Latest Posts

Topic: Suggestion: starting ware multiplier

DragonAtma
Avatar
Topic Opener
Joined: 2014-09-14, 01:54
Posts: 351
Ranking
Tribe Member
Posted at: 2021-06-14, 20:36

I've noticed that the AI sometimes has trouble handling its wares long enough to reliably make soldiers; this is especially true on unusual maps. Therefore, I recommend giving every player a starting wares multiplier, from 50% to 1000%. If everyone starts with double the standard amounts, the AI (and newer players!) are less likely to run out before they get mines up and running. And as a bonus, it lets us add handicaps; if you're a better player than your friend, then you can start with 100% and your friend with 120%.

Speaking of handicaps, you could also add long-term handicaps with a similar method. If you give someone a 120% handicap, then a building that normally produces one ware will instead have a 20% chance of producing two wares.


Top Quote
the-x
Avatar
Joined: 2019-01-19, 13:23
Posts: 967
Ranking
One Elder of Players
Posted at: 2021-06-14, 21:03

You can find the Addon "Royal" by the-x which does exactly that, except that it adds only 50% but the problem is more in the AI --> AI works efficiently only on large and huge maps, not the usual 1v1 maps, maybe we can correct that by letting the AI build trainingssites earlier?


Top Quote
hessenfarmer
Avatar
Joined: 2014-12-11, 23:16
Posts: 2646
Ranking
One Elder of Players
Location: Bavaria
Posted at: 2021-06-14, 21:10

DragonAtma wrote:

I've noticed that the AI sometimes has trouble handling its wares long enough to reliably make soldiers; this is especially true on unusual maps. Therefore, I recommend giving every player a starting wares multiplier, from 50% to 1000%. If everyone starts with double the standard amounts, the AI (and newer players!) are less likely to run out before they get mines up and running. And as a bonus, it lets us add handicaps; if you're a better player than your friend, then you can start with 100% and your friend with 120%.

Speaking of handicaps, you could also add long-term handicaps with a similar method. If you give someone a 120% handicap, then a building that normally produces one ware will instead have a 20% chance of producing two wares.

Applying such factors on the normal production will be difficult. For starting conditions there are some Options available now, including "trading Outpost" for weaker players which gives them some extra wares from time to time or the "royal" start condition in the addon the-x mentioned.
handicaped Sc are available as well by addon and in standrd game.


Top Quote
hessenfarmer
Avatar
Joined: 2014-12-11, 23:16
Posts: 2646
Ranking
One Elder of Players
Location: Bavaria
Posted at: 2021-06-14, 21:19

the-x wrote:

You can find the Addon "Royal" by the-x which does exactly that, except that it adds only 50% but the problem is more in the AI --> AI works efficiently only on large and huge maps, not the usual 1v1 maps, maybe we can correct that by letting the AI build trainingssites earlier?

In fact it builds them quite early, long before it can support them so it would be better perhaps to delay them. Anyhow the underlying problem is that we only have one AI which should be working with all tribes and all maps more the less. However the current implementation is only reactive but not proactive. Which makes it slower by definition as a human player.


Top Quote
the-x
Avatar
Joined: 2019-01-19, 13:23
Posts: 967
Ranking
One Elder of Players
Posted at: 2021-06-14, 21:23

Yes, the AI works very nice on large maps, if you give the AI enough time, it solves many problems. Maybe we can improve it in short games, competitive games or smaller maps? Yes, probably only one AI makes sence, and we cannot use the "weak" AI buttons for different timings in AI play? Like late and strong, just simple upgrades or heroes but fast?


Top Quote
hessenfarmer
Avatar
Joined: 2014-12-11, 23:16
Posts: 2646
Ranking
One Elder of Players
Location: Bavaria
Posted at: 2021-06-14, 21:36

in the past years the AI improved a lot in my opinion. So we already had a complaint about it being to hard for beginners so finally we might reach a stste where the weak buttons make sense.
Making it good enough for an experienced player on small early contact maps will probably never happen with its current design. And there are only a few developers understanding this design. So changing this is quite improbable.

So any suggestion of "make the AI doing this or that" is more the less useless as this is not how the AI works or will ever work.

On the opposite reporting any specific and reproducible misbehaviour like "AI never does this, or never builds more then, or always builds to much of this" are very welcome to identify code bugs in the over 7500 lines of code (a good couple of them was fixed in every recent release) .


Top Quote
the-x
Avatar
Joined: 2019-01-19, 13:23
Posts: 967
Ranking
One Elder of Players
Posted at: 2021-06-14, 21:40

Yes, the AI definitely is a masterpiece ;- ) still i wonder, in easy words why cant we focus the AI on the output, like maximize the number of upgrades in a trainingssite? spiecific AI does not this etc mistakes are very rare and there we are on a very high level, rather many games are played when a player says lets play a small map, i have only time an hour and therefore maybe we can focus the AI more, also i am very happy when i would met the AI and it already has an upgrade whilst still i dont have one.


Top Quote
hessenfarmer
Avatar
Joined: 2014-12-11, 23:16
Posts: 2646
Ranking
One Elder of Players
Location: Bavaria
Posted at: 2021-06-14, 22:05

the-x wrote:

Yes, the AI definitely is a masterpiece ;- )

I have never said this. so no need to be sarcastic. I know the Ai has its shortfalls. However knowing about this and solving it are different shoes.

still i wonder, in easy words why cant we focus the AI on the output, like maximize the number of upgrades in a trainingssite?

If you can point me to the part of the code you would improve for that and how it should be improved I'd consider doing the improvement. However with my small knowledge of the AI Code I am not able to find a solution for this. So please help me out.

spiecific AI does not this etc mistakes are very rare and there we are on a very high level,

so at least our efforts over the past years was not pointless, thanks

rather many games are played when a player says lets play a small map, i have only time an hour and therefore maybe we can focus the AI more, also i am very happy when i would met the AI and it already has an upgrade whilst still i dont have one.

As said making the AI competitive on small maps for experienced players is nearly never going to happen in my opinion. So better start a long game and use the savegame option whilst playing against the AI.


Top Quote
the-x
Avatar
Joined: 2019-01-19, 13:23
Posts: 967
Ranking
One Elder of Players
Posted at: 2021-06-14, 22:10

I like the AI a lot - its not sarcastic. I am doing the code at the moment but also prohibited till doesnt help, i want the AI to (only if possible) after a stable wood etc production to focus already on upgrades in the trainingssite. We can also help it out by giving some trained soldiers, to make it competitive?


Top Quote
hessenfarmer
Avatar
Joined: 2014-12-11, 23:16
Posts: 2646
Ranking
One Elder of Players
Location: Bavaria
Posted at: 2021-06-14, 22:29

With code I meant the approx: 9000 lines of defaultai.cc dafaultai_warfare.cc and defaultai_ seafaring.cc c++ source code.
not the small leftovers of AI hints in the Lua files.
There are only small possibilties to influece the AI with Lua code.


Top Quote