Morten Monrad PedersenDenmark
If there’s one type of story about game design, then it’s stories that give a peek behind the curtain of a solo mode development process. It gets even better, if that process is inspired by my “Automa Approach” .
Today we get just such a story from the 1-Player Guild’s very own Joseph Pilkus who’s designing a solo mode for Tradewars: Homeworlds. For me it was a very enjoyable peek behind Joe’s curtain. Hmmm… that might have come off wrong .
Before I embarrass myself even more, I’ll pass the keyboard to Joe.
Peeking behind Joe’s curtain
I’ve enjoyed solo-gaming for a long time…before we even discussed it as a board gaming topic, I had played countless games in a solitaire mode, even if they were never intended as such. More than 40 years ago, my father taught me Chess and from that point, my love affair with gaming began and manifested itself in many ways, from family games, to military war games, to RPGs, to finally, designer board games. Before introducing low-level characters to my party of adventurers, back when D&D 2nd Edition first came out, I would run the NPCs through a mini-adventure to get them some X.P. and a few items which would benefit them. No handing-out some random amount of X.P.s for me. While stationed in England from 1995-1999, I played more than 50 “solo” games of Axis & Allies, with a deck of cards I created for the Allies to focus their attention. In short, I love the solo experience and I want others to enjoy it, as well.
I would be remiss if I didn’t thank a number of people for the opportunity to share my AI development experience with all of you. First, Morten graciously invited me to provide a guest piece for his blog and I’m grateful to him for the invitation. I’ve followed his blog closely these past two years and enjoy the Automa for both Scythe (which I used effectively while serving as a play-tester for the Invaders from Afar expansion this time last year) and, of course, Viticulture, a game I adore (and both my girlfriend and my daughter love it immensely). To the designer, Kris Kycia, who reached out to me almost two-and-a-half years ago and asked me to be his developer for a deck-builder, a genre with which I had at the time, limited exposure. The publishing team consisting of Mike and Stan Strickland at Outer Limit Games, has been warmly receptive of my ideas. Finally, I owe a debt of gratitude to the fine members of the 1-Player community who communicated their desire to take our game to the next level by introducing an AI and truly give it the attention it deserved.
From the Beginning
From my earliest days engaging with Kris’ deck-builder, Tradewars ~ Homeworld (TWHW), I gravitated toward the Solo Scenario, The Derelict. It played fast (under 30 min, on average); it kept me on the edge of my seat; and the AI was by-and-large simple. Roll a die, check the chart, and apply the result. Since this was a scenario in which an unknown starship appeared in your Homeworld’s space with an intent to destroy everything, the AI didn’t need the “I” as the “A” would suffice. In time, I grew to love the scenario, playing it one week more than 50 times…all in the name of play-testing. Now, those of you out there who know anything about play-testing, “playing” and “play-testing” are light years apart, but there’s definitely an element of play-testing every time you play a game.
With a few vocal members from the 1-Player community articulating their desire for a more robust solitaire experience and a nod from my designer, I had all that I needed to go forth and develop TWHW’s AI. But what did that exactly mean? Moreover, how do I even begin? Well, in life you occasionally encounter those moments which are truly characterized by the adage “better luck than good management.” While driving to work and listening to one of a handful of podcasts, the latest one from Gabe Barrett’s Board Game Design Lab highlighted an interview with Morten himself. Not only was it a great interview, it answered many of the questions I had about the process and a few I didn’t even think to ask.
The one quote which I’ll paraphrase here, is to find the balance between “simplicity” and “simulation” when designing an AI. Thus, I can make a rather unintelligent AI in a matter of minutes as it will replicate nothing or by contrast, spend a year on an AI which simulates best the very action it should make, taking into account the current state of the decks, both its own and that of the player; the number of deployed ships; and myriad other facets too numerous to mention here. That was my invitation to begin the design: start with everything and shed from it those areas inconsequential to the overall design, leaving behind a streamlined AI. My David was in the hunk of marble. All I needed to do was chisel away that which wasn’t David.
