<< Back to Warzone Classic Forum   Search

Posts 21 - 39 of 39   <<Prev   1  2  
Something Different: 10/20/2015 20:08:11

inquisitor
Level 56
Report
I want there to be an AI difficulty setting.


Nice idea, but I doubt it will happen. Skip the programming part. Try to write a complete step-by-step instruction to tell the AI how to play all its move during a game. It can't be vague. The instructions must be 100% clear and exact. It is going to a big project.

Fizzer made the single player version to teach new players how to play, and help them to familiar with the gameplay. It is where the AIs are mostly for. I don't think he will invest a lot of time to improve the AI.

Having said that, there is a Warlight AI coding competition.
http://theaigames.com/competitions/warlight-ai-challenge-2

There are some bots which can play decently. Maybe Fizzer could contact a few coders and see if it is possible to port some good bots to Warlight, but I doubt he would do.

If there is never anyone to play, then how do I practice and get better? Well, AI difficulty in Single Player.


Only if the AI improves a lot and can really compete with good players.

Currently AI is a noobie who makes tons of silly mistakes. You will learn very bad habits if you try to learn by playing with AI. It does not help you to get better at all.
Something Different: 10/20/2015 20:38:12

wct
Level 56
Report
Second link: https://www.warlight.net/MultiPlayer?GameID=9684185

This time Riryo will start as a team mate so he can see my playing if he wants.

[Edit: slight screw-up. New link updated.]

Edited 10/20/2015 20:42:18
Something Different: 10/20/2015 20:54:13

wct
Level 56
Report
Inquisitor, I'll join yours after I finish Riryo's. I liked your idea of changing the humans to turn into AIs, so I tried to modify the second of Riryo's to do that, but then I realized he would become an AI ally, so I scrapped that idea and just kept him as a surrendering team mate, so he can watch if he wants.

Screenshot for game #1, turn 17. Soon to complete South Dakota:

Something Different: 10/21/2015 00:17:16

wct
Level 56
Report
Turn 28:




I've encountered a strong AI opponent (pink). Presumably this is what Riryo is thinking of, that it's a roll of the dice which player will get the best starting position with few enemies nearby and a headstart on expansion.

If that's the case, then this will be a good example of how to win despite an AI headstart, since I started in the middle of a clusterfuck of AI opponents and expansion was slow and tedious at first.

Indeed, the strong AI is threatening some of my bonuses, and this does indeed make the game slower and more tedious, but at this point it's really just a matter of time before I manage to crank up my income to be on an even level with the strongest AI. The AI simply is not smart enough to break my bonuses and keep them broken, whereas a human can easily do this to an AI, using the right techniques.
Something Different: 10/21/2015 06:15:17

wct
Level 56
Report
The second game started a bit easier. Here's turn 26:

Something Different: 10/21/2015 12:11:49


powerpos
Level 50
Report
[Wolf]Orthrus Echo Five said:
I'm not the best at WarLight, so I like to play more experienced players, to get practice and become better at the game.


Then don't play against AIs, they won't improve your skill after you learned the basics.
Secondly, get your bootrate under 15%, nobody wants to play against someone who is not even decent enough to surrender when they 've lost, so no surprise you're not seeing any games.
Thirdly, get your 1vs1 winrate up, auto-games should be good enough for this,
this will allow you to play games against better players.

TL;DR: You want skillz you need to put in effort.
Something Different: 10/22/2015 23:02:28

wct
Level 56
Report
Second game, turn 31:



I decided to end it here, since the rest is just tedious steamrollering to the map borders. Here's the game link again, so you can check for yourself that there's no real competition left (largest competitor has a small fraction of my income): https://www.warlight.net/MultiPlayer?GameID=9684185

Edited 10/22/2015 23:25:05
Something Different: 10/22/2015 23:59:24


DW: Soz, NGL, I Play SLOW. UV BN Warned! 
Level 57
Report
:O

Awesome job, wct. :)

But since I play a fair amount of single-player,
and we are unlikely to get a more difficult AI to play against,

I'm kinda glad that I can't pwn them as well as you do,
because then I'd be bored on single-player. :-/
Something Different: 10/23/2015 00:04:14

wct
Level 56
Report
Here's inquisitor's 39 AI challenge, turn 24:



This one will be easy, too, I'm already at 141 income and haven't hit any strong AIs yet. I will probably finish this one first before trying to finish the first of Riryo's games, which is the hardest of the three.
Something Different: 10/23/2015 00:14:22

wct
Level 56
Report
Dublin Warrior, as I said earlier on, the trick is to play against teams of AIs to increase the 'difficulty setting'. I'm only able to beat so many AIs because they are unteamed, and they all tend to fight amongst themselves. And AIs tend to be pretty bad at eliminating other AIs, so they have a hard time gaining and maintaining bonuses and super-bonuses with all these piddly little 5-income enemy AIs nibbling away at and within their borders.

When two teamed AIs gain a shared border, that's when they have a chance to put up a much stronger fight, as their effective border strength is essentially doubled. And that's not even mentioning the fact that the base income of the team is double the human player's income for teams of 2 AIs each.
Something Different: 10/24/2015 06:27:42

wct
Level 56
Report
inquisitor's 39 AI challenge turn 39:



There are two major competitors, Bronze on the East, and Sea Green on the West. Both have approximately the same income as I do, perhaps more. My strategy is to keep the big AIs' stacks relatively isolated, while I swarm around them and into their bonus territory. Meanwhile, I'm steam-rolling into the weaker AIs' bonuses to gain income. The AI is very predictable, and this allows me to use just-enough defenders to keep my territories, with just-enough surviving attackers left over (after delay moves) to steal neighbouring AIs' territories. It's a bit tedious, but it's also pretty much inevitable to win.

