AI Co-Op TeamPlay

[Scouty]

Had a game against Seraphim Turtle instead of Aeon on Harbours. I feel Aeon is better mostly because of the Colossus poking its head out of the water in all the shallow shore areas. The Ythotha is much easier to kill without any danger to your navy.

It's too bad you can't prevent nukes without banning Seraphim battleships. Does the AI even build those nukes?

And maybe we should've put the AIs on Sorian Water so they spam more T3 subs...

In case you're interested:

[Adraius]

The AIx can sometimes play poorly because it gets an engineer stuck trying to leave a plant. I don't think that was the reason for the big difference between the two games.

In the game you say you easily won:
1. Many players would find what you did there difficult ...not easy lol
2. Luck was on your side ...although I believe you make your own luck.
3. The middle player johnj was very strong compared to the mid player in the 1st game. Also Pucu played better in the 2nd game.
4. After a few games on a TeamPlayAI map you learn the AI tactics and therefor know the AI weaknesses, also you learn the sequence of AI attack. T1 land, T1 bombers, Commy attack from Rush AI, T2 land, T3 strat bomber, ect...knowing what comes next helps.
5. Learning the landform is also a big factor. It is at first very hard to see the slight height changes across that map.
6. I think you had the [email protected] which is high enough for most people to enjoy a good fight. You could set the [email protected] but I don't think it will make that much difference. I may be wrong as I have not tried it that high.
7. Only other way to improve AIx performance would be to use just 2 AIx Sorian Turtles and set a buildtime to allow AI time to expand. Say 5mins.
8. Most small maps (5x5 and 10x10) can be easily beaten by your 'Charge of the Light Brigade' tactic. That is why most of my maps are 10x20 to give better separation and make it harder to rush the AI before it gets properly set.

Thanks for the replays they are an excellent learning tool to improve future AI.
roj

Heh, well, I'm using the word "easy" in a relative sense. =P And you are correct about knowing the AI's "battle plan;" I knew to build T1 mobile AA to counter the bombers, a T1 PD to stop the commander rush, etc. What stunned me was the effectiveness of the arty. Normally I use a factory spamming tanks to provide a suicidal protective screen when I need to push out with engineers to expand the perimeter, kill T3 mobile arty, etc. I think that game I chose arty instead because they could out-range Auroras... but when I moved them out to put pressure on the enemy base, I fully expected them to die under a stream of Auroras. Instead, they laid waste to the outskirts, so I kept producing them and pushing further (you can see them strip 1/2 the HP off the enemy commander in one volley, muhahaha). As for the dreaded elevation ditch, it didn't hinder me much (for once), and I even got some revenge when I attacked the center with my arty and the line of enemy PDs shot the ground in front of them. =)

Charge of the Light Brigade... I like that name. And I'm beginning to brainstorm the best ways to export my brand of terror to your larger maps... Agenda 21 looks like a prime candidate.

EDIT: And I found watching Gyle/felixlighta's latest cast to be EXTREMELY vindicating. xD

[AwarE]

@Scouty: That was a good comparison between Turtle and Water AIx playing side-by-side. I agree that Sorian AIx Water did far better than the Turtle. The main reason is that the water AI builds transports and expands very quickly. Also Turtle AI builds too many T1 land units that really just get in the way and mill about. If you check my recommendation under the map I do say to use Water AI at all levels of player skill.

On the question of faction ...I did recommend Aeon for these reasons:
1./ The GC has better HP for underwater travel.
2./ The chicken's death charge is wasted on subs.
3./ I found Seraphim has poor anti-air on water ...seems to build T3 SAMs on water because its Cruiser is useless.
4./ The AI gets confused whether to have destroyers submerged or on surface ...seems to prefer them submerged.
5./ Aeon has better Subs, Cruisers, Destroyers and has a Battleship as well as the long range missile ship.
6./ Aeon T3 LRC have good range and accuracy.
As you point out there are other factors:
a./ The GC can kill navy in the shallows because its weapon is out of the water.
b./ The Seraphim does not build its best naval units ...Its Battleships are restricted[due to nuke weapon] and the Yathsou is ignored. This T3 Sub Hunter has Anti-Air, Sonar, Submersible, Torpedoes and Torpedo Defense. Why wouldn't the AI use them ...because Sorian probably thought Battleships were better option so they are not included on the AI's build orders.

I guess the AIx multipliers [email protected]?
Khesh was a very worthwhile partner that actually beat the AIx in 'Mass Spent' by 125% and 'Energy Spent' by 150% and that is not easy to do. Increase the multiplier to 1.5 next time you play with him and consider putting an AI Tech on the right-hand isle.
I'm glad you tried a different faction as we need to check if there is a better faction than Aeon. Cybran or UEF may do better than I expect, it depends what they build under pressure in a big match like the one you played.

