Forums Tyrant

Very simple "AI control" suggestion.

46 posts

Flag Post

Ok, we all know that AI plays dumb.

Like, super-dumb. I’m convinced it simply chooses at random out of 3 cards in his hand.

We also know that “true” Ai is difficult to code. However, I think there is something that is rather simple to implement, works much better then current (let’s be frank – hopeless!) system, and gives players a feeling of control (and further aspect of “strategy”).

We all know how a deck design field looks like, right?
COMMANDER -- 10 full-card-size slots:

/\ /\ /\ /\ /\
\/ \/ \/ \/ \/
/\ /\ /\ /\ /\
\/ \/ \/ \/ \/

Where <> is a full card.

Now, you can safely put a dropdown “choose a number from 1 to 10” menu at bottom right (for example) of each card in the same window (disregard dots, these are for formatting only):

/\. /\. /\. /\. /\
\X \X \X \X \X
/\. /\. /\. /\. /\
\X \X \X \X \X

it should actually be X(1), X(2)… X(10).

Now in that dropdown menu we choose priority “group” for each given card. Why “Group”? Because if it’s plain 1-to-10, then cards N9 and N10 are bound to always be played last (with 3 cars in hand), which is not necessarily desirable.

So with this system, the end screen could look like:

/\. /\. /\ /\. /\
\2 \1 \2 \3 \2
/\. /\. /\ /\. /\
\3 \3 \2 \4 \2

What happens next? In the game, AI checks if there is a card froum “priority 1” in his hand. If it is, it always gets to be played first.

Then (or if none is) he checks if there are 1 or more cards from group “priority 2”. He then plays one of those at random (or whichever algorithm is already used).

Repeat each turn with new draws. If there are neither 1 nor 2 cards present, something from group 3 gets played.

You can see, in my example, only 1 card is marked as “group 4” which 100% guarantees it’s played last in the game.

However, because there are full 10 groups for players to choose from, this system can incorporate anything from the current “anything at random” system (just assign everything the same group, say, default “group 10”) as well as “strict hierarchy” (1-to-10) system of earlier “Controlled AI” suggestions.

For ease of learning, the whole dropdown-hierarchy-number-choice can be hidden behind some “advanced editing” toggle in the same deck construction menu the game has now.
Then if player uses “simple” (exactly as now) everything gets assigned to the same “group 10” by default and plays the same as now. The moment he clicks “advanced”, player gets access to this fine tune.

And to top it all off, the whole coding this system will require is truly affordable. Sorting algorithm with pre-determined priority levels plus dropdown menu interface? no rocket science, for sure.

All in all, I really hope this could give the game a serious gameplay improvement for “hardcore” players without interfering with learning curve (players dont’ need to learn anything new at all, compared to current, before they click “advanced”) and without too much effort on part of developers.

 
Flag Post

Big thumbs up on this one. The uncontrolled decks certainly make it straightforward for people to rise in arena rank, but it takes a great deal away from enjoying faction wars. Arcanis’s suggestion sounds very easy to implement and is completely transparent for casual players who don’t care about their defense deck.

 
Flag Post

Very very big support from me. Excellent idea and, speaking as a coder myself, it is very easy to implement. No excuse for this not to be put in game.

 
Flag Post

+1 to arcanis for this awesome suggestion. The AI decks are fail right now so essentially faction wars become who has more energy to spend. With this, defense will finally serve a purpose.

 
Flag Post

Great idea

 
Flag Post

I like this a lot. Hope it gets implemented or at least seriously considered.

 
Flag Post

This is an excellent suggestion. Anything to make defense a true factor in faction wars would be greatly appreciated, but I have a feeling that the designers prefer offense having the advantage because it makes war bonds energy refills more important than defensive deck design.

 
Flag Post

The issue is, while you will of course value one card over another, it’s also got to do with what you’ve got in play. No point in dropping a four-wait guy infront of something already active, but on the other hand, if I have two more guys out than my opponent anyway, I don’t want the computer to prioritize Tiamat over Titan.

 
Flag Post

I had a small idea to add…should the auto mode for players be given this improved A.I. too? Since it appears to use the same code…only asking cause I agree with everyone else and hope they use this but it might make the auto mode useless

 
Flag Post

Solace: Of course it’s not a fool-proof decision making system. It’s just more effective than a completely random AI, and very easy to code. If your defense deck relies on cards that are situational, it’s going to be a crappy defense deck either way :P

 
Flag Post

Seems like a pretty good idea, developers should at least consider something along these lines.

 
Flag Post

maybe also adding an algorithm that the AI plays anti air(if in the hand ofc) if 50% or more of the players cards on the field are air units. also saying the AI only uses offensive action cards if player has 2+ cards on the field. i write code so i know this isn’t hard to do

 
Flag Post
Originally posted by dkblap:

maybe also adding an algorithm that the AI plays anti air(if in the hand ofc) if 50% or more of the players cards on the field are air units. also saying the AI only uses offensive action cards if player has 2+ cards on the field. i write code so i know this isn’t hard to do

Which would make AI holding his Daemons in hand and stuff…

Guys, it’s not a fool-proof system nor true AI nor nothing, basically. It’s just a quick and dirty way to both fix the obvious Artifical Idiocy which causes too much laugh all around and give players even more “control elements” aka gameplay – all without much ado.
Trying to make the “actual AI” is way more hassle, not even in code but in logic: it’s difficult to invent the whole set of adequate rules (such as that “play AA cards only if the field is air” or “play EMP only if the field contains 2+ creatures” etc – notice how both these rules are by far not foolproof!). Besides, wouldn’t you like to see how AI does what YOU – not devs, not RND, yes YOU – told it to?

