Robot Algorithm.

[korexus]

Han, trewqh and I have been discussing the development of WoK: Duel quite a bit by email, but, as some other people are playing and hopefully even more are watching, we thought it would make sense to bring the ocnversation over here.

The first point is the algorithm for robot movement.

Currently, a robot starts in its home province and on turn 1 attacks the lowest numbered neighbour, converts 2 sets of POP to ARM in this new province, moves 5 armies back and sets the aim in both provinces to +MIS.
On future turns, the robot attacks from the province it took last turn to the lowest numbered legal target, then converts ARM, moves 5 ARM back and changes the AIM as before.

Now that duel is automated, the lowest numbered legal target is chosen as the turn runs, so it is possible for savvy players to redirect a robot during the turn by taking one of its provinces, thus allowing a new lowest numbered legal target.


There are two possible changes which we thought of. Any comments on them or offerings of other ideas are gratefull accepted.

1) The robots could prioritise who they attack. (So for example they attack the lowest numbered human owned target if it exists, if it doesn't then they take the lowest numbered neutral, if that doesn't exist either then they attack the lowest numbered robot owned province.) This amounts to a smal degree of cooperation between the robots against both players. (In the current test, all four robot players are pilling onto each other, meaning that they're not really a factor for the humans to worry about.)

2) Currently, if a robot's 'active province' is taken or if it fails to take a neutral it will stop doing anything else for the rest of the game. We could modify this with something along the lines of: If the robot's main attack force no longer exists, it attacks from the lowest numbered province which still has armies in. As it will probably only have 5 armies we could also let it missile the target province before hand. This really allows players to manipulate the robots to their advantage, missiling out the main attack force of a robot so that it attacks from an unexpected province or leaving missiles in the path of a robot player so that it's attack on an opponent is more effetcive than it would otherwise be.


There's quite a lot to take in there, especially for people who haven't played yet, but any thoughts would be great!


Chris.

[Mullog]

Interesting stuff!

But why make the robots so predictable? With a little more randomness we can use the same algorithm to run robot-players in regular wok-games as well.

I see the argument that the robots should not be the main competitors to the human players, but with some randomness it will be harder to exploit the robots behaviour and they will be more useful.

It sounds reasonable that a robot player will continue playing even if its 'active province' is captured. Maybe the robot should spend one turn doing one pop-arm transformation before performing new attacks?

I think we should focus on making it harder to exploit the robots, for that will take focus away from the main points of the duel - fighing your opponent.

[korexus]

I like the idea of an AI player in WoK games, that would be fun, but probably a way off yet. (And it would really stuggle with NAP requests...)

I think the idea of the robos here is that they are exploitable. Players can use them strategically to god advantage which makes the duel more skillful. If that became dependant on a random factor players would probably ignore them at which point they may as well not be there to begin with.


korexus.

[Hannibal]

I agree with Kor. My idea for the 4 Roboplayers was that they would NOT be serious (AI) competitors, but would provide a changing landscape for the player-on-player duel, an added compexity to plan for (rather than an added random element). They just make the geopolitical map more interesting! In game terms, you have choices as to whether to get out of their way, ignore them, or exploit them .... they ARE exploitable:

1) you can plan for their forward movements to shield you .... 2) they don't convert any POP in their start-prov, just send 5A back to defend it .... so, at some point, that becomes a more attractive target than the next neutral ..... 3) they run out of steam, since their eff gets eroded by no replenishment, such that, by around turn 4 or 5, they are leaving POP behind (they only try for 2 POP-Arm transforms), and only send 5A back to defend them; so, neutrals get harder and harder to capture, but left-behind Robo-provs don't add 2 extra armies per turn, nor .2 def, and also lose Eff - so they become easier to take than neutrals ...4) they attack out with all Armies, and only send back 5 defenders to the place they attacked from on "Moves" - so, you can always nick the place the "head" attacked out of, if you go later than him in the OOP, before he moves armies back to defend ..... Plus other opportunities!

So, IMHO, they are already interesting enough! No need to try harder to make them more interesting! Even being totally-predictable, they already add enough complexity and options ..... and are a skill-element without adding randomness to what they'll do.

