

 

[Infixo]

I added a section on AI handicap bonuses on the wiki a bit ago. These are the details currently.

Great! That's what I was talking about.

Then... If I understand correctly those variables used, the situation is even worse then expected.
1. First, iYieldHandicap is calculated.
2. Then, it is adjusted based on event type (*2 or /2)
3. Then, for some events, cities are excluded (GP, Wonder, Dig, CS, TRs)
4. iYieldHandicap is added to ALL cities as Culture, Food and Production (if cities are included).
5. Civ gets always iYieldHandicap as: Gold, GAP, Culture and Tech.

So...
a) all yields are treated as equal
b) there are cases when Culture is added multiple times (one per city). That's actually the default approach, only specific events are excluded. Could you provide also a full list of events HISTORIC_EVENT_XXX? The code just lists events that do NOT add yields to cities.

 

[ElliotS]

Great! That's what I was talking about.

Then... If I understand correctly those variables used, the situation is even worse then expected.
1. First, iYieldHandicap is calculated.
2. Then, it is adjusted based on event type (*2 or /2)
3. Then, for some events, cities are excluded (GP, Wonder, Dig, CS, TRs)
4. iYieldHandicap is added to ALL cities as Culture, Food and Production (if cities are included).
5. Civ gets always iYieldHandicap as: Gold, GAP, Culture and Tech.

So...
a) all yields are treated as equal
b) there are cases when Culture is added multiple times (one per city). That's actually the default approach, only specific events are excluded. Could you provide also a full list of events HISTORIC_EVENT_XXX? The code just lists events that do NOT add yields to cities.

Historic events are everything that gives humans a historic event. (aka entering era, winning war, great person, building wonder, etc.)

Also @Gazebo can correct me if I'm wrong, but I'm 90% sure the city-culture is only for border expanse. If it's real culture then they're getting MORE culture than science, when they should be getting less.

And yeah the AI gets equal amounts of every yield. I'm not sure if that should change. Thoughts everyone?

In Deity, the AI seem to be getting (slightly) too much Science and significantly too much Culture. I had one game as Tradition Morocco with Uluru in my borders and Incense as my Lux and was 2 policies behind by turn 81.

I think the trouble with Policies is that a lot of Policies are self-enhancing - lots of Policies increase your Cultural income which means you then get more Policies. They snowball, and significantly. This is much less true for Science - most Techs don't unlock anything that gives you Science.

Try playing a Deity match with Ethiopia in. It's not winnable.

Judging by initial reports it's also too easy on easier difficulties. (Please only report if you play the game to completion. It's a curve now, if early game is pre-curve strength it will wreck you later.)

 

[Infixo]

Quick game report. Turn ~250, I entered Industrial and managed to catch up will all Civs except Inca.
I have 43 techs and 14 policies. The AIs that still matter have 40-43 techs and 12-16 policies. Inca has 44 techs, 20 policies (!), 2nd biggest army, an biggest Production and Gold output.
So, I suppose that is not the matter of handicap but a runaway case. I'm gonna check now how late game looks on Emperor, if AIs will start to speed up.

 

[CrabHelmet]

What sort of historic event might fire before turn 10? I've been watching AI behaviour carefully, and often you'll notice AI on Deity with 5 techs and 2 policies by turn 10, when most others have 2 and none, respectively. Clearly something has triggered the handicap, but I can't figure out what?

 

[ElliotS]

What sort of historic event might fire before turn 10? I've been watching AI behaviour carefully, and often you'll notice AI on Deity with 5 techs and 2 policies by turn 10, when most others have 2 and none, respectively. Clearly something has triggered the handicap, but I can't figure out what?

Can I get a screenshot? That sounds like a bug. That would be way more than 1 or even 2 events.

 

[tu_79]

In my current game, after some failed games, I was already the last civ in Emperor, even with a tuned down handicap. This is France. No early bonuses at all. It feels right. No bonuses, no leading.
If you are playing China you should be the big baddass civ in the world, at least in Classical.

So, I propose you compare early game with civs with early power and civs with late power. Be cautious about your lucky starts, though.

 

[notque]

iHandicapA * iEra * iEra + iHandicapB * iEra + iHandicapC

Shouldn't that be

(iHandicapA * iEra) + (iHandicapB * iEra) + (IHandicapC * iEra) ?

I guess I fundamentally misunderstood the bonuses... okay, so they stack. in a weird way.