Thanks for your input toward making Co-Op verses AI games better.
roj

PS. I agree it would be better to restrict just the nuke weapon ...I have tried with other maps but it seems to cause other problems with the AI build. Crazy things happen like the AI won't upgrade mass extractors to T3.

[stalewee]

I hate AI chickens. They control the ghost and move it into your base.

[AwarE]

Yes ...they are excellent on land as the sparking charge does as much damage as the live chook.
roj

[Adraius]

Because you so kindly provide us with profiles on your latest maps, I thought I would return the favor and write a review/guide on one of your newer maps, Agenda 21.

First Priority
Because of the minimal number of mass points on your starting landmass, you need to secure yourself a chunk of the mainland ASAP. Two tempting targets are the mass point cluster in the far north, and the cluster on the mesa to the first cluster’s south, both of which are best reached by early ACU push or engineer drop. However, reaching and fortifying the mesa is difficult, as the AIs begin spamming T1 bombers as early as the 3 minute mark.* In the meantime, make good use of the plentiful reclaimable rocks strewn about the map and put the mass towards fortifications and mex upgrades.

Naval Aspect
As the main AI bases are in range of cruise missile-equipped cruisers, cruisers can be a huge asset; targeting clusters of energy storages will result in the destruction of much of the main bases. (suicide strat bombers also work against unshielded energy storages) However, the cruisers will require significant escort to get behind the AIs’ bases. Even if you do not pursue this strategy, the AI will build a significant naval force, making the construction of a defensive fleet a wise idea. However, as the AIs’ cruisers do not possess cruise missiles, it is possible to completely cede the waters without many negative ramifications. Also note that the AI heavily prefers T1 naval units.

Land Aspect
AI land assaults are surprisingly minimal. Expect periodic waves of Auroras, progressing to T2 and T3 units. Early on, these are tricky but completely defensible, particularly if the commander prioritizes T2 point defenses or tanks. The unusual geometry of the mesa makes defending it with point defenses, or any direct-fire unit, a tricky proposition. For this reason, any force operating near the mesa should include a contingent of T1 mobile artillery. In my game, GCs were produced but not used offensively. Still, some method of dealing with GCs would be wise.

Air Aspect
The AIs’ main offensive threat is the Czar, which it produces in significant numbers of on this map. Producing sufficient numbers of ASF to down a Czar is difficult before the first Czar is constructed. 50 SAM launchers are needed to 1-volley a Czar, at a cost nearly equal to the Czar itself; 25 (enough to 2-volley) is more reasonable, but will give the Czar time to air-crash onto your base. Be prepared to make repairs until air superiority can be established – and DON’T keep your ACU at your forward base.

Oddly, the AI did not use nukes, although if I remember correctly several silos were built. (another oddity failing is their total lack of stationary radar)

Overall, I found the AI to be unusually weak on this map. Aside from Czar threat, the AIs’ offense was largely ineffectual. Past T1, land assaults are spaced far apart. The AIs’ control of the center bay is mitigated by their lack of cruise missile-equipped cruisers. Perhaps most crucially, the AI only builds a single forward base. This means that you are spared from the barrage of tactical missiles and stationary artillery shells that characterize some other maps. This may also have something to do with the lack of ground assaults and offensive GCs. In conclusion, the only usually difficult things about this map are the lack of mexes on the starting landmass (use reclaim) and the Czars. If a player can handle the early bombers, minimize harassment from early Auroras, and muster enough defenses to stop a Czar, they are golden.

---

I still think that, despite the large map size, a properly refined Charge of the Light Brigade strategy could achieve victory, but would be much harder that I had hoped, and probably needs a player with faster hands and better unit control than I.

I’ve attached the replay of my victory.




*Also note that if you secure your mesa but your ally does not take his, both AIs will focus their air attacks and much of their ground forces solely on you. >.< See replay.

[AwarE]

@Adraius
GG ...well played by both you and Typo91. That was a true TeamPlayAI 'Allied Victory'.

You both used the UEF cruiser's cruise missiles to great effect. I must admit I did not consider that type of attack when shaping landform or placing AI defense markers. In fact I deliberately set the mountain heights around the AI base lower so that Air was able to clearly fly over them.

I thought you would have been ousted off the mesa when the other AI came around the back and attacked from the rear, while at the same time under heavy bomber attack. You were getting double teamed at that point.
The base that Typo91 established down the right side did draw forces away from attacking the mesa.