In our 2nd or 3rd Beta-test, Duel #04, trewqh, Kor and I thought that maybe we'd generated a map where the Robos would be boring (hence the debate on a new algorithm for them); Not boring at all! trewqh just sprung a surprise on me by battling through a Robo to get to my rear, so the Robo was not a curtain ..... Me, I've dodged through where Robos would be about to cut me off from the main action, by spotting that I could get in and through before the Robo attacked .... and the other time by going through the prov that the Robo had just attacked out from, therefore undefended till "Move" phase. And I've taken two Robo-start-provs as easier targets than neutrals after turn 5 or so ... The Robo's have been tactically VERY interesting, not at all boring, even in this game where we thought they were maybe only up against each other .....

I reckon my original design for the Robos works and gets the balance right. You don't lose a lot of interest if two of them clash and stop attacking onwards, what they leave behind is an exploitable landscape!

Now, I know that (quoted by Kor above) I myself suggested a possible revision to the robo-algorithm .... that they would prioritise attacking humans before neutrals before other-Robos. But I'd now like to withdraw that suggestion. Duel #04 has shown that they are interesting factors as they stand. A problem with my idea of prioritising humans-before-neutrals-before-other Robos, is, now I think about it, that it can just make it TOO hard to get out of the way of the oncoming Robo ..... you'd spend all of Yellow's efforts trying to keep moving out of the prov that the Robo was going to attack .... for several turns in a row. Unbalances the game. The beauty of the current Robo-rule (always attack lowest-numbered adjacent prov), is that, even if your Yellow finds itself in the path, it can decamp to a HIGHER-numbered adjacent prov, and then just watch the Robo plod by on its lowest-numbered prov course. Which is GOOD, because it doesn't make Robo's take over the game, at least for Yellow expending all its turns to just get out of the way ....

SO, I offer 2 alternatives for discussion:
1) As original: Robos attack (with their leading "head") from their last-taken prov, to the adjacent prov with the lowest prov-number, regardless of whether human, neutral or another Robo.

or 2) A compromise between my original and my human-before-neutral-before-other-Robo idea (which I withdraw) : the compromise is that all trhe 4 Robo's are "allied if possible" !! ie. the Robo scans around, as to where his lead-prov with his leading armies is ....ie the last prov he took .... and then does only ONE step of prioritising: he treats any humans or neutrals as equally attractive, and then will only attack a fellow-Robo if there is no adjacent human or neutral. So, lowest-numbered adjacent prov that is human or neutral; if no such, then lowest numbered adjacent prov that is another Robo. Dead clear, easy (?) for the players to plan around .... But (for those of you who read this far): it means that a human can still "dodge-to-a-HIGHER-numbered prov to get out of the way of the plodding Robo", rather than spend turn after turn moving everything one more space away from the Robo's inevitable attack-after-next .....

So, what I've done/proposed here is keeping the Robo-algorithm very simple and predictable (not trying to give them AI and make them more active).

The Robo-algorithm is: (idea 2, or reduce back to original 1): The Robo's start with a start-prov with the same values as PLAYERS, not with the values of neutrals. They then always make ONE attack only; from their last-taken-prov (start-prov on T1); they attack with all armies; they attack the lowest-numbered-prov adjacent to that prov of theirs; except that, loosely allied with the other 3 Robos, they will prioritise the lowest human or neutral adjacent prov ahead of any Robo-controlled adjacent prov; they'll only attack a prov controlled (at start of turn) by another Roboplayer if they have no alternative human or neutral prov adjacent to attack instead; even if the prov number of the adjacent Robo is lower than the prov-number of an adjacent human or neutral prov. They prioritise humans/neutrals equally, ahead of other Robos.

The Robo also moves 5Armies back from the prov he attacked to the prov he attacked from; obviously voided if he failed to take the prov he attacked. Lastly, the Robo switches every prov he owns, captures or started with, over to AIM of missiles; he never fires any; they just build up and it just serves as an incentive for the humans to take a Robo's prov! Yep, it needs that incentive, to get the balance of options right for the humans ....

