Posts made by Radde

@Uveso
Hello There!
I've been testing other AIs for a while.
M27AI is the most impressive when it comes to eco management and great placement of structures but it lacks naval logic.
There are a few things that you could implement to improve the Uveso AI.

1. Always placing a t3 shield that covers a t3 arty and a group of t3 engies supporting it.
   It's a great defensive thing that M27AI does and it makes any AI extremely hard to crush in arty wars.
2. I've noticed that Uveso likes to use gunships while M27AI is more bomber focused.
   Spamming t1 bombers to counter incomming land experimentals would be a great thing.
   Also using t2 and strategic bombers to kill unprotected mexes would make games more challenging.

Since the no rush for game enders option Uveso struggles to pump all of his available eco into units. It's always overflowing alot, not making a good use of mass untill it can start making experimentals, arties and nukes. I would suggest improving its behaviour before the game ender timer, making it swarm players with everything it has (t3 land, air, navy).

I've also noticed a strange behavoiur with walls. It feels like the AI doesn't know how to counter it and gets stuck. The AI is piling lots of units outside of the walls and doesn't attack it or change path to overcome the obstacle. Unfortunately i don't have a specific replay that i can post to show the problem... it's something that I forgot to mention when I gave my feedback the first time. To fix this I would suggest simply making the AI treat walls like a standard enemy with low priority. Simply getting rid of it by attacking may work well.

I also have a suggestion regarding no rush for game ender option.
Could you split the option into 4 categories?
One for land and air experimentals, second specifficaly for nukes and nuke subs, third for t3 static arty and last for the t4 game enders.
It may get rid of the eco problem that i've mentioned before, because if we let the AI to make the land and air experimentals earlier, before other game enders, it would pump mass into them.
It also would make the game more fun for the players if we can decide about the flow of the game. For example if the AI can start making land and air experimentals min 20, we know what to expect and it's possible to get prepared, then the next phase would be nukes by min 25, then static arties after 30 and t4 game enders late in the game 35~40 min.

I'm looking forward for the future improvements!

@Uveso

Hello there!
I've been playing against AI Uveso for several years. I often host the most insane games when it comes to difficulty. 2.0 build cheat multiplier and from 2.5 to 4.0 resources cheat multplier, so i guess I am experienced in this topic and I would like to give You my feedback.

First of all, great job in creating such a fun mod! As i said before i've been using it for several years and it's because it's so much fun that i keep challenging myself in trying to win harder and harder games against your AI.

I used to play against classic supcom AI back in the days... and it was not only weak but it also had several issues such as slowing the game the longer it lasts. I was happy that FAF has implemented the new Sorian AI which didn't have this issue, and the AI was much more challenging and fun. But unfortunately Sorian AI had issues aswell... for example it didn't make t3 arty or game enders due to a bug. When i installed Uveso AI and tried it i was really impressed about it... AI destroyed me in few minutes and i couldn't believe how smart and reactive it was. The tactics it used were on another level... something that sorian couldn't pull off. Air was used correctly in hordes, gunship ambushes, teleport SACUS, crashing experimentals on strategic buildings, smart nukes.

It took me few years to master how to play against it and there are few things that could be changed to improve players experience. In early versions of AI it used to launch nukes as soon as it is loaded unless AI scouts an smd in players base. It used to be annoying to rush the smd before min 13 just to prevent AI from nuking. The 35 minute mark was a great improvement since we don't have to rush an smd anymore, players can focus on other stuff and when the time gets closer to ~30min we can start building an smd. Here is my question:

Is it possible to expose some parameters in a game lobby such as:
Time before AI can start building t3 arty
Time before AI can start building game enders?
In current version of this AI there is similar problem to nuke time mark, and the problem is with arties that AI can start building very early due to high economic boost. Most of the time AI finishes an arty before min 22 which is hard to deal with, and the greater eco boost the earlier the arty is built. I would rather deal with more units or earlier experimentals before 25 min than being focused on tanking an arty and building my own to deal with it. Making it possible to decide when players can expect such dangers would improve our experience!

Experimental crashing is the thing that i would get rid of. It's either annoying or pointless and let me elaborate on that.
First of all, most of the flying experimentals are used by the AI to be crashed... i don't remember if it was the other way back in the days, but i haven't seen many flying exps being used the ''normal'' way, for example bombing with ahwassa or harrasing lone mexes with the soul ripper. When we see an experimental we're like ''Oh, i wonder where its gona crash'' cuz it's used that way 95% if the time... same goes for gunships but gunships most of the time are used properly, but sometimes they're flying towards something in a straight line and crashing at pointless stuff.
Second of all, it's annoying when experimental crashes perfectly on the most important building, for example pgen next to t3 arty which ends the game for players... 20 minutes wasted because AI has decided to perfectly position his ahwasa and crush it on something that is the only hope for the player team.
And the last thing about experimental crushing is that AI most of the time is targeting pointless stuff. Crashing an experimental on t3 mex is a huge mass donation, but it may target things like walls on the front (It may sound absurd but AI actually does that and i have some replays of it). That's why its either pointless or annoying, it may end the game by crashing on our smds or t3 arty or it may make a huge mass donation.

I'm looking forward for the future updates! Great work so far