[h1]Summary[/h1]
Essentially Space Invaders, but with a true natural selection system which facilitates evolution of enemy behaviors, resulting in progressively greater challenge (and virtually inevitable defeat for the player, given enough time.) Heaping helpings of player ship, weapon, and ability unlocks. Somewhat of a rogue-lite sensibility is present here as well, in that the object is to adapt and survive as long as possible against an equally (or more) adaptive opponent. But unlocks are persistent, so your tools expand over time.
[h1]Full Review[/h1]
How many different takes on the basic formula Space Invaders and its kin created 40 years ago have we seen down through the deaces? Probably too many. I find most to be boringly derivative, and very few that are at least fun, let alone compelling. Fewer still, though, can be called truly remarkable. Darwin's Demons firmly positions itself within that handful of Space Invaders re-interpretations to distinguish itself in a truly novel way.
"Adaptive difficulty" was briefly a buzzword in game design a few years back, along with procedural generation, verticality, and the like. But as with those, few games truly leverage the concept to do anything particularly noteworthy. And rarely has it ever been truly adaptive, so much as simple difficulty scaling.
It's surprising then that no one has thought to marry the basic premise of natural selection to a wave-based SHMUP before. Because the result - at least this implementation of it - is astonishing both in its "Why didn't anyone do this before now!?" DUH-ness, and the sheer fun it provides.
Other games do incorporate something along these lines, albeit in a less direct form (Boss Constructor comes to mind, wherein you can pit designs against one another to determine fitness, and then either use those designs yourself or deploy them as enemies.) But the simplicity of the Space Invaders inspired gameplay here makes the product of