AMAI Vs. Blizzard

How do I change it from random to something else? Where should I be looking? If the manual covers this, which section should I read?

What I'd really like to be able to do is something like IF in an odd numbered team THEN Blizzard, otherwise AMAI.

Is that even possible?

Miss other thread?


Warning: Do not use any map with mines (or fountains!) blocked by trees, rocks, water or elevation.

If AMAI has to expand multiple times, it will inevitably get stuck trying to reach one of the inaccessible mines. Its army will plaster itself against a treeline, or shore, and ignore all else. It doesn't matter if there are other mines available, once it decides to creep clear a blocked mine it will go into an endless loop trying to reach it.

Blizzard AI, likewise on longer matches, will inevitably run out of money and be unable to expand. As HU or OR, it will walk peons over to another mine. Without income, Blizzard AI as NE or UD will sit idle in base, pretty much no matter what size army it has.

Unfortunately, this means no Lost Temple, Twisted Meadows, etc.

And expect long matches :) Suprisingly, I'm not seeing many decisive victories from either side. With this bug resolved, hopefully next release AMAI will handily beat Blizz AI each and every time.

I think AMAI often loses because it either:

a) Tries to be too clever (like raids) and when they don't work out it doesn't extract itself quickly enough.

b) It doesn't focus fire, its targetting is as vague as Blizzards so even with an effective counter it often ends up dishing out a lot of damage but not killing anything.

I too was surpsied, especially when it lost a 1v1 on Two Rivers as Orc against a Human. Hopefully we can identify WHY that is. It could just be that particular personality (which makes testing a nightmare).

I noticed that while it would keep its main army at its expansion (and it expanded later than Blizzard) it would keep a small contingent at its main base. It failed to unite them when making its big push, even when one retreated past the other.

I'd forgotten all about that post, thanks, I'll check it out.

It would be a good idea to track which profiles win or lose. Tough part is the replay renamer utilities do not save custom games. Doh.

How exactly can you do real time AI? For instance, the AMAI should focus armor types that take extra and the strats should be more solid against a particular race, (for instance ghoul/garg vs ne and hunt/archer/dott/dryad vs ud). For instance having the gargs focus hunts and ghouls focus archers starting with nearest enemy and continue on that way.

It appears to me that the AI has issues because it has no real time decision making (other than blizzards already done stuff). Also, all computers run away when they have a significant advantage. Perhaps there should be something that sends weak units home instead of sending entire army home to heal, then remember to heal the units sent home when the heroes get home.

Right now it looks like AMAI just attempts to counter what it sees by going random strats to counter it, whereas the strat shouldn't necessarily change, just add more of one unit and less of another depending on what the enemy has.

My thoughts anyway, I'm sure you guys know more about this, but I really like AMAI and would love to help out.

Turtle Rock, human mirror match, AMAI (Team 1) vs. Blizzard AI (Team 2)

AMAI resoundingly massacres Blizzard, 12 to 2.

Didn't really watch much, few minutes here and there. Sometime soon after game #15, I saw Bliz vs Bliz playing!?! There's a glitch somewhere.

Sorry other scores are mixed in, I forgot to reset leaderboard prior to the tourney.

-----------------

Deathrose, 1v1 random, AMAI (Team 1) vs. Blizzard AI (Team 2)

AMAI cleans up in six consecutive wins. Impressive.

Game #7, AMAI errs badly in a too cramped NE base build. A single moonwell traps its Level 1 DH and all its huntresses in base. Crippled, AMAI fights off Bliz OR with a Warden, archers, faeries, and MG's. Bliz eventually marches to the NE base with a 90 food army of taurens and demos for the win.

Game #8. Deathrose (bonus map) sports nasty red creep camps loaded with spiders guarding corner mines and merc camps. During the past seven games, both sides have struggled with these creep camps, losing units and even heroes. This game, AMAI picks a particularly bad time to creep, losing its hero and many units, just as Blizzard descends on its main base inflicting heavy damage. AMAI never recovers, Bliz smells blood and after repeated pushes AMAI surrenders.

