Blogs - JonnyKnows

It didn't take me long to make the adjustments, but I did want to make an even more difficult challenge.

But actually playing this new version is... not fun, the AI are too strong with 400% income.
Now luckily for myself I did explain that it was a nearly impossible challenge with it.

I think it stands true really, I gave it my all and I can hardly slow down the dang AI's.

Very hard is an understatement of the dang mod.

If I do make another version of Brutal AI, it will be after 1.05 is finished.
I'll make separate versions of it so that you can choose to play against brutals who use 300% or 400%.

Honestly 300% would be more challenging and probably not impossible like it is now.

You're probably wondering why I don't make a new difficulty setting or change hard to 300% and keep brutal 400%.

Well for the difficulty setting, I have no idea how to do that and I don't think it's possible.
As for changing hard to 300%, it would take a couple days of work to change all the AI's to act like brutals on hard difficulty. My mod does indeed affect the hard difficulty and it has a lot of hard specific things in it.

So for me to remake the hard difficulty would mean editing each personality to use brutal specific stuff. This would take a long time, especially with all the AI I currently have implemented in Superior AI.

My only option will be to make alternate versions of it and pack it into one download.

The tournament was successful, Japan AI proved to be pretty dang good.
Although the finals was between two allied AI, the Japan did well and lost their semi final matches.

At any rate, I feel satisfied. I'm happy to release this amazing version of superior AI.

Kind of crazy thinking about how I've worked on AI for ra3 for over a year now.

I hope some people other than myself have had enjoyment from the mod.

I get excited when my AI find a way to beat me, it's a strange feeling really.
On one hand I hate losing, on the other I'm excited that I was managed to be outplayed by AI.

Anyways I'm tired as ever right now.

Enjoy the new version once it's authorized.

For a long time now I've wondered these particular questions:
"What makes allies AI so much better than the others?"
"What can I do to improve the Japan AI?"
"What makes an AI better than another?"

It took me until 1.04 to finally figure out something that may or may not work good for the Japan AI's.
Actually, one of my new AI's showed me something cool.
His name is Kyoichi, he is similar to Hine.

He defends like a Paranoia AI, and attacks likes Aggro.

However unlike Hine, Kyoichi is successful. With a record of 7-3 Kyoichi has done quite well during my exhibition matches vs other AI. Not to mention he has performed well in my last tournaments between my AI's.

He's shown me the effectiveness of Japan defenses, he has proven that they are one of his keys to being a stronger more formidable opponent for Allied & Soviet AI's.

It's not that he turtles, it's that he just uses defenses in general and it works from what I've observed.

Continuing, I also seen some success of mecha tengu use all game long, especially as an anti air unit.
So I've implemented these two improvements into all my Japan AI's.

As tradition I shall do a tournament before release 1.04 to everyone.
This will truly help me determine 2 things:
1: If I actually improved the Japan AI's.
2: Bugs, these tourneys include every AI including each form of aggro, paranoia, ace and complexity.
So tournaments are quite necessary in order to make sure no AI crashes the game.

So how long will you have to wait until 1.04 will be released...
7 days maximium. I run the tournament all day and I don't watch each match, I simply run them and mark results. So it'll take about 3-4 days to complete, there's a lot of AI's ya know.

It may take longer if I end up playing some games myself, which is very likely.
Then it'll take like 12 hours for the release to be authorized by moddb.

As a reward for reading this, I'll say this.
There's more new AI's than 1.03.
One of which builds absolutely no defensive structures, and another which loves to build century bombers.
A soviet AI who starts with 3 barracks.

AND A KIROV RUSHER!

Well as I was feeling a bit bored and unsure what to do, I took a look through some files of superior AI.

So during that I found some motivation to do some major work with the Japan AI.
Using my guinea pig AI, Ace. I had begun experimenting a lot, and at last I've made
some serious improvement to the Japan AI.

From what I'm seeing from Japan Ace, it's performing way above my expectations.
In 2 hours I transformed my weak Japan AI's into more balanced competitors.

