In a few days, we'll send an update on all of the City Conquest defensive tower types, as promised. In the meantime, we wanted to give you a special update to talk about the coevolutionary system we're using to tune the game.
It occurred to us that passing along some screenshots of the output of our Evolver program would shed a lot of light not only on how our tuning system works, but on the optimal strategies for playing City Conquest.
The screenshots below show the two best players from the most recent run of Evolver pitted against one another in a post-evolution playback session. Although their cities aren't perfect, they show some remarkably intelligent strategies on the part of both players.
Keep in mind that all of the screenshots you see below show the strategies of AI players that were produced entirely by evolution. Other than the pyramid-shaped Mining Facilities you see on the crystals, every single building was placed as a result of the coevolutionary process pitting 500 red player AI scripts against 500 blue player AI scripts over the course of 2000 runs earlier this week (roughly 3-4 million games over 48 hours). The coevolutionary process built these bases because they turned out to be the closest the computer could find to "optimal" generic strategies.
[For more on genetic algorithms, see Popular Science, Koza, Brainz.org, and Wikipedia; for some interesting research on the (relatively recent) application to game design and design tools, see Togelius, Avery, Gillian Smith, and Adam Smith, to name only a few]
Screenshot 1: HD
There have been rumors of an upcoming ultra-HD iPad 3 with a screen resolution of 2048x1536. We really hope these rumors are true, because if they are, it's nothing but good news for City Conquest -- the game will be a perfect fit for that sort of ultra-HD screen!
I wanted to make screenshots at that resolution, but alas, that screen size is too big for me to simulate on my Alienware laptop. So I had to settle for a screenshot 50% between iPad 2 and iPad 3 resolution, at 1365x1012 pixels.
(EDIT: Apparently, Kickstarter resizes images sent in updates, so the image attached to your e-mail will not be the correct size. To see the 1365x1012 image at full-size, go to http://imageshack.us/f/42/fullscreens.jpg/ and then hit the "magnify" button).
Keep in mind that in City Conquest, you can zoom in and out at will, so if everything in the screenshot looks far too small to you, rest assured that with a quick two-finger "stretch" you can zoom in to a much larger scale almost instantly, just like you see in the Kickstarter video.
Also, bear in mind that this is only a barren test map for running the Evolver program; actual City Conquest maps will be much prettier and will have far more decoration. Also, we will pass along screenshots a number of fully finalized City Conquest maps later this month.
Screenshots 2 & 3: Blue Player
The second and third screenshots show the blue player's base.
We wanted to point out the following interesting evolved behaviors, all of which we think are things a good player would generally do in a typical City Conquest map:
- 1. Point of entry: the blue player has only one open entrance to its Capitol, in the rear. This allows it to focus on a single strategic choke point where it can mount its final defenses, and forces the red player to send its units all the way around to the back, giving the blue player ample opportunities to attack the red player's units along the way.
2a and 2b: Northern and southern routes. The blue player divides and conquers the red player's forces, splitting them between two routes around his city. Since one route is usually longer than the other, enemy units will arrive at different times, which makes them much easier to deal with as they will reach the Capitol at different times.
- 3. A well-placed Disruptor near the entrance allows him to interrupt the red army's cloaking, shielding, and healing effects at exactly the right time, just as they're beginning to get truly pounded by the blue player's defenses.
- 4. A fully-upgraded Shield Generator allows the blue player to protect nearby units from damage.
5. A set of 3 Lightning Towers are cleverly dispersed at the front, side, and rear of the blue player's base, ensuring that they rarely waste time targeting the same units at any point in time.
Screenshots 4 & 5: Red Player
The last two screenshots show the red player's base. As you can see, the red player's Capitol building has more health, so he is clearly winning. However, evolution forced him to adopt many of the same intelligent strategies as the blue player.
- 1. Point of entry. Like the blue player, the red player forced the opposing army to attack at the rear of his base.
2. Separate entry routes. Like the blue player, the red player splits the opposing army into two different paths to help upset their arrival timing.
- 3. A perfectly-placed Disruptor allows the red player to interrupt cloaking, shielding, and healing effects all along the front and side of its base (all along the north edge five Mining Facilities you see), as well as when enemy units circle around to the rear and finally close in on the Capitol.
4. A Lightning Tower at just the right spot gives excellent coverage, allowing the red player to zap enemy units at the front, side, and rear of its base. This placement ensures that this relatively expensive tower will see as much utilization as possible.
- 5. A pair of Ground Slammers, one at the front of the red player's base and another at the rear, allow the red player to cause earthquakes and give area-effect damage to all ground-based units that pass by.
6. A pair of Grenade Tossers is also nicely distributed in a very similar way, with an upgraded Grenade Tosser in the front and another non-upgraded Grenade Tosser in the rear. These allow the red player to hit all units (excluding Gunships) in a wide area with a moderate amount of area-effect damage -- enough to weaken them considerably and allow the other towers to take them out.
We hope this update has given you insights to the unit balancing process we're using on City Conquest. It has certainly proven to us that computational intelligence has enormous potential for making positive contributions to the discipline of game design.
Because a game's balancing determines its dominant strategies, Evolver has been invaluable in helping us to identify those dominant strategies as our game's balancing changes over time. In other words, it tells us what the best players do, so that we can tune those parameters accordingly and ensure that everything in the game matches its intended purpose.
We're looking forward to also applying Evolver to tuning our single-player game maps in the near future.
Rest assured that all of our missions will also have an "Easy" difficulty level for those who are feeling intimidated by all this!
Thanks again for your contribution, and stay tuned for this weekend, when we'll post an update on the game's defensive tower types!