When you strip away all of the fancy illustrations, curious hieroglyphic-like iconography, and flavor text from all deck-builders you’re left with mechanics. In essence, what do you do and how do you do it? Mechanics don’t care about victory points. Mechanics don’t care what you just did or what you plan to do during a future turn. It cares, desperately about now…and only now. So, for me in this Zen state was to perceive the “now” in a way that I hadn’t really looked at it before. In TWHW, players operate in three broad areas: Select a role and take the corresponding Action; Buy cards from a set of different piles; and Configure Starships to battle others and attempt to destroy their Homeworlds. Now, I had to reduce these broad areas into their constituent parts and determine what I needed to tenaciously hold on to and what I could easily abandon.
First, however, removing decision-making from the player was paramount. But, the AI still needs to make decisions, which may be made manifest through die rolls, a spinner, a deck of cards or possibly other ways to randomly generate sensible decisions...and that's the trick, you want the AI to remain intelligent. I thought about dice and a series of matrices, taking me back to my flowchart days, but that would prove cumbersome. I mentioned a spinner just for laughs...moving on. Finally, a deck of cards seemed to be the most elegant solution so that's the direction I took the development process. So, we'll now look at the three areas affected by these decks...the Action, Buy, and Configure Starship phases.
Actions are the life blood of most deck-builders, from Dominion to Thunderstone to Ascension. In the case of TWHW, the cards don’t dictate what you’ll do, the player does, ala Puerto Rico and other board games. Thus, I needed to have the cards emulate this level of control. But, not all roles are created equally...as the Chancellor focuses on the various decks (Starship Shipyard, Military Academy, etc.), I thought it best to simply remove them entirely from the game. Also, the Espionage Agent (Spy) relies on deciding what to shed from the hand of cards, and as the AI has no hand of cards, I removed this role from consideration, as well.
As to the Buy Phase, with the elimination of Draw Decks from which to make purchases, the player need not make decisions about which Starship, Weapon, Crew, or Tactic the AI will add to their complement of cards. Instead, the cards themselves would provide information on both the activated Action and the purchased cards from which to Configure Starships. In the end, I designed a deck of 18 cards which provide a Configured Starship at the top of the card and the bottom half is dedicated to the possible Actions...either those with dependencies (Admiral, Captain, Commander, and Engineer) and those with no dependencies (Treasurer, Smuggler, Saboteur, and Trader).
Play testing or "Are you having fun at work today?"
During the first few rounds of play testing, I realized a few things which I needed to change. First and foremost, I had included basic level configured starships with "1-3" level Weapons and Crew along with "4-5" level ones, which threw off the balance. The amount of credits added to the AI's Treasury, however remained at parity with the amount I had garnered, so I knew that part worked exceptionally well, but needed to change-up the complement of weak and strong starships. Enter, a second deck. I retained the 18 cards and had modified some of the configured starship values, but now had two different decks, an Initial deck consisting of 12 cards and a Development deck consisting of 6 cards. The player begins the game by completing their Action, Buy, and Configure Starship Phases, and then reveals the topmost card from the AI's Initial deck and follows the instructions on the card. This continues throughout the game. When the player must shuffle the AI's discard pile (now consisting of all 12 cards), he adds to the mix the top two cards from the AI's Development deck.
While play testing shall clearly continue in the weeks and months ahead, allow me to reveal some of the findings from more than a dozen games, thus far.
- The AI's Treasury continues to grow at a pace consistent with the player
- The AI's Space Lane fills at a rate nearly equal to that of the player
- Battles between and among the player's and AI's starships are tense
- The AI has claimed nearly 60% of the victories
Those designers planning on creating an AI for their game should prepare themselves for a lot of work and at times, a lot of math, in pursuit of maintaining a delicate balance. It's the most fun I've ever had with math as my background is firmly rooted in history, international relations, and foreign language. I'm excited by what's been developed so far, I look forward to getting it into the hands of very knowledgeable folks in the deck-building community; and providing the 1-Player Guild with a fantastic AI to further enjoy Tradewars ~ Homeworld. I certainly welcome your thoughts and inquiries.
Taking back the keyboard
Morten here, taking back control of the keyboard.
Before I let all of you get to the comment section I will mention that Tradewars: Homeworlds is currently on Kickstarter. It’s at well into the stretch goals and has 4 days left.
A blog about solitaire games and how to design them. I'm your host, Morten, co-designer of solo modes for games such as Scythe, Gaia Project and Viticulture.
A glimpse behind the curtain of AI development – guest post by Joseph Pilkus designer of an AI for Tradewars: Homeworlds
26 May 2017
- [+] Dice rolls