Again, you're not really understanding my point. I am not asking for the AIs to play worse. I'm asking for the AIs to PLAY BETTER. From my last, playing as Austria is not my concern; it's a general example. I'm asking you to provide feedback to the developers, not "fix" anything. There are a few shortcomings that could be addressed in a future effort. Whether or not it is addressed isn't the point. I work in IT and have worked with software developers on many occasions. They need feedback. No software effort is one and done.
Specific feedback items are:
1. Bad Endgame Strategy:
1a. Each country needs to prioritize capturing their 18 centers. The marginal centers may vary, but there are usually a dozen or so specific centers that a specific country cannot do without, varying with each country. In the endgame, countries need to prioritize attacking these centers not already in their possession. (Ex. A game with a very strong Anglo-German alliance, France eliminated, Russia eliminated, Austria eliminated, and a moderately sized Turkey and decent sized Italy: An Anglo-German alliance makes sense in the beginning, maybe even the middle, but England cannot win a solo with an equally strong Germany in the way. England needs to prioritize attacking the German centers instead of allying with Germany to attack Turkey when both England and Germany have more centers each than Turkey. This is especially true when Italy is doing a decent job keeping Turkey contained on its own just from being in the way in the Med.)
1b. Likewise, each country needs to be programmed to identify a country rapidly expanding and prioritize stopping the expansion. (Using the same example above, Germany needs to be aware that the English stab is coming, in pursuit of it's solo effort, and have forces in reserve to address that stab, instead of keeping Turkey contained. Likewise, England needs to see an equally strong Germany as a potential threat, since Germany will require some centers that England already has for its solo effort, in Scandinavia for instance.)
1c. Going with the two above, the AIs need to know about STALEMATE LINES. Once three or four countries are eliminated, holding or breaking through the stalemate lines should be everyone's top priority. Also if an AI has a good stalemate line held with distracted enemies in front and an "Ally" behind this line in the rear, the AI should stab the country behind the stalemate line to either conquer or contain them, which is related to the two points above.
(Again with the previous example, maybe Italy and Turkey have been bitter enemies all game, but now they must come together and support each other to prevent the English solo.)
2. Home centers should be a priority in the beginning:
A lot of AIs get completely destroyed relatively early. They seem to not care about Country A taking home centers in the back because they are obsessed with attacking Country B in the front. A human player would do their best to patch relations in the front and swing to the back to retake their home center. AIs don't or can't change their strategy like this.
3. Go beyond gunboat:
Allow a human player to explicitly message the AIs with a move request, and have the AIs consider it. For example, if an AI can help a person take a center, but there's a 50%/50% proposition as to which is the attacking unit, the message can settle it. Likewise, let the AI send a message to the human player requesting the same assistance with a move request too. If the support goes well, "good-will" is established. If it does not go well, "ill-will" is established.
The above are a few notes that I've come to after playing with the AIs a lot. There could be more specific feedback to be discovered later. Now, regarding my previous posts, here are some notes that could be useful feedback to the developers as well.
http://webdiplomacy.net/board.php?gameI ... ocache=635
In S12, both England and Germany decide to attack Turkey, who never did anything to either of them and was trying to get in position to help hold them to prevent a French solo. They didn't even hold each other. Instead, they help France solo, despite the fact that France conquered them both. What gunboat message did France send to get such cooperation? Indeed, in S13 England does it again, and in F13, England just gives up and supports the French solo.
http://webdiplomacy.net/board.php?gameI ... rand=49028
In A05, Italy decides to help Austria destroy Turkey despite it being the first gunboat interaction between the two of them. A06, Italy attacks for Smyrna, helping Austria. S07, Italy attacks for Smyrna again, but Austria gets Smyrna instead. A07, Italy attacks Austria in Smyrna. Repeat, Italy ATTACKS Austria! S08, Italy leaves. Two things going on here. First, Italy is being very stupid. His actions allow Austria to de facto win the solo should the game continue, if it had been a real game. Second, despite Italy attacking Austria, Austria elects to NOT take Venice with Trieste in S08. So you can tell me, "Input orders of support, gain a friend. Input orders of attack, gain an enemy," but that did not happen here, and no other variables were in play.
http://webdiplomacy.net/board.php?gameI ... ocache=962
Russia and Austria attack for Galicia in S01. Russia supports for Galicia in F01. Austria does the same in S02. Austria takes Warsaw in F02. Russia takes Warsaw back in S03. Clearly, Austria and Russia are not friends. All the while, Italy is battling for Austria's home centers the whole time. But, Russia decides to commit suicide and help Austria by screwing over Turkey in A05. So despite open aggression with Russia and Italy the entire game, Austria decided to stab Turkey, and Russia recognizes that this specific stab is coming? Then again in A06, Germany supports England, who supports Austria to help attack Turkey, which allows Italy to attack Turkey too. Germany and Turkey are usually considered natural allies in the game at this point, but the AI doesn't recognize this. In fact, only France was left out of the party of fighting Turkey, which is funny because Turkey is the easiest country to contain.
Again, I'm just asking you to forward feedback to the developers. Let them decide what to do with it. Even if it just goes into a folder with the code and they do nothing with it, maybe future developers will.