 

[Questdog]

I destroyed a Barb camp and returned a City State worker to its owner and the worker never moved; just sat there where he was liberated.

 

[Gothic_Empire]

I destroyed a Barb camp and returned a City State worker to its owner and the worker never moved; just sat there where he was liberated.

Poor Bartleby.

 

[notque]

Wrote up a quick thing to show me the yields over time so I can better try and choose some numbers for myself.

https://play.golang.org/p/OLxHSqsgPl

 

[Infixo]

Well, I think I'll give up. Inca has entered Modern era and has just jumped 4 techs and 2 polices ahead, in like ~10-12 turns.

 

[notque]

I am going to try this out and see how it flies. Bonuses heavy early, and lowering over time, since there are so many more gps/arch digs/trade routes etc later.

https://play.golang.org/p/2C9QSOzIwx

 

[vuko]

I am for the idea to disabling historic events on archeology digs. There are just to many sites and the first AI (off course its the runaway) gets just insane instant bonuses for being first in researching archeology.

Why would such "normal" event be historic? The amount of "skill" needed for this action is in my eyes equal to killing a single enemy unit.

 

[Moi Magnus]

iHandicapA * iEra * iEra + iHandicapB * iEra + iHandicapC

Shouldn't that be

(iHandicapA * iEra) + (iHandicapB * iEra) + (IHandicapC * iEra) ?

I guess I fundamentally misunderstood the bonuses... okay, so they stack. in a weird way.

No. Your proposition is kind of weird. There is no difference in your case between A B and C, while in Gabezo's formula, it they have different effects:
+ The bonus C does not scale with era, so is an "early game" bonus.
+ The bonus B scale with era, so is a "mid game" bonus
+ The bonus A scale quadratically with era, so is an "end game" bonus.

 

[Owlbebach]

iHandicapA * iEra * iEra + iHandicapB * iEra + iHandicapC
Shouldn't that be
(iHandicapA * iEra) + (iHandicapB * iEra) + (IHandicapC * iEra) ?
I guess I fundamentally misunderstood the bonuses... okay, so they stack. in a weird way.

Ehh... What you wrote is equal to iEra*(iHandicapA+iHandicapB+iHandicapC), whats the point of different coefficients then?

 

[notque]

No. Your proposition is kind of weird. There is no difference in your case between A B and C, while in Gabezo's formula, it they have different effects:
+ The bonus C does not scale with era, so is an "early game" bonus.
+ The bonus B scale with era, so is a "mid game" bonus
+ The bonus A scale quadratically with era, so is an "end game" bonus.

There's no point going over what I thought, but for some reason I thought the intention was separate bonuses for early middle and late. In the formula, they each contain each other.

It doesn't matter, it was a mistake.

 

[psparky]

As I only have one completed game of VP, I'm not in a position to help with game analysis. However I've been following this work with great interest as I think it could be a very important improvement to the game. Indeed the single biggest attraction of VP to me is that it already greatly improves on the vanilla "Ai starts way ahead, catch me if you can (and then most likely win)" model, making a greater range of strategies possible.

Anyway, I have an observation and a question

        iHandicapA = pHandicapInfo->getAIDifficultyBonusEarly(); 
        iHandicapB = pHandicapInfo->getAIDifficultyBonusMid(); 
        iHandicapC = pHandicapInfo->getAIDifficultyBonusLate(); 
        iYieldHandicap = iHandicapBase * (iHandicapA * iEra * iEra + iHandicapB * iEra + iHandicapC) / 100;

I checked on Github and it appears (to this novice Github user) that this code is still current. If so, then the Early and Late named configuration values are being applied to the wrong term in the equation.

In the first two game eras, the factor that applies is just A+B+C. Once a set of values that work well for the first two eras is established, when A or B is changed, C should be changed by the same amount in the opposite direction so that A+B+C remains constant. Do we have a good value for A+B+C for the early eras yet?

 

[Gokudo01]

the culture runaway is clearly an issue. and because culture is so strong in vp, the AI which leads, snowballs culture which snowball everything else. I mean my last game immortal, AI picked freedom, finished statue of liberty 5 turns later and then was 6 tenet in freedom while most AI and I were still trying to finish the renaissance trees ....

 

[ElliotS]

So far my list of proposed changes to start:

• AI gets 50% culture compared to other yields.
• Digs are not counted for Handicap bonuses (leads too easily to snowballs)
• Trade route distance modifier removed, overall yields decreased by 25%. (Keep 1 per city thing. That's enough for some diversity.)