Originally posted by whoownswho:

I had a small idea to add…should the auto mode for players be given this improved A.I. too? Since it appears to use the same code…only asking cause I agree with everyone else and hope they use this but it might make the auto mode useless

Sure! The same system with the same setup/interface/everything can obviously used for autoplay on your part. Matter of fact, before posting this, I actually played some games “emulating” such a system on offense (e.g. never thinking about board position but following pre-determined hierarchy groups I wrote for myself on a paper, only had to resolve ties in one group by actual dice roll, tnx my old stockpiles;). Not only it worked, it told me some interesting things about my deck.

Huge thanks for your support, too.

 
Flag Post

Shameless bump :P

 
Flag Post

Support!

Until then, I’ll be building <card> x10 decks

 
Flag Post

Seeing how people are happy with way more inferior (when not that much simplier in execution) solution, this definitely needs a bump.

 
Flag Post

+1
and bump

 
Flag Post
Originally posted by ArcanisX:

Bad variant of my earlier sug
http://www.kongregate.com/forums/65-tyrant/topics/152101-very-simple-ai-control-suggestion

But for people who absolutely, in o way can read (only write, common clause on teh internetz) that’s good enough.

your suggestion wasnt earlier it was made 2 and half weeks later actually.

Your suggestion was made after, is different, and is more complex. so id call your suggestion a complex variation of the earlier one (the one that me and some others came up with in chat). The complexity of your idea would be more time consuming for the developers to implement effectively for the missions ai.

I dont even know what your trying to say in that last sentence (tho from what i can make out seems rather ironic).

.
.
Orignal Ai improvement Petition below.
http://www.kongregate.com/forums/65-tyrant/topics/147580-ai-improvement-petition-sign-here?page=1
(ai needs some improvement we can all agree on that, im sure the developers can come up a few different options. so sign just to show you want to see change)

 
Flag Post

A+ … Suggestion is a lot simpler than it looks.

War Tyrant’s worthless AI drains the enthusiasm and stamina of its player base. Please divert energies from designing new promo cards or cardsets and implement an AI system such as this one.

Other thoughts:
1. If I could only look at one additional variable in the game state, I believe that air units / armor piercing / fear / mimic / etc. would be less relevant than simply (# of opp cards in play) – (# of my cards in play).

2. What are your reasons for having number groupings (1-1-1-2-3) vs simply numbering 1-10? The former system is more random, but a deck prioritized as 1-1-1-2-3 is always equally likely to play out 1-2-3-4-5 as any other allowed play order. The only advantage I could think of for number groupings stems from players being less able to exploit lack of randomness during rematches, rather than improving the AI’s first-game performance.

 
Flag Post
Originally posted by BrokenFixer:

2. What are your reasons for having number groupings (1-1-1-2-3) vs simply numbering 1-10? The former system is more random, but a deck prioritized as 1-1-1-2-3 is always equally likely to play out 1-2-3-4-5 as any other allowed play order. The only advantage I could think of for number groupings stems from players being less able to exploit lack of randomness during rematches, rather than improving the AI’s first-game performance.

Originally posted by ArcanisX:

because there are full 10 groups for players to choose from, this system can incorporate anything from the current “anything at random” system (just assign everything the same group, say, default “group 10”) as well as “strict hierarchy” (1-to-10) system of earlier “Controlled AI” suggestions.

 
Flag Post

Even though I find your idea extremely interesting and well thought through in terms of how to make it attractive to players, I think IT adds little value to the game, and that at a hardcore level it would even decrease the defenders chances. My points:

Your system applies to Auto attack decks and Defense decks in war and arena (If I am understanding everything right)

In auto attack it’s great, it makes your auto deck even more efficient.

In defense, however, and if we are talking about hardcore players, it just makes decks more predictable (aka weaker). In the long run, after one or two fights vs a deck it will be somewhat easy to “guess” what’s more likely to come up next (or come up reasonably early). With decks as small as 10 cards, trends of in which order players tend to put cards in play with this sort of system would become very easy to identify, adjust to and transmit to your faction mates, and I think that is detrimental for the game.

I’d rather see real pvp like we see in Kongai than this =( I was gonna post my idea for it, but it is something that doesn’t belong in this thread I think.

Kudos for your proactivity Arcanis. (proactiveness? damn it!)

 
Flag Post

Well, for predicatability, keep in mind that unlike in “direct” 1-to-10 card ranking system, my one still allows you, as a deck builder/ algorithm maker, to allow as much randomness in as you’d like.
For one, you can simply put all your small creatures in group 1 and all your big ones in group 2. That way, there sill still be tons of randomness in each game, but at least AI won’t succumb to DoA by playing cards that are bad openers first when he got good openers in hand.

Oh, and once people figure your def algorithm, you can always notice it (by decrease in defwins) and, for example, assign a different one. To the very same deck. End result? More variability (aka depth of gameplay) with the sme cardpool.

 
Flag Post

I had been thinking of direct 1-10 card ranking based on how you order your cards on the deck design field, but thought it would end up too predictable. I like your idea of prioritizing better, but thought it would even work if it went one step simpler. How about a simple toggle for each card on the deck design field for whether you want it high priority or not? If it is higher priority then it will get drawn before the lower priority cards… and if two cards in a hand are of the same priority then the current AI will decide which card comes out first. Its still basically your idea Arcanis, but might be an even simpler interface.

 
Flag Post

Really good idea, shouldn’t be that hard to code it and it would prevent f. example throwing EMP as an opening card by that so-called-AI ;)

 
Flag Post

Signed