You shouldn't count on an AI update because then all the single player records and what have you will be null and void. I got this off the Warlight WIKI by the way.

Fizzer has stated on multiple occasions that when he'll make an update to the AI, all single player results will be reset. *However*, every time I've seen him say it, he did so in the context of "not making a minor fix, even if it's quick and easy". I cannot remember ever having seen a statement he didn't want to update the AI at all. Quite the opposite really; to the best of my knowledge, he is indeed planning to give the AI an overhaul at some point, it's just not currently at the top of his priority list (I think the mobile clients are??).

So, it will probably happen, it's just one of those "later rather than sooner" things.

The most recent statement I have seen Fizzer make about this, is that he wants improved AIs, but also wants to keep the single player stats valid. As such, he wants to make a system that allows you to pick your AI (eventually). After that, he will be able to make changes to the AI, even without changing the single player levels or invalidating their statistics.

Thanks for the suggestions

For the record I did some test games with complete fog but giving everyone spy cards

Seems the AI can still tell where the blockades are? !!!

Grumble grumble
[Link text](http://WarLight.net/MultiPlayer.aspx?GameID=2514192)

Take note of turns 63 and 64 in Antarctica. The AI clearly could tell that the blockade was gone and now occupied by a human even though it should be able to see this

It is a bit annoying, I've never played a stalemated game like that though, lucky me.

I guess this shows that humans are cleverer then computers.

The only solution i can see is a trustworthy human being the zombie, maybe you could on the ZombieRageVirus account. I dunno though, I'm no expert.

You are lucky Arun, stalemate happen more often than not unfortunatly. Thats the main reason you only get 1 blockade

no it doesnt have the same feel when a human plays as the zombies because that person can actively try to kill off one person who is being annoying. The main fun for me is that you dont quite know what the AI is going to do.

Turn 19, the orange player blockades Sudan. Earlier that turn the AI captured it first territory bordering on Sudan, so it can't have any information about the state of Sudan yet (from previous, failed, attacks for instance).

Turn 21, the orange player abandons Somalia. At the moment that happens, the AI isn't even bordering on it, so it can't possibly know *anything* about the status of that territory.
During that same turn, it captures Congo, so now it has a border with Somalia (but, there's complete fog, so it still shouldn't be able to tell anything about it yet).
(During turn 20 and 21, it never attacks Sudan. To be honest, it doesn't have too many armies in the area, so it's possible it would like to attack but simply can't spare the armies.)

Turn 22, the AI has 114 armies in Congo. For ease of reference, Congo has connections to the following territories (in alphabetical order):

- Angola (blue player)
- Botswana (blue player)
- Cameroon (AI)
- Chad (AI)
- Kenya (blue player)
- Somalia (wasteland)
- Sudan (wasteland)

The orders given this turn are (in chronological order):

- 2 armies attack Botswana
- 110 armies attack Angola
- 1 army attacks Congo ("original order sent two armies", details below)

Or, looking at it another way:

- Do not transfer any armies to already owned territories (makes sense).
- Attack all surrounding, *player-owned* territories in the standard "plenty armies available" way, apparently assuming each territory only has a single army on it (as expected).
- Do not attack the wastelands *at all* (this is extremely suspicious behaviour to say the least).

I think you should report this in the bugs forum. Mind you, Fizzer has repeatedly stated he won't make incremental changes to the AI, so it won't be fixed shortly. However, I do think he needs to be aware of this problem for when he finds time (and priority) for a massive AI overhaul.

---

Details about that weird 1-army attack

As described in the [Possible bugs in AI attack logic](http://warlight.net/Forum/Thread.aspx?ThreadID=3143) thread, when the AI has an overwhelming number of armies available and surrounded by territories with one defending army each, it will attack all of them with 2 armies and boosting one of the attacks with "everything else". However, it doesn't account for the army which needs to be left behind, resulting in the last attack always "requesting" one army more than is available. Except when the last attack is the "everything else" one, this means it will be a 1-army attack.

Also look at turn 19 in the same game linked above: Nigeria starts the turn with 6 armies, receives 3 reinforcements (for a total of 9, obviously planning four 2-army attacks). Then the attack into Nigeria gets 3 armies ("everything else", just not a very spectacular one), leaving only a single army for the last attack (into Ghana). If you hover over the order it will still show "original order sent two armies" though.

Hm, I wondered why the AI sent so many 1 army attacks. It's actually not a bad idea when it only has one army to send, but it frequently does it when it has plenty of armies.

Thanks for doing the research on that.

Actually, it is a very bad idea. Using default kill rates and luck, these are the chances:

Defender Defender
killed survives
Attacker killed 42% 28%

Attacker survives 18% 12%

Not so bad you say? Well, the point is, if both the attacker *and* defender die (the most likely case actually), the defender is magically resurrected (since otherwise the territory would have to revert to neutral or something else weird).

So this boils down to a success rate of just 18%, a chance that nothing really happens of 12% and a pointless suicide (attacker killed, defender remains) of 28% **+ 42%** = 60%.

If you only have one army to send, don't send it, it's probably a waste. Instead, wait until next turn and for now, just let it defend. If your opponent decides to do a counter-attack, 2-vs-1 has a better than 88% chance of success, whereas 2-vs-2 has less than 17% chance. If he attacks with 3 armies it's > 96% vs < 69%; one additional defender is very valuable, so don't sacrifice it unless you're very, *very* desperate to break a bonus or earn a card piece or something.

If you have zero armies to spare, but taking that territory is still worthwhile, then it's not a bad idea to send that 1 troop alone, unless that one is still needed for something.

Watch pwn Trackers play sometime (he sends tons of 1v1 and 2v2, still rated above 1800 on 1v1 ladder) to see how often the attacks work out. 1/5 is pretty low, but that's 1/5 games where he gets a bonus one turn earlier, breaks an enemy bonus he would not have otherwise, etc. Having a strategic territory, or denying it to your enemy, is generally worth much more than the single troop used to do it.

It's clearly situational, and if you have the extra troop(s) to send, it's almost always better to do so, but that doesn't make the attack worthless.

Quite a few #1 players have done it in games they won. I first saw the Impaller do it in a game. I think that Trackers does it too often, but in the games it works, he gets a big advantage, while the disadvantage for when it doesn't work is generally quite small.