The idea was to increase mecha tengu use throughout the match.
Along with testing a new way for it to spend money, both on defenses and units.
Like Soviet and Allies, Japan Ace is playing in a much more balanced way and is defending itself more effectively against it's opponents.

What does this mean for 1.04?

It means that Japan AI's in general should ignore units less often and be less obsessed with other units like Wave force artillery. Of course that depends on the personality.

Some AI are supposed to be a bit obsessed, speaking of which. I've changed paranoia's spending's to prefer defenses even more. Paranoia is supposed to be a bit more defense spammy than other AI.
It's essentially a turtle AI, yet it plays pretty aggressive at times.

One of my biggest tasks is trying to make the AI fit their play style appropriately.

1.04 Is almost ready to be released.

Even I'm wondering what exactly comes next. Could be anything really, I have a few ideas in mind.
But nothing official yet, I suppose I should do some things to make the AI more functional in infinite money map situations.

I've done a lot with the AI, but I wonder how it can ever be even more challenging.

Believe it or not the AI really do end up defining themselves in the end, none are intentionally better than others. Some are intentionally more unique though.

But overall, they prove themselves to be good or not. I've spent hours trying to figure out what really separates the good AI from the bad ones, making AI similar to the ones that are good isn't something that works.

I've tried it before, a few times now. It usually ends with bad results, the skill of the AI seem to be pure coincidence.

So I wonder, what will it take to make superior AI go even further.

Perfecting what I have will make it better, but the AI won't be more difficult with that.
Why is it that, I can beat my AI while they get double income, but of course I'm by no means a noob or avg player, I've got plenty of experience too, I suppose that would be a good explanation of why I can do it so easily.

BUT, still the AI have twice my income and it really shouldn't be a problem for them to out build me.

But with the right strategies, all my AI's are pretty easily beaten.

However, there is one simple solution; with the release of 1.04 there's gonna be a big surprise for all the guys who also find the AI to be a tad bit too easy. So if anybody reads this blog, in the next few months when I complete 1.04 look forward to that "surprise".

To make AI for another mod, like one with new units and such.

It would be interesting, that's for sure. I mean anybody with some understanding of the ra3 AI can make AI for it, but to make the AI truly effective is much harder.

There's a lot you need to take into account when trying to get the AI to be more difficult.

I've surprisingly done quite a good job at improving the AI, the reason I say surprising is because it's difficult to make it.. better.

I mean you can always do simple changes like fixing their build orders, changing certain tactics up a bit so it's more effective.

But changing almost anything of the AI can drastically change it for the worse. I'll be honest and say many of my new AI's in 1.02 aren't very good, and probably wouldn't beat some of the original AI's.

However, some of those AI's are very good. Where my AI excel the most is taking advantage over free time. If you let up on the AI, they won't hesitate on harassing you and holding you back while they expand as they please, they'll also tech during those moments too.

The largest flaw in my AI is their defensive play, and there isn't too much I can do about that.
My AI rely on gaining an offensive advantage, when they fail at that they tend to get awkward.

Of course AI micromanagement is still, really bad and overall AI always get themselves into weird situations.

In the end I do indeed have idea's for what to fix in 1.03, if that day comes.
I've added a crap load of new AI's, what else can an AI mod accomplish other than making AI even more of a challenge than the last update.

I suppose 1.03 will be a hole filler, for example.
In 1.03 I plan on making AI capable of recognizing their playing on a map with lots of starting cash, like and infinite money map or 40000 money starts. I'll make the AI have more production throughout their build order in such an occasion, and have them want to build multiple structures right after their build order is complete.

I also want to give the AI more unique island map build orders instead of using the same old ones.

I have some other adjustments to make as well, specifically allied AI unit choices, they tend to over produce IFV's and apollo fighters. Then they go onto to not building aircraft carriers, if an AI has too many assault destroyers the AI tends to not attack with them. That could be changed too.

See what I mean? There's still stuff to finish, but nothing significant to make the mod any more enjoyable.

Maybe I'll think of something one day.

I'll write on the news page eventually, but I really want to take a break from making AI for a while.
I hope 1.02 pleases, once it's finally an upload working.

