Single Player bugs and crashes v38 plus (SVN) - After the 20th of February 2018
- Thread starter Dancing Hoskuld
- Start date
Noriad2
Emperor
- Joined
- Oct 23, 2014
- Messages
- 1,153
I usually play with Start as Minors on (because it is historical) but over the years I occasionally played with off, and didn't see much difference.
Just curious, has the claim that SaM hurts the AI ever been properly tested? As in let the game autorun from the same starting position, with SaM on and off? And AI scores compared afterwards?
Thunderbrd
C2C War Dog
No. It's based mostly on some behavioral observation factors. By all means, run a test to confirm.
Why easier than noble handicaps have iconstruct/iresearch/itrain percent changed (40/60/80 for settler/chieftain/warlord) ?
Instead of that you could use different steps for these handicaps.
Is that for no AI multiplayer? For quick testing?
Spoiler :
While ladder up to immortal is fine, Deity/Nightmare still are bit nerfed when compared to old system.
AI is bit slower than used to be in old system with buildings/projects/techs on Deity/Nightrmare but it is compensated with units.
Deity AI speed for projects is reduced by 20% and for buildings/techs by 10%.
Nightmare AI speed for projects is reduced by 20% and for techs by 30% when compared to old system.
Basically it seems like there was not single but double step up in old system for projects/techs (for Immortal->Deity and Deity->Nightmare).
Last edited:
It is meant for multiplayer and the flexible difficulty mechanic. It offers a significant handicap differences between inexperienced and experienced human players as well as being an interesting buff for the weakest AI players when using flexible difficulty. Why not treat the lower than Noble difficulties differently?
You are only looking at three factors within the handicap file, if you compare all the changes you would see that Deity and Nightmare are more difficult now (rev. 10052) than what they were before I started changing difficulties back in rev. 9976.
Why are you so obsessed with not making any changes to the mod, should I, in your opinion, revert all the changes I made and focus only on replacing the iXPercent tags with completly equivalent iAIXPercent tag values?
Last edited:
Now that you explained, why handicaps below noble are treated differently, and that AI doesn't need that much research boost on Deity/Nightmare anymore its fine now.
I wonder how @Noriad2 will like changes - he complained, that its easy to out tech AI even on Nightmare.
I agree with Noriad2. It doesn't matter if the AI can build and train faster if it's consistently behind in the Tech Race, even on Nightmare, with the new system.
And I gain would like to point out that multi-player games still offer no significant benefit for lower difficulty players, again because of the Tech Race being the biggest contributing factor, and the build and train speeds are also equal between human players, only the costs of maintaining are different, and that's irrelevant as one doesn't need hordes of units for a good strategy. IF amount of units played a larger role then maybe, but it doesn't, and I don't see how it can be set to do either.
My ongoing test with no changes to system at all on SVN 10051 I am getting nearer to Tribalism and am a clear tech leader. Closest rival is 6 turns from Atlatl Making when I am 8 turns from Tribalism. I am playing with TD and WFL BUT inactive for me, the player, only AI get those benefits.
I do have SaM on but planning on doing a comparative test with it off after checking if Unit Supply still gets skewered.
And I gain would like to point out that multi-player games still offer no significant benefit for lower difficulty players, again because of the Tech Race being the biggest contributing factor, and the build and train speeds are also equal between human players, only the costs of maintaining are different, and that's irrelevant as one doesn't need hordes of units for a good strategy. IF amount of units played a larger role then maybe, but it doesn't, and I don't see how it can be set to do either.
My ongoing test with no changes to system at all on SVN 10051 I am getting nearer to Tribalism and am a clear tech leader. Closest rival is 6 turns from Atlatl Making when I am 8 turns from Tribalism. I am playing with TD and WFL BUT inactive for me, the player, only AI get those benefits.
I do have SaM on but planning on doing a comparative test with it off after checking if Unit Supply still gets skewered.
