The SP experience (aka the CPU opponent)

[Bullman]

It is understood that this PC game is an adaptation of a board game with a set of rules to be played and understood by two players playing head to head. There are at least two other ways players can play each other using a digital platform/interface. The one thing that NaWD has that none of the other platforms have is a SP CPU opponent. This is great for those who want to just play and experience the game whenever they want without having to rely of finding/organising/waiting on a human opponent.

However I will pull no punches in commenting on the CPU opponent that comes with this digital version of NaW. It would rank it as one of (if not the most) laughable CPU opponents I have ever experienced for a game of its type, so much so that I can only wonder why the developer even bothered. A CPU opponent as bad as I have experienced should never form part of a professionally released PC game title.

I support my claim by showing you screenshots of the moves played out by the CPU opponent when I randomly decided to try the scenario "A Little Tea Before We Fight" vs the CPU. This is a exit mission for the German forces that are trying to exit the western/left side of the map. I have attached some examples of the moves plotted by the CPU opponent. They are essentially moving in circles.

I understand that most CPU opponents for games like this are best played when they are on the defence, but still, the moves you see plotted by the AI is laughable. Even simple orders to just charge in a straight line towards to western/left side map edge would be better than doing circles in the middle of the map.

I really don't expect much from any CPU opponent in any wargame, as I normally get games like this to primarily play multiplayer (and avoid any of the usual shortfalls developers have in coding a CPU opponent for SP in games like this), but it has got me wondering how much the developers really care about the SP aspect of board game adaptations like this one. Do they really think their game is best showcased when playing it SP? Or do they acknowledge that the only real challenge a player will ever possibly face is in a MP game?

Designing a compelling board game based system (like NaW) is one thing. Making it in to a digital version with a SP aspect to it (and hence a coded CPU opponent) is a completely different unrelated thing needing a completely different set of skills and experience.

I really hope the developer can comment here, as I know this is not the first and last time I have ever encountered a similar thing in a game like this.

 

[Daniel Karsai]

Admittedly, the AI (Artificial Intelligence, aka the CPU opponent) was tested playing against itself prior to launch, and I couldn't watch its every move (several simulations were running in parallel while i worked on something else). This was good enough to determine the AI can win against itself on either side in each scenario, but not good to track down nonsensical actions or to measure up against a real human, so I should have been more thorough with it.

What you have in the screenshots boils down to one thing: An "imbalance of tactical value". The computer wants to exit, but doesn't want to risk the units dying, since they need to reach the edge. For each possible action, it will generate a (really inaccurace) number to determine that action's value. It goes wrong when it chooses self preservation halfway through approaching the exit, because more units came into line of sight, or (close) attack range. (it thinks the safest thing is to increase distance, turning it around)

I just looked at the code and the AI completely ignores if the enemy units are ops complete (unable to opportunity fire), so this is a bug I can fix soon-ish. The question remains, will the units be too exposed in the next turn, when the ops complete markers go away? That is where it gets really complicated from a programming point of view, though I think it's more fun to fight an AI that's foolishly brave rather than overly cautious, so it should be an improvement.

It's challenging to track down problems with the AI, especially those that only happen on occasion. There's a lot of information required to figure out the cause, and playing the game over and over hoping to catch it is really time demanding. (it's really easy to burn out too, when you're doing it for work rather than pleasure)

I appreciate you bringing this issue to my attention, it really helps me solve it.

 

[Bullman]

I fully understand the challenge of coding a CPU opponent for a game like this, let alone any other similar game that has a board game heritage. It is relatively straight forward and easy for humans to read and understand the rules to a board game, learn the game and play it against each other with the knowledge that their opponent is just as likely to make rational and educated decisions throughout the game to improve their chances of winning, because that is what humans kind of naturally are cut out to do. Not so straight forward if you are expecting to code a CPU opponent to essentially "stand in" for a human player. So many decisions, so many factors to consider that have to be turned in to complex coding algorithms.

I do agree that a more aggressive AI is favourable over one that is so overly cautious, it is incapable of completing its objectives. Its good to know you are still looking at improving the AI and I am happy to assist.