Some simple AI improvements that don't need cpu power

Feature Requests and Game feedback
1

Hi,

I am sure you don’t hear for the first time, that the AI plays very weak and makes wrong/awful decisions most of the time. This results in decks being built using Death touch or other “destroy” skills that rely on rng instead of right decisions/tactical moves.

The limitations for the AI, as far as I read in the forums is the limited CPU power, as it has to be able to run on mobile devices without wiping their battery in 15 minutes. That’s ok.

So I thought a bit about how we could improve AI without using additional CPU and I came up with these ideas, that would cause almost no impact on calculation needs on the AI.

FIRST IMPROVEMENT:
There are decks, that rely on skull damage (high-attack-creatues) and decks that more or less ignore skull damage as they to their damage in another way (Mab, Elemaugrim, many AOE creatures are far more efficient when casting their spell instead of giving away move priority to 3 skulls).

What happens during play? The AI chooses more or less in this order:
5-match over 4-match over skulls over either attack skill OR a 3match that fills up an attack skill. Some defensive skills are rated lower than single 3-matches sometimes.

While this is a logical order to some point, as I said, some decks would prefer their skills to be launched when ready even OVER 3 skulls.
So what about a simple Radiobutton below the team (when choosing the defender team) that offers two choices as a hint to the AI, what the player would like the AI to do with the deck?
(o) Skulls over Skills
( ) Skills over Skulls

This means, WE can decide, whether skills should be launched instead of giving away priority to a 3-skull-match.
One additional “if…” statement at this point could be that the AI just checks whether those 3 skulls would kill an own creature if used by the opponent. That would be the point where the AI could overrule and use the skulls to avoid losing a creature.
This simple option would make a big difference in the “danger” of defending decks and it could result in a lot of way more interesting fights in pvp.

IMPROVEMENT TWO:
A second idea, maybe an extension to the first would be, that we can “prioritize” our creatures in the deck when building a defender team in terms of “which skills are the most important”
What I mean here is, I can put a priority from 1 to 4 on my 4 creatures that shall help the AI decide, which skill to launch first, if more than one is ready at any point in time.
And it also should impact which colors to match first if more than one 3-match is available (which is the case almost always).

Those two things could change the decisions of the AI drastically, even make is less predictable and it would also change the value of some cards (up and down, both directions would happen).

At the moment the AI only has the chance to get a full-traited mythic deck that can’t lose easily.
With these changes, we as players would have to take way more care against other decks and maybe a deck 2000 points stronger than your own would no longer be an auto-win for the lower deck just because it’s played by a human.

(My deck currently has 6k and I can’t remember when i lost my last match vs an 8k deck… I beated 8k decks while my deck was less than 5.5k easily).

What do you think about that?

Cheers, H

1 Like
2

I wish they sort out the cheating a.i in pvp and gw it a joke in some games i get 1 turn that is it every move a.i makes a 4 given him free turn for the whole game.

Then you have the soul dragon ability that might bring him back so far my soul dragon has died 23 times and still no re spawn and then in gw I had to kill the same dragon soul twice he gets back to back re spawns for dragon soul and I can’t get it once

3
4

Thanks for your replies but none of them answers the question, what you think about the ideas

5

  • i like spell priority amonth units

  • skull vs spell priority might be tricky (a little too rough ai if that is done properly)

6

Where did you get the impression that CPU utilization was the limiting factor in the AI’s complexity? The AI could be ten times as sophisticated in its decision-making and this would have an immeasurable impact on CPU activity or battery life. The AI runs for a few milliseconds every time the AI takes a turn. This is an insignificant fraction of the total processing time.

What does impact battery life:

  1. Ambient graphical effects (idle animations)
  2. Screen brightness (player-controlled)
  3. Heavy network utilization
  4. Persistent CPU operations (notably, not one-shot-and-idle)
  5. Heavy I/O operations (flash/disk operations)
  6. Ambient sound effects (though only slightly)
  7. Vibration/GPS/Gyroscope and other less-used peripherals
5 Likes
7

I read it in the forums.
But, if it is not the cpu, what else could be a reason for such a bad playing AI?
I mean… an AI for this game is not a complex task, I wrote AI’s for far complexer games that were light years ahead of the AI of this game in a few days.
And I just refuse to believe, that they have not the skills to write a good AI.
So, according to this and to some posts regarding AI-limitations “because of the mobile devices” the conclusion was logical, that it must be the cpu (at least for me).

8

Apple’s regulations concerning scripting. There have been many posts about this.

9

That’s a strange conclusion, but you’re free to believe what you like. My view is that the AI was performing acceptably given the relatively small investment they made in the algorithm, so they didn’t bother improving it. Limited resources means you focus on what has highest priority, and once the AI was suitable (if simple), there were other tasks that needed doing.

Besides, for a casual game like GoW, most people want to win more than they want a complex opponent. Considering how many people are convinced that the Console AI (which is only slightly more complex) is some sort of cheating doper, I think any substantial improvement to the AI’s planning around extra turns and the like will simply invite additional rage.

2 Likes