Final score: AMAI 6 Blizzard 2.

---------------------------

Timbermaw Hold, 2v2 random, AMAI (Team 1) vs. Blizzard AI (Team 2)

AMAI: 2 Blizzard: 3 (Each game counts as 2 w/l on the leaderboard)

Despite having the advantage of using health fountains and healing pots, AMAI averages a dead heat against Blizzard in team play.

A few things I noticed:

Frequently AMAI spent more time micromanaging units than attacking with them. Case in point, first game AMAI drew double orc and quickly fielded a massive grunt rush while Blizzard was teching and creeping with token armies. What should have been a quick rush shutout game was drawn out well into tier 2 despite Blizzard being hopelessly crippled. AMAI was overly preoccupied with zigzagging its army and retreating red units when meeting the slightest resistance. What do you do when you have two dozen grunts and an opponent with no army and no expansions? You FLATTEN his base and WIN. AMAI needs to know when to ignore losses when a win is imminent.

Repeatedly, both side's armies were scattered everywhere; healing, waiting for healing in base (Blizzard), creeping with a smaller force, strung out while retreating, etc. I never saw a well coordinated massed attack by both team members. One or the other always arrived late if in formation, or was strung out in a single unit line.

---------------------------

Deadlands, 3v3 random, AMAI (Team 1) vs. Blizzard AI (Team 2)

AMAI: 1 Blizzard: 3

A fairly large north vs. south type map with big juicy expansions at 17K. Makes for nasty fights and long games, and so AMAI and Bliz indeed have four very long 1hr+ matches. Lots of expo attacks and repeatedly rebuilt expos, high level heroes each game, ultimates going off every which way, tier 3 armies in each game.

Many inconclusive attacks, for when each side forced a tp or retreat at an expo, they didn't press the offensive onwards to the main base. Following up after a successful attack (or counterattacking after a successful defense) is a given with people players. My deadland RT games average 25 minutes.

Instead each side repeatedly withdrew to base to rebuild forces after any action. In each of the four matches, the end game drew near only when one of the three team members, on either side, ran out of money.

In game #3, Chick3n fast techs to necro/wagon/stats and builds a respectable force. Only problem is, especially in a 3v3 game, this is a SUPPORT army which needs a hefty meat shield from a teammate before advancing. Chick3n finally attacks ALONE, face first into Bliz's Wyrm/Garg flotilla. Scratch one fast tech necrowagon strat. Chick3n apparently didn't budget for expanding, and spends the remainder of the game, heroless, with a few statues and wood ghouls sitting idle in base.

With people players, you are expected and required to fight with whatever you got. Those statues were needed for the surviving armies, and the ghouls would have been ideal for scouting expos and expo/base harassment. For extra credit, Chick3n could have unsummoned a few building to revive its heroes. At level 6, they could have put up a good fight solo.

Not that Blizzard didn't have idle armies in base as well when it ran out of money.

Again, saw evidence of AMAI dividing armies. Or when apparently ordered to attack in a team effort, armies arrived late, strung out in single file. Blizzard ultimately won, because 1) kept each army together 2) moved that army in formation 3) and as a result, stood a better chance at combining 2 or 3 teammates for attacks and defense.

Specifically for the Deadlands map, people players know they must get their armies across one of the four land bridge choke points intact. Fighting on the bridges can be disastrous for armies; melee units are blocked, some ranged units are out of range, and the natural choke points are choice sweet spots for AOE spells. Once across into enemy territory, they know to rally up as a team.

Some kind of team oriented Assemble Teams at Rally Point order is needed.

I ran Turtle Rock games to, doing 2 AMAI Humans Vs. 2 Blizzard Humans etc.

Blizzard won every game except the Human V. Human. Only ran one of each though.