You can try new game without SaM on SVN 10052.
I also want to check if SaM does hinder the AI that much. The only thing I can think of is the player being able to take out their hunters and scouts and trackers, which should not make that much difference IF they can spit out units at a much higher rate, as has been claimed.
But, yes, just saw that the levels have been changed again so should do a new game with that, but after I've checked if the Unit Supply problem still persists.
But, yes, just saw that the levels have been changed again so should do a new game with that, but after I've checked if the Unit Supply problem still persists.
JosEPh_II
TBS WarLord
I can't agree with your supposition and position on this Toffer.
JosEPh_II
TBS WarLord
Does starting with SaM give you the player a Clubman still to start the game with? Big advantage player if it still does. (Was supposed to have been changed, was it?)
We have already "proven" that with the AI focused on war from the start, slows down their empire development. Contrary to T-brd's assertions that it's only been, "No. It's based mostly on some behavioral observation factors.".
SaM was made for having early warfare to satisfy the war mongers' thirst for battles as early as possible. The AI is focused on building units and only those buildings that aid in this effort. Has this been denied? Has this been proven to be wrong? I do not think so. By more than just this Modding team. But from the AND team as well as other Mod makers. I did work with Afforess for awhile on AND and with 45* too. Both easily match the talent that is on This team.
You're right, "Much more" was an overstatement, it's about the same difficulty on deity now as it was before. The focus has been shifted a bit, economical penalties for the human player and faster unit training for the AI is now compensating for a somewhat slower tech progress for the AI.
Thunderbrd
C2C War Dog
I am only speaking from my own knowledge on SaM. I have done nothing to prove anything, only have suspicions of inneffective behavior there. If you know its been proven more conclusively, cool.
I updated to SVN 10055.
AI is lazy to build cities: on 100th turn of Blitz/Noble/Gigantic everyone should have two cities at least.
Game crashed somewhere after 100th turn.
Added autosave of last turn when game crashed.
By the way some AIs are wasting time on processes.
This sometimes appears in Xml_MissingTypes:
[5867.494] info type 'ATTACHABLE_CHIMNEYSMOKE' not found, Current XML file is: modules\Natural_Wonders\Reef_Shark_CIV4SpawnInfos.xml
AI situation is much better on 200th turn.
This time China. Babylon and Egypt are having a lot of fun.
AI is lazy to build cities: on 100th turn of Blitz/Noble/Gigantic everyone should have two cities at least.
Spoiler :
Game crashed somewhere after 100th turn.
Added autosave of last turn when game crashed.
By the way some AIs are wasting time on processes.
This sometimes appears in Xml_MissingTypes:
[5867.494] info type 'ATTACHABLE_CHIMNEYSMOKE' not found, Current XML file is: modules\Natural_Wonders\Reef_Shark_CIV4SpawnInfos.xml
AI situation is much better on 200th turn.
This time China. Babylon and Egypt are having a lot of fun.
Last edited:
Thunderbrd
C2C War Dog
The crash is very confirming that @alberts2 is on to something with the selection groups not cycling properly. The crash takes place when a function call goes to the entirely wrong function during an initiation of a selection group for the animal team during a new spawn process. If selection groups aren't recycling properly when animals die, then on a Blitz/Gigantic map, it is entirely conceivable that the maximum number of selection groups can easily be exceeded within a hundred turns or so. I suspect that this exceeding of the selection group cap for the player causes a stack overflow that messes with the ram and sends the next function call to the wrong location.
I'm not sure what's wrong with that or how to fix it... hopefully alberts2 is still looking into it for us and will find a solution. Could be a very very VERY longstanding bugbear of an issue for us.
That said, the cool side of this is that if you save and reload now and then it will clear these junk selection groups. So after the crash, the game can be reloaded and continued. Thus it's not the worst bug of all time. Just really annoying, especially if you have a large animal spawn rate in a large space. Some epic wars could cause this pretty quickly too I suppose. Barbs could also trigger the issue.