The AI navy kept control of the rivers early on when at only T1. I agree the AI navy should have gone T2 earlier, and soon after that T3. In the tests I ran the AI did T3 navy early.

The AI resource build was excellent 350K score in 30min and both AI built a Czar by 35min. Once you took down those Czars that were attacking Adraius's forward mesa base you were in good shape and by 40min Typo91 had an awesome airforce(asf and gunships) to defend against Czar or GC. You both had escorts for your cruisers and at 43mins the AI suffered a devastating blow from missiles by losing all power, that was followed up by hover, then percy drops to finish it off nicely.

Its good to see that you and others are now getting the TeamPlayAI setup correct.
I assume that the AIx multipliers were @ 1.4 ...replay showed 28 output for AIx Pgen.
The only thing I noted that could have been set differently to improve AI was AIx Omni was OFF.
Please try the map with AIx Omni ON. The AIx lost 10 Experimentals but only built 6 ...that is on the low side as it should produce more like 15-20. The only way to be sure to get more GCs would be to restrict Air.
I have not tested other factions on Agenda21...Some other factions may be good or even a mixture of factions if your cpu can handle it.

If you want to up-the-anti then add a AI Tech or AI Sorian Rush in the top mainland start positions 3&4 ...that will really cause some trouble when trying to make a base at the mesa ...Please leave start positions 5&6 vacant.

The replays are great feedback thankyou.
roj

[Adraius]

Here is a second review for you, this time on Crush the Castle 3. I tried beating this map a few times before my hiatus without success. Thanks to an extremely capable teammate, (rank of over 2000) we were able to defeat it today with the multipliers set at 1.4. (however, I did neglect to change the AI Omni to ON, and will rectify that for future matches)

In all previous attempts, (spawning in the north-east) my strategy had been to build two defensive fortifications in the open areas to my south and south-west while assembling a naval bombardment force. With my defenses absorbing the majority of the AI’s assaults, my hope was to allow my partner the freedom to tech up and spearhead a counteroffensive. However, this strategy did not succeed, both because of inexperienced teammates and the extremely rapid AI progression to experimentals.

Threat: the AI progresses to producing experimentals very rapidly.
Even in the victorious attempt where we were able to deny the AI mass points outside of the “castle,” the huge number of mass points inside allowed them to construct their first GC by approximately the 22 minute mark. (oddly, this first GC seemed unwilling to attack, although it had in previous games, generally tramping my northern defenses) This time period is reduced even further if the AI is allowed to access mass points further north.

My partner was Chinese and so our communication was limited to map pings, but by chance we both chose very aggressive strategies, which were met by varying levels of success. I brought inadequate tanks and anti-air to cover for my ACU as he attempted to set up a forward base, and compounded this error by failing to build forward factories. After finally managing to set up a firebase, I continued pouring my economy into 1) T2 Navy, 2) the defensive line, 3) mex upgrades. My insistence on navy made kept my defense vulnerable to a GC.

My ally, on the other hand, opted for a more dynamic and minimalistic approach utilizing very few stationary defenses and relying largely on the Ilshavoh’s superior range, health, and mobility to destroy several times their own mass in attackers. This sometimes resulted in trouble when the minimal shield and PD he used as a fallback was bombarded by tactical missiles. However, with only a paper-thin front line, he was able to devote his mass to upgrading mexes and constructing a chicken, which finished at the same time as the AI’s first GC.

Several accidental things synergized perfectly to enable our victory. Because my defensive line was closer to the AIs’ bases, it received the brunt of the AIs’ attacks, letting my ally focus on the chicken. My ally entirely ceded the sea, but because my naval buildup attracted the very few T2 naval units the AIs built, his only losses were the mexes along the water’s edge. I also managed to attract the AIs’ second GC, allowing me to bring my main concentration of firepower to bear while my ally sandwiched it with his chicken. (my ally’s prudent control of the chicken over the course of the game is especially commendable) Even the dead GC’s proximity to my forward base was a bonus, allowing me to construct a Fatboy.

Conclusion: the AI on this map is a powerful opponent.
The concentration of mass points in the “castle” and diffuse spread of them elsewhere play into the AIs’ hands, giving the AI easy access to mass and making it difficult for the player to defend any mass points in the south, making ceding even more mass to the AI tempting. Finding an economical and relatively early way to defeat a GC is paramount. Remember the AIs’ knack for mass T1 bombers to avoid any setbacks, as you don’t have the time to spare. UEF/Sera navy has a lot of potential on this map for the same reason as on Agenda 21, but the game progresses so quickly the match can be decided before it comes into play; despite the fact I prioritized my naval buildup, it was too late to play a key role in our victory. (it primarily served as extremely expensive bait for the enemy navy)

With the huge store of mass available to the AIs, there is definitely no need for Rush AIs on this map despite its middling size. =P This map is a definite challenge and promotes teamwork between the players.

