Name: 硬いAIを倒す - スタークラフトIIのPython AI チュートリアル p.6 (Defeating Hard AI - Python AI in StarCraft II tutorial p.6)
Uploaded: 2021-01-14T12:27:53.000Z
Duration: 33 min 34 s
Description: VoiceTubeの動画で発音を聞きながら英語表現を覚えよう！学べる英語：

Everybody and welcome apart six of the Starcraft two.

What we've been doing up to this point is defeating the easy in the media.

May I, which both were easy and medium, I would say, is not medium.

We could not defeat the hardy I, but given how simply we defeated the easy immediately I just by simply creating some fighting units, I think defeating the hard eyes prime Not gonna be too difficult.

But before we do quick shoutout to my most recent sponsors group bought deep pox Amir Florian, Joe Wow, Tyler Durden and Faust and Nick Bender.

Thank you guys very much for your support.

If you didn't know you could support the channel, Click in that beautiful blue button looking Look at it.

Thanks again, you all for your support and let's get into it.

So at least looking at this there's a few major things that we that we have going wrong.

So first of all, this pot has no notion of time like how much time has elapsed in the game.

So, for example, when the game immediately.

Starts are bought has a goal of creating three nexus is or command centers whatever you wanna call him.

Uh, you know that that might be a worthy objective to have just immediately.

Is that are those three nexus is gonna be your primary objective, right?

Nexus is before you build a single military unit.

Maybe, but unless that's your goal, you need to focus on that.

Do we want to build three gateways right away?

Or do we just want one gateway and then take those resource is because the gateways are our version of offense or defense for a base, as opposed to, say, resource collection.

Do we want three immediately, or should we take some of those resources and start building an army?

Also, three's pretty good for the first few minutes.

But what if the match, for whatever reason, gets us toe?

Our enemies probably got many Maur, you know, unit producing buildings on, and eventually it's just they're just going to suffocate us, so a fixed number doesn't make any sense.

This needs to be a variable and probably depended on how much time has elapsed.

Or you could make it a variable based on how many resource is you have.

But this fluctuates so much you could maybe do like a resource rate or a variable, based on how many workers and gay or how many workers and nexus is.

You have something I'm going to do time next.

How many units is it gonna take for us to attack?

Initially, 15 units is a lot to send to an enemy base or if an enemy was the send 15 units to you.

But later on in the game, especially 15 of these little based stalkers.

And then finally, the other major issue is that as we produce new units, they kind of run towards the enemy.

And at least in this case, with even three, you have a scenario where one's maybe dying.

One's on its way, ones also on its way, and we just continually just send new units up one by one, though they're not clustered, they need to come together before they go.

Not really gonna solve that issue here, but just know that is an issue.

So what we're going to do first is give this bought a notion of time.

So coming up here, basically, you see, on this on step there is an iteration.

We can use that adoration to our advantage.

Now, we could just arbitrarily say, if thean aeration is greater than X or whatever, Um, then that would give us some sort of, uh, way to make rules around this.

But I think at least for me, I want to think in terms of minutes.

Hi checked in federation, there's about 1 65 iterations per minute.

Or if that is just how quick your processors are, I really don't know, so you'll want to check that for yourself.

I found it to be 165 or very close to 165 Didn't.

There's no empirical testing that went on there.

So, um, what we're gonna do is we're gonna define an initialization method here on basically yourself.

All we want to dio is define a constant here for generations per minute.

Just so basically, you want your bought to always be in portable.

Hopefully at the end here, if this video doesn't make it to, like, 6000 hours long, will probably end up copy pasta in a little bit off of the text based tutorial just so I can get through everything because never mind.

So what we need to do is self dot It orations her minutes 165 Now we can reference self doubt iterations per minute everywhere.

So self dot federation it oration equals it, Horatio.

Just so we can access this everywhere in all of these methods.

And then we can start building things with that in mind.

So the first thing that I want to do is like building a simulators like pylons and assimilate er's we just if we don't have the supply builds and pylons, right?

Pretty simple stuff and actually are pylons.

Our pilot should be much more intelligently built.

Um, hopefully I'll explain later to a little more or I'll show you guys.

But at the end of the day, this logic, despite not being like fully logical before we can afford a nexus, we can afford other things.

So it's only at the point where we have full accused basis or something's already being built, that we expand to a noon like to a new nexus.

Uh, you probably want to put some logic in there, but it just doesn't matter because of the cost of things.

But anyway, um, so the first thing we want to focus on is offensive force buildings because while we only need one, theoretically, probably a good idea to have, like, two cybernetic scores.

Maybe because then if you just have one and someone destroys it, you're in trouble.

But anyways mostly want to focus on the gateway.

So if the land of gateways is less than three I proposed we'd like to have one gateway per minute.

So at the nine minute mark would be nice to have nine gateways, right?

So, about once a minute, I'd like to have a new gateway.

So rather than being less than three, let's just say I want to know it.

Is that less than self dot It oration derision, divided by self dot It orations per minute.

So if, um and this is just this is super, um approximate.