So, I've gone to great lengths to keep the Robos predictable, scary at their "head" ie leading-attacking edge, but vulnerable behind that, and increasingly vulnerable as their Eff drops, such that at some point they will become easier targets than the always-growing neutrals; a deliberate balance to give the 2 humans both predictability and options. Without the Duel game overbalancing into being TOO MUCH about the Robos .....To give them a role as "changing landscape" rather than as active players .....

Guys, this Duel game, after 2 and a half beta-tests, is ALREADY riveting and highly skilful, both tactics and strategy. Including dodging and/or making use of the "plodding" Robos .... Until you've tried it, don't even THINK about ways to make it even more complicated!! It's fully complicated enough with the the Robos being fairly simple and entirely predictable. Just try; you'll be glad that at least the Robos were predictable, for one or other of you to exploit!!

Cheers,
Han

PS Eagle Eyes has said he'd like to play Duel. Kor is busy; I'm taking a break; trewqh is unbeatable champion (subject to Duel #04 surprises), so would anyone else like to take Eagle Eyes on at it?

[korexus]

I'd say don't put any prioritising in at all. The Robos are quite fun as they are.

The only tweak I would suggest is that Robos select their targets when they step up during the OOP, not at the start of the turn. This would allow a player to take a robot's province and redirect it that turn, as opposed to the next turn. I really like this idea as the robos are just too easy to avoid right now...

I'm happy to run another game if Eagle Eyes wants to play.

Chris.

[Hannibal]

OK, done. I quite fancied the middle-course of the Robos being "slightly allied", ie. avoiding each other if there was a human or neutral alternative. But I'll go with your call on that. No prioritising, always lowest-numbered adjacent prov. - as per original design! This has the bonum [;-)] that it's easy to DESCRIBE in the rules, and easy for new players to remember - always next adjacent prov with lowest number. (I always like to keep rules simple, tactics complex).

Which brings me to your other point, ie your suggested tweak of the Robo choosing lowest-prov MID-TURN, at their point in the OOP, rather than being ruled by lowest-numbered-prov at the START of the turn. Hmmm. I'd MUCH prefer to keep it simple: you look at the map, you can see which way the Robos will go, without having to factor in/remember that the Robo only chooses the lowest AT its point in the OOP. For one thing, it's a whole extra paragraph in the rules! And people would forget it. And I'm sure that it would most often apply when Robos are bumping into each others' adjacent provs rather than from human redirection. So you'd constantly be adding that needless complexity only for Robo-on-Robo, effectively.

Y'see, I think you're overestimating the relevance of the humans being able to RE-DIRECT a Robo. IMHO. It'll virtually never happen! For one thing, you'd have to take the adjacent prov (to re-direct it into) BEFORE its turn in the OOp.....so 50%. Then, it would have to happen to be the LOWEST-numbered prov, or else he doesn't change course at all .... 33%? And then, you'd only do it if redirecting him to there actually helped you ........20%? Because his redirected onward-route is pretty unlikely to smash into your opponent .... only if the lowest-prov route then happened to hit him ..... and even then, it would prob hit one of his empties ...... and of course, that Robo's first move is INTO the prov you just took from him; it'll be a ANOTHER turn before tha Robo actually reaches your enemy!! And, given all thast, you're assuming that a normal player will actually spend ANY time looking at it as a possibility?? No, even on the rare case where it might actually work, it's 80% that the player would never even have considered it!

Net, it'll be a rare-or-never event. Definitely not worth complicating the rules for the remote possibility. Let's keep it simple where we can. Best if all nap-restrictions and robo-attacks are all governed by the provs owned AT THE START OF THE TURN, simple.

After all, as we've just seen in Duel #04, the Robos present an interesting ingredient even if no-one ever redirects them. For me, it's not a minus that they are "easy to avoid"; it's a plus. they just change the "map", and your options, giving you the need to get out of the way, and chances to exploit them. They already add complexity; and the game is already complex enough in the choices you have! Let's keep the rules as simple as poss, - it's just tactics that then get complex!

Cheers,
Han