At this point, I'm steam-rolling my way South to complete Arkansas, Mississippi, and Louisiana, at which point I should be in a good enough position to start beating back the big competitors.
Something Different: 10/24/2015 06:57:46


Sephiroth
Level 61
Report
I play a ton of Real-Time games. They are my favorite to play, but I always see more Multi-Day games on the front page. I'm not the best at WarLight, so I like to play more experienced players, to get practice and become better at the game. If there is never anyone to play, then how do I practice and get better?
join the real time ladder and play 1v1 auto games
Something Different: 10/24/2015 07:11:48


Master Ryiro 
Level 63
Report
well done man!good job
Something Different: 10/24/2015 19:36:38

M. Poireau 
Level 57
Report
My favourite way to increase AIs' strength (if you want it to be 1-on-1):

* Start each player with a Sanctions card, and set the duration to something very long (e.g. 500 turns). The AI doesn't use them; you will play your own on yourself. Now that AI has a significant production advantage against you and is a serious opponent.

You can also:

* Make Reinforcement cards very common. However, make a rule for yourself to never use them. (So only the AI has the advantage.)
Something Different: 10/27/2015 09:05:37

wct
Level 56
Report
I decided to end inquisitor's 39 AI challenge when it became clear that the two big competitors were no longer able to grow or defend themselves from my encroachment. Here's the link again so you can see for yourself: https://www.warlight.net/MultiPlayer?GameID=9690458

Final turn (51):


Income history:


The end-game gets exceedingly tedious, especially on such a large map, but you just essentially have to wear the big competitors down, while keeping their big stacks isolated and constantly shrinking. For a sanity check, the last several turns took about an hour and a half to complete each one. In the 50-ish total turns, I made 11,650 orders (233 per turn on average): about 4250 deploys, 4087 transfers (incl. delay moves), and only 3313 attacks. The two big competitors only made a grand total of 5,364 and 3,876 orders each, about half as much or less than mine.

And, as it turns out, I screwed up the first of Riryo's games by getting booted (forgot it wasn't single-player and was focused on finishing inquisitor's first). I think I'll let that one count as a loss, and not bother to start a new one, since Riryo said 2 out of 3 would be sufficient. I might have been able to win it in the long run, but as you can see how tedious these games can get, it would probably have taken a great deal of time to dig myself out of that income disadvantage. Here's the link: https://www.warlight.net/MultiPlayer?GameID=9683128

(In that game, there was a budding competitor on the East coast, who would have eventually made contact with the big competitor on the West. If I were able to survive that initial wave from the East, I would have been able to let those two duke it out while I steal away their territories bit by bit until my income got up to their levels. At that point, I would have been able to win. But again, it looks like that game would have taken several days of tedious play to get there. I think I'll be satisfied with just a 2 out of 3 record.)
Something Different: 10/27/2015 09:34:50

wct
Level 56
Report
One thing you can do to eliminate the end-game tediousness is to play with an AI team-mate, and once you have dominated the income, like the two games above, you let your AI team-mate grow until it's safe enough for you to surrender and let your team-mate take over to 'clean up' the end-game. You can help this along with gift cards so that before you surrender you can give your team-mate a head start in gobbling up your territories, to ensure some other AI doesn't get them first.

If you want to keep the difficulty even, then play in teams of 2. If you want to increase difficulty, keep your team at 2, but increase the size of the other teams: 3, 4, 5, etc.

One nice thing about this is that you'll learn how the AI works when it's playing as part of a team. It can be really frustrating until you figure out its quirks, but once you've figured it out, it's pretty straightforward.
Something Different: 10/27/2015 19:01:25

inquisitor
Level 56
Report
Well done, wct.

I can see how tedious it can be to play with AIs in such a very big map.

Personally I don't think it is joyful to match against AIs. They are far too stupid. They are very predictable if you take some time to observe their behavior. Then you can exploit their stupidity and make full use of it.
Something Different: 10/27/2015 19:08:58


TBest 
Level 60
Report
One thing you can do to eliminate the end-game tediousness is to play with an AI team-mate, and once you have dominated the income, like the two games above, you let your AI team-mate grow until it's safe enough for you to surrender and let your team-mate take over to 'clean up' the end-game. You can help this along with gift cards so that before you surrender you can give your team-mate a head start in gobbling up your territories, to ensure some other AI doesn't get them first.


Do NOT do this. If you surrender the AI will VTE. (Exception, in a tourney a rnd winner is picked)
You can however give your land to your AI, and commit "blank" turns were you only deploy. But you need to be alive. {This applies to multiplayer}

Oh, and good job wct. I have done a similar challenge before and lost my patience way earlier.

Edited 10/27/2015 19:10:35
Something Different: 10/28/2015 04:29:38

wct
Level 56
Report
Do NOT do this. If you surrender the AI will VTE. (Exception, in a tourney a rnd winner is picked)

I think that may only be the case with multi-player games. In single-player, you definitely can have AIs playing against AIs until the end, as I've done it many times. Just play around with the settings until it works the way you want. It is definitely workable.

There are some maps which are better than others for this kind of game. I used to play Land of Kalador (https://www.warlight.net/Map/9071-Land-Kalador) quite a bit because different areas on the map are structured differently, so the game-play keeps changing as you move from one area to the next. Also, with random starting positions, you can find yourself with different challenges in the opening-game each time you play. Issander's Huge World (https://www.warlight.net/Map/6496-Issanders-Huge-World) is also good like that.
Posts 21 - 39 of 39   <<Prev   1  2