Question about difficulty

[KungCheops]

Does the AI play at it's best on prince, or is it better on higher difficulties?

I'm just wondering because in the new patch notes it says that the AI is going to use GG's in offense more often in the early game. Is that just because of their bonuses or are they programmed to play better on higher difficulties?

Personally I always play on prince (or which ever one is the equal one) because I like playing with fair odds (even if I win every time).

 

[Strategist83]

The patch notes are simply addressing the fact that the AI has been reprogrammed to make better use of GGs. In other words, the AI is upgraded in the new patch - that will affect all difficulty levels.

The AI doesn't play better on higher difficulties, it just gets larger bonuses.

 

[Trias]

The AI doesn't play better on higher difficulties, it just gets larger bonuses.

If they truly implemented everything they talked about prelaunch, then the AI should play slightly better as difficulty is increased.

The way that the AI basically works is that it rates each different possible move with a value, and then picks a random move with the rating acting as a weight. At higher, difficulty the AI supposed the weigh the rating heavier making it more likely to pick the move that it rated the highest.

In theory, that should make the AI play better at higher levels, while avoiding that it always makes the same choice for options that are almost equally good. In practice, the algorithms that determine the rating for each move, are not (yet) so good and therefore do not always effectively distinguish between good moves and bad moves.

Also, it is not clear if any of this was implemented in the final AI routines. Hopefully, the AI code will be released in the future so we may find out.

 

[Txurce]

Personally I always play on prince (or which ever one is the equal one) because I like playing with fair odds (even if I win every time).

In addition to what Trias wrote - that there is some reason to believe the AI makes better choices at higher levels - the AI has a built-in happiness bonus at every level. So Prince is actually not a level playing field. Buckets has a mod that removes this bonus. It is also balanced in the WWGD mod.

 

[nokmirt]

Does the AI play at it's best on prince, or is it better on higher difficulties?

I'm just wondering because in the new patch notes it says that the AI is going to use GG's in offense more often in the early game. Is that just because of their bonuses or are they programmed to play better on higher difficulties?

Personally I always play on prince (or which ever one is the equal one) because I like playing with fair odds (even if I win every time).

The AI is the same for all levels of difficulty. The base or nuetral level is Prince for the AI, from what I learned on another thread. The AI just gets bonuses on higher levels of play.

Myself, I have been playing King and the AI keeps just beating me to victory, so right now it is challenging. However, I like to go back here and there and play a game on Prince. I am not a good loser.

 

[Strategist83]

The AI does NOT make better choices at higher levels. Anything above Prince uses the same AI. Or, rather, let me elaborate on this: It is possible there is some subtle difference from Prince and up, but if there is, it is so insignificant as to be non-existent. Thus, regardless of what the code specifically says, you may as well conclude that there is no difference (though I believe the AI is noticably handicapped on lower levels, I wouldn't know since I've never played anything below Prince). The difference and the reason why you might think it plays better is things like starting out with lots of units on high levels. The AI has tremendous problems coping with barbarians (that's why you find so many workers/settlers captured in camps) but on these very high levels it starts out with so many units it manages to defeat them regardless. This is what causes the real difference: The Prince AI loses all its settlers/workers and so falls hopelessly behind the player. Immortal/Deity AI have so many units they recapture their settlers, bad AI or not. But there are absolutely no real differences in the AI's ability to 'think' - it's all about what starting conditions it has and how large the bonuses are.

Arguably, there is no such thing as a 'level' or 'fair' AI. You get to choose from either an AI that anybody but extremely poor players will utterly steamroll without even trying or an AI that still isn't challenging but uses its intense bonuses to throw a literal wall of units at you. The only meaningful way to play the game is either multiplayer or playing against yourself, such as making special rules for what you can and cannot do (roleplaying) and seeing if you can outperform your own records.

 

[Txurce]

The AI does NOT make better choices at higher levels. Anything above Prince uses the same AI.

Even after editing, that's still a very emphatic statement. Could you back it up, or is it personal opinion?

For all I know you're right, although that's not my interpretation of what the devs when the game came out, as Trias basically explained.

 

[nokmirt]

The only meaningful way to play the game is either multiplayer or playing against yourself, such as making special rules for what you can and cannot do (roleplaying) and seeing if you can outperform your own records.

You could duel yourself with hotseat. I wonder though, I know I am not the best player, perhaps playing more multiplayer will help make me better. If you can learn to beat other humans the AI should be a cinch. At least thats my theory.

I play at King level, but I still fall behind in tech early and then after its a bit too late to win, I push science as much as possible. I was thinking of praticing by playing with just one victory condition per game, so I can master each one. When I beat one, I can go on to the next and so on.

 

[aatami]

The way that the AI basically works is that it rates each different possible move with a value, and then picks a random move with the rating acting as a weight. At higher, difficulty the AI supposed the weigh the rating heavier making it more likely to pick the move that it rated the highest.

In theory, that should make the AI play better at higher levels, while avoiding that it always makes the same choice for options that are almost equally good. In practice, the algorithms that determine the rating for each move, are not (yet) so good and therefore do not always effectively distinguish between good moves and bad moves.

Oh, did not know that, thank you for the info.