I had an immediate problem with 1.02 involving AI not landing tengu's, I've fixed it but I had to change the file superior AI 1.02 was using.

I hate having to wait for authorization.

I'm off to a tough start, the very first tests against warren has been failures.

Don't worry though, my japan AI had trouble at first too.

I'll just have to think about it and make adjustments until I figure out where I should be going with it.

More often than not, it's the style of attacking which can make or break an AI.

Sometimes AI have a tendency to sacrifice armies and make illogical attacking decisions.

But then there's times they do good. Unfortunately making an AI defend more efficiently is nearly impossible. On offense you can make the AI at least build up an army before attacking, with defense the AI doesn't make good positioning choices and doesn't really use tactics for defending. In fact it's like all the units are set to guard the area and they just engage no matter what.

Sometimes the AI don't try to counter things they should as well, they underestimate early infantry and infantry armies. There really doesn't seem to be a way to make them fear infantry more!

So when defending AI tend to lack, but Brutal AI have the double income advantage, which can even out the odds by giving them the capability to have very harsh offense.

My AI specialize in increasing their offensive advantages when they can.
With 1.02 they will even switch unit compositions once in a while!

One moment you'll be fighting off tsunami tanks and infantry, then next you could suddenly be up against a handful of tengus. The new AI will no longer ignore the usefulness of earlier units.
This will be most noticeable when playing against a Japan AI.

As soon as I make my official changes to the soviet AI, I'll be sure to explain how awesome some of them are!

Oh and on an end note, I played against 1.01 shinzo on fire island today.
I've played so much against 1.02 shinzo not too long ago. The difference between these AI are amazing. My old shinzo tried to steamroll me, but he lost in the end, because he ignored the usefulness of vx, infantry and tengus. My new AI won't have this problem anymore ;)

But he was tough, and I thought I was going to lose for a little while.

Now that I'm done taking a break from making AI, I'm gonna begin to work on the last part, Soviet AI.

Changes will be massive for the Soviet AI as well.

Once I finish the improvements and test them a bit, well... The 1.02 update will be ready for release!

So let me explain how long this might take me to finish.

If I work on it as fast as I did for the allies or japan AI's, it'll be in the final testing stage.
I've been very open yet secretive with the update, I haven't said how many new personalities will be in the game, or what each personality is capable of.
Or even how I came to make these personalities.

So altogether, if I'm putting in the work at my usual pace. The release of 1.02 could be within 7 days time at the latest.

I work extremely hard and fast when I want something finished.

So to explain more about what I have left to do let me, well... explain.
With both the allies and japan AI's I needed to make my AI better with attacks, and strategies.
I also had to make the AI capable of being versatile with the units they use for attacking.

The method I am using to make my improvements is a simple one.
Basically I make a brand new AI, borrowing only certain things from my superior AI.
I use this new AI and make different approaches to many ideas I had in the past and see if I can improve on them from a fresh start.
I also proceed to add new harassment tactics and improvements upon old harassment tactic ideas.

And finally, I make sure this AI can beat the original, Warren, Giles, Shinzo, Kenji, Moskvin, Oleg.
I do this because, once upon a time I made my superior Shinzo, fight the original Shinzo.
The original shinzo won 2-1.

Because of that, I now make sure my AI is capable of outsmarting the original AI before importing the changes into the entire superior AI mod.

Because, in the 1.00 and 1.01 versions of superior AI, the AI were more of a refreshed AI. They were different, but not entirely better.

In 1.02 the AI are based on improvements that made my AI better than the original, so you can count on a better challenge from my AI mod. But for skilled players it still won't be enough.
Trust me, I know that and I hate it too.

To get the best out of my AI mod, play on larger maps and don't rush them.
Anyways, hope some people are looking forward to the update.

Finished up all my allied AI upgrades, I haven't played against em much to be honest.

They seem to struggle with early attacks, but they feel enjoyable to play against now.
Especially with their hydrofoils.

To see a list of my new changes which has just been updated. Follow this link!

Mind you, most of the recent ones are allied related. But there are some global changes once in a while.
Moddb.com