The rest of this is obviously known issues to improve on.
I have Barbarians on. Earth itself is actually Large sized map.
All current spawn rate modifiers:
Code:
<Define>
<DefineName>NEANDERTHAL_SPAWN_MODIFIER</DefineName>
<iDefineIntVal>50</iDefineIntVal>
</Define>
<Define>
<DefineName>BARBARIAN_SPAWN_MODIFIER</DefineName>
<iDefineIntVal>50</iDefineIntVal>
</Define>
<Define>
<DefineName>ANIMAL_SPAWN_MODIFIER</DefineName>
<iDefineIntVal>100</iDefineIntVal>
</Define>
<Define>
<DefineName>SEA_ANIMAL_SPAWN_MODIFIER</DefineName>
<iDefineIntVal>75</iDefineIntVal>
</Define>
<Define>
<DefineName>BANDIT_SPAWN_MODIFIER</DefineName>
<iDefineIntVal>0</iDefineIntVal>
</Define>
<Define>
<DefineName>KILLERRABBIT_SPAWN_MODIFIER</DefineName>
<iDefineIntVal>0</iDefineIntVal>
</Define>I wonder what are those last two entries.
Crash happens every 50 - 75 turns or so.
I have TD/Tech Trading/Tech Brokering enabled but not WFL.
One of AIs is already in Renaissance on 300th turn.
I guess 4 civs with contact to each other vastly accelerates research rate.
While AI Autoplay doesn't show effects of interaction with human player it is fastest way to test but accuracy may suffer.
Tech leader isn't as advanced as human would be both in farthest Xgrid tech researched and total techs researched.
And they all are on Noble.
I noticed AIs pack cities densely.
Is that result of Barbarian option being on?
If I had WFL then everyone would be in X and X+1 era at worst.
Spoiler :
My current autorun is so realistic when you aren't looking at date: Aztecs, Incas and Aborigines are in Classical era.
Egypt/Greece/Babylon/China are in Renaissance era, maybe someone still in Medieval.
Tech Diffusion will be very strong, when they meet civs stuck in Americas/Australia.
Thanks to aggressive animals, and barbarians option being active Aztecs still didn't met Incas.
Turn time speed report:
18 minutes passed between 300th and 310th turn.
That is 1.8 minutes per turn or 108 seconds per turn.
There are 7 AIs + 5 NPCs and autosaving.
That is it takes around 13.5 secs per AI to process turn.
Last edited:
Dancing Hoskuld
Deity
These are works in progress. Nothing actually exists yet.
- (BANDIT_SPAWN_MODIFIER) Possibly another NPC to cover wild bandits/criminals that have no cities and are a nuisance to everyone. A group that ignore the Great Wall and can be used for slave and serf revolts and other events in all nations. They prey on trade and merchants not units or improvements. They have minimal effect on the crime property but may inspire others to criminality.
- (KILLERRABBIT_SPAWN_MODIFIER) more of an Easter Egg for the Bad Karma options.
Found what's causing this while browsing the source code on my phone. I try to fix this later today or maybe tomorrow.
AI reduces properties even though they are deep into safe region.
I'm running AI autoplay on Blitz/Noble/Gigantic (Earth is Large sized).
Babylon is most successful civ in this test.
You could add global -1 to happiness to each crime and AI wouldn't even notice, as it would try to push down properties.
As for calendar it should be 20th century around 500th turn on blitz, but that will be fixed after release.
I'm running AI autoplay on Blitz/Noble/Gigantic (Earth is Large sized).
Spoiler :
Babylon is most successful civ in this test.
You could add global -1 to happiness to each crime and AI wouldn't even notice, as it would try to push down properties.
As for calendar it should be 20th century around 500th turn on blitz, but that will be fixed after release.
Last edited:
Similar threads
- Sticky
- Sticky