[AwarE]

Confrontation TeamPlayAI




Size: 10x10
Players: 2 or 4 Humans verse 2 or 4 AI

This is one of the few TeamPlayAI maps where I have set the markers for more than two on the AI team. So on this map a much wider choice of AI numbers and types will give good AI performance.

For example your AI opposition can be either:
2 Sorian AI Turtles(rear) <---- easy
2 Sorian AI Turtles(rear) + 2 AI Tech (front)
2 Sorian AI Turtles(rear) + 2 Sorian AI Rush(front)
2 Sorian AIx [email protected](rear) + 2 Sorian AI Rush(front) <---- hard
4 Sorian AIx [email protected]
2 Sorian AIx [email protected](rear) + 2 Sorian AIx [email protected](front) <---- extreme

Like all other TeamPlayAI maps the rear position markers are set for AEON Sorian Turtle AI only. Any other AI will not build correctly as they place their initial structures in the wrong positions causing an inefficient build.
The front AI positions are much less critical and almost any AI type or faction will be ok.

Please turn AIx Omni ON in lobby options, for hard to extreme games.

Good Luck Commanders!
roj

[AwarE]

These maps are big and if four humans are to verse four AI then all CPUs will need to be SB [email protected] or better to maintain good game speed. Restricting Air will improve game speed and provide an interesting alternative AI play style.

TeamPlayAI Setup Notes:
1./ Aeon faction AI and AIx will be the strongest enemy because the map markers have been optimized for Aeon ...Any other AI faction will not run as well ...Seraphim would be next best.
2./ Set AIx Omni = ON
3./ 2x resource mods half the AI strength. NO 2x
4./ Set units = 1000
5./ Fog of War = Explored
6./ Pre-built units will cripple the AI build. NO Pre-built.
7./ Do not restrict Water on 'Bay of Plenty TeamPlayAI'.
8./ Set Civilians = Neutral on 'Hardcore TeamPlayAI' to place 'capturable' engineers of all factions.
9./ Ally all AI or AIx on same team.
10./ Follow the instructions in each map's description to place AI in the correct positions. Only in these positions have markers been set for the optimal AI build.
11./Set 'Victory Condition' to Annihilation. The AI will fight on even if its commy dies. If any Human commy dies the Human team has lost the battle. Your game will continue so that you can play the final stages to see the AI end game tactics. Bear in mind that although you may destroy all the AI it is still a LOSS to the Human Team. Simply put ...the AI WINS if any Human commy dies. I am working on a map script that will end the game when any Human commy dies.


Bay of Plenty TeamPlayAI

Size: 20x20




Two Humans verse two or four AIs.
Two Humans in North start positions 1&2 verse Two teamed Aeon Sorian Turtle AIx in South start positions 7&8. Please leave positions 5&6 vacant as this is essential for the AI build. Pros can add AIx in start positions 3&4 to increase difficulty if needed.

Restrictions: No sat, LRC, para.

Difficulty:
noobie: 2AIx Aeon Sorian Turtle AIx @1.0
mid: 2AIx Aeon Sorian Turtle AIx @1.2
good: 2AIx Aeon Sorian Turtle AIx @1.4
elite: 2AIx Aeon Sorian Turtle AIx @1.4[in South] + 2 Aeon Sorian Rush AI[in North] (Rush is non-AIx)
Pros: 2AIx Aeon Sorian Turtle AIx[in South] + 2 Aeon Sorian Rush AIx @1.4 [in North]


Hardcore TeamPlayAI

Size: 20x20




Up to four Humans verse two or four AIs.
Humans start in positions 1, 2, 3 and 4.
Sorian Rush AI or AIx start in positions 5&6.
Sorian Turtle AIx start in positions 7&8.

Restrictions: No sat, Experimental cannons, rapid-fire cannons, para.

Difficulty:
noobie: 2AIx Aeon Sorian Turtle AIx @1.0
mid: 2AIx Aeon Sorian Turtle AIx @1.2
good: 2AIx Aeon Sorian Turtle AIx @1.4
elite: 2AIx Aeon Sorian Turtle AIx @1.4[in South] + 2 Aeon Sorian Rush(non-AIx) AI[upper South]
Pros: 2AIx Aeon Sorian Turtle AIx[in South] + 2 Aeon Sorian Rush AIx @1.4 [upper South]

Merry Christmas 2012
roj