A new player has entered the arena and went on to smash all records with a very impressive turn 4 capture of Berlin even with a 10% income-bonus for the Axis! Wielded his canopener like a BattleAxe!
And I was fairly confident my own turn 7 achievement would hold out quite well.....
We could use some more competition here. Next months challenge I'll leave up to you to make!
1 Bel Finish: Round 4-7 (British turn) GOTM2_bel.tsvg 2 Zjelco Finish: Round 7-7 GOTM2_Zjelco.tsvg 3 Ice Finish: Round 8-7 GOTM2_Ice.tsvg Though he didn't really try very hard he said... (Included you to make the field seem larger :-)
Whew! That was work! The Russians entered Berlin on turn 31 (no 10% bonus for Axis countries).
Since I am a slow, methodical player, this will give every one an easy target to beat.
It was fun to play. I recently checked in and found out about the Hard AI, a fantastic improvement. I played the Allies in Ultimate Pants and Germany took Moscow. Then I tried Ultimate World: Expanded and lost Moscow to Germany, again. The AI plays land powers quite well.
I think NWO is easier for the Allies to defend Moscow and ultimately win because of the cheap infantry. I think that keeps Russia in the game. Also, the lack of a Japan means every bit of fighting power can face west.
I am not a great fan of Low Luck. It definitely helps the human. Non-Low Luck allows for great victories and defeats as found in real warfare. It also forces players to keep reserves, just in case, thus inhibiting the human.
Indeed lowluck does help the human player. And the AI is not as good at reaping the benefits of Low Luck.
Next months regular challenge is Dice again. Depending on feedback and requests it might stay dice in the future.
If you send me the savegame (TripleA.GOTM@hotmail.com) i can include it in the results, or you can post it here for others to enjoy.
First let me say congrats on a 4 turn finish, that is really impressive. I didn't submit this month, but let me offer some thoughts
NWO isn't a very enjoyable map to play against the AI. This was basically a game of "find the best way to exploit the AI", and I wasn't really enjoying it (I started a game but never finished). Because the AI has little logic for capitals, can openers, and nations that are mixed land-sea powers, I feel like I'm basically just playing against a randomly moving opponent. The AI also had really weird purchase choices, I saw Germany spam elites.
The bonus income has very little effect on this challenge, because with or without the Allies could still outproduce the Axis AI by a substantial margin. I would suggest a larger amount than 10% be used next time.
For NWO, the difference between a well played axis and a poorly played axis is enormous. Unless the AI can execute a decent opening, it is completely different than playing a human vs human game (for a map like Revised, the difference isn't as big). I think a map where the gap is smaller would be a better choice.
- Can openers are probably the biggest weakness of the AI since it currently doesn't understand turn order. These tend to be more common in large/complex maps such as NWO. This is essentially the next thing I'm going to focus on implementing.
- crazy_german is pretty much spot on. These 3 things are definite weaknesses of the AI which are important on NWO: little logic for capitals, can openers, and nations that are mixed land-sea powers. These are also fairly complex issues that beginner/mid level human players struggle with as well. The AI also tends to play the Allies much better (if you run all Hard AI on NWO the allies win pretty handily every time).
- While there are definite weaknesses on NWO, if you want to see how far the AI has come then turn on the Medium AI and you'll see there is a big difference.
- I haven't gotten much feedback around the AI on NWO as most of the consistent pre-release testers haven't played it much so specific feedback is appreciated.
- 10% income bonus for Axis on NWO is way too low to even make things interesting. This was more of a how fast can I exploit the AI game. (I'd actually guess that even with both sides all Hard AI the Allies would win with that bonus)
- One benefit is now I have a few test games to use as I implement turn order :)
Cernel, do you have experience against the AI on 270 BC? I think 270 BC is a likley poor choice for a GOTM, the reason being that the AI will probably think every nation is a land power (like it does for Italy in NWO), and purchase no navy and avoid trying to transport troops by sea. This makes power located in the Mediterranean play rather silly generally.
