ATTACK OF THE BOTS (by Tom Russell)

Most of our games are for two players, and like a lot of two-player wargames, there's relatively little to prevent a gamer from playing both sides when he lacks an opponent. One of the weirdest quirks of wargaming is that there's a long and storied tradition of solitaire play, a tradition that looks positively bizarre when viewed from the outside. Heck, when I first got into this side of the hobby, I thought it sounded absolutely bonkers and solipsistic. I soon learned better, and there are times when I even prefer solitaire play, particularly when I'm first learning a game. And I'm not alone in that. So it makes good business sense for Mary and I to publish two-player games that are suitable for solitaire play, including solo-only games like my Agricola and Charlemagne, and several other solo titles we have planned for 2018. 

We also of course have published a few games that seat more than two. An Infamous Traffic plays with up to five, For-Ex with up to six, and Dynasty seats either three or four players. And one thing that people have asked us over the last year is if we're going to release a dedicated solitaire mode for these games, or create a series of AI "bots" to facilitate a smoother solitaire experience for these multiplayer games. Several multiplayer wargames - particularly asymmetrical ones like the COIN series - now utilize some rather sophisticated flow-charts to manage non-player factions; why don't we do that here at Hollandspiele? 

It is, I suppose, a fair question. There are gamers who are interested in this title or that one, but they'll never have the three-plus players they need to make it work, and so they cannot justify the purchase; if there was a bot for the game, however, then they could play the game and would purchase it. We'd have made one more sale, and they'd have one more game in their collection - in my book, that's a win-win. 

The problem is that the type of multiplayer games that we've published don't really lend themselves to this approach. Bots work best in asymmetrical games, in which each player (or bot) takes on a specific player faction with specific goals. The bot and its flowcharts simulate a specific playstyle that likely maps reasonably well to what that faction might do if controlled by a human player. 

But all three of the multiplayer games we've published are largely symmetrical (though Dynasty is also asymmetrical given its Emperor/Governor split). It's not about the interactions of different factions, but of different players, with different play-styles and personalities. For-Ex and An Infamous Traffic especially function largely as studies in multiplayer dynamics, ripe with possibilities for emergent and temporary alliances in which each player's business gets hopelessly entangled with another's. I'm going to help Bob here until it's in my interest not to help Bob, and meanwhile I'm leeching off of Susan, who is in turn applying leverage to Phineas, who in his desperation to catch up is doing exactly what Bob wants as well.

It's difficult enough for a human being to successfully navigate that space; I don't know how on earth a flowchart could ever provide even the barest facsimile of those interactions, especially given the complexity of the decision space. How many if-thens would it take, for example, for a bot to make a decision in For-Ex, given that there are twenty-one different currency pairs, and that the future contracts are fairly open-ended - having ten possible trades for each currency pair? But beyond the hopeless complexity of that game, there's still the matter of the essential relationships between the players, and the inability of a bot to model that. 

Let's look at my Winsome game Northern Pacific, which is all about those relationships, and little else. The game is often described as "cube or train", in that the player has two options available. Cube: they can place an investment cube in a city; each city can hold a certain number of cubes depending on player count. Train: they can place a track to connect two cities; the train is largely unidirectional, always branching off the last placement. When a train arrives in a city, the player gets double their investment. The player with the most cubes when the last track is placed wins the game. 

This is a game about incentives and turn order: if I do this, then he'll do that, and she'll do that, and then this will happen, unless she does this instead, in which case, et cetera and so on and so forth. So though the game is very, very simple - perhaps irreducibly so - I think there's more to think about than is typical of a filler game.  So: how would a bot make decisions in that space? How can a flowchart recognize the tempo of the game, and anticipate that if he does that she will do this and then this? I don't think it can, even with the simplest possible game of this type. 

And that's okay. I don't think every game needs to be solo-friendly. That might mean that you can't get it on your table, and that it's not the right game for you - and that's okay too. There are lots of games that I'd love to have but that realistically aren't right for the people I play games with and the time I have available. And so I don't buy them. 

And again, I want to emphasize that I'm not anti-bot; I for one welcome our new flowchart overlords. They're completely appropriate for certain types of games, particularly games with defined asymmetrical factions. But when the decision space is less about modeling the interplay between the government, organized crime, and rebels, and more about open-ended opportunities for collusion and sabotage, bots aren't really a viable solution.