I would suggest using Big World, Pact of Steel, or a WW2vX map. Something that can show off the talented AI, though a totally new map would also be useful for feedback purposes. If someone wants an interesting challenge, play the AI in minimap (I think it might be invincible)
I'd vote for World War 2 v5 as the next challenge.
When you compare v5 to another world war 2 game like v3 (AA50), I think v5 is a demonstrably better game for the AI right now.
v3 = AA50: canopeners abound, especially with Italy or UK, which can be very potent and prone to exploitation by the human. Also, AA50 has objectives and technology (both of which are problematic, the former because the AI can't track objectives, the latter because the AI invests too much into tech rolls.) AA50 also has more total nations in play with weird China rules to boot.
V5 = 1942.2: doesn't have any of those rules distractions and just 5 player nations, so I think it would be best for a fast game.
v6 = 1941 starter board: would likely work as well, though that map is kind of crazy, and kind of small, and the lack of artillery has a way of grinding the pace despite the simplified map/unit roster. So I like v5 over v6 as a way highlight the recent AI advances for most players.
Perhaps allow a dual challenge? 1 for Allies and 1 for Axis, with different goals, or different bonuses?
I think the AI functions pretty well in v5, at least in terms of game-pacing, since the scope of the map is more limited than some other tripleA games and there are less total nations/rules to deal with. For a Game of the Month Challenge, a map that showcases what the AI does well right now, would be more fruitful than one which shows its current weaknesses.
I would suggest steering clear of Global for example, or games that rely heavily upon the politics or national objectives feature, since those kinks still need to be ironed out for the AI. But v5 feels solid. For a harder challenge, you can just give the machine an income bonus or a bid.
I think crazy_german has it right with the weaknesses of the AI. Though I would say a challenge like game of the month is always "find the best way to exploit the AI" ;)
To me better capital defense would be the most important improvement (which would involve handling can openers). Many other things can be balanced through higher income for the AI. But if he doesnt defend his capital properly income bonus does not help.
Oh and sometimes it would be good to check if he can hold the capital when recapturing. Because that way you can basically "farm" the PUs of a defeated nation.
well for future games of the month I would suggest 2 ideas
first is a survival game: you have to defend a certain land (e.g. finland) against an extremely strong AI (like 300% income) as long as possible
second is switching sides: this means playing the part the AI played before. For example trying to win the game as axis after Berlin has been taken. (it would need an interesting savegame though - something that is challenging but not impossible)
By the way AI does not always defend accordingly to good TUV exchange. There is a example of AI not defending his fleet properly in turn 2 of my game (It was +20 TUV for me). It happens in land situations too.
@ Cernel: I agree it should be in the warclub, so the mainthread actually is there. But as Redrum pointed out, traffic is low, hardly anyone comments there. More important, forum options really suck. It's a pain or impossible to upload pics and savegames. Two good incentives for me to improve on that.
That is why I decided to put the challengethread here. I am also advertising it at Axis & Allies.org, though I'm not so familiar with that site. Maybe will move in time.
@Bel: Those are nice idea's! Survival-mode seems fun, and the reversal I already played several times when I got the AI pretty on it's knees. Though with this AI it will have gotten a lot harder!
I've posted a savegame to the AI development thread where Italy gave its fleet up turn 02, I'm also noticing many more poor TUV exchanges from the AI in larger maps. They are a pretty rare sites in medium sized games (unless you exploit turn order of course)
Here is another idea for a game of the month. Your allies are played by the AI, the human gets only a single nation. So you have to maximize the impact of your nation without being able to coordinate well.
Also, consider implementing a hard bonus (+15PUs per turn) rather than a percentage boost. This keeps the income high even when losing territory or getting bombed heavily.
@crazy_german - Agree. Large maps it tends to be much more fun to only play a single nation say germany, UK, russia, italy on NWO then play an entire side. Sort of have to figure out how to make the biggest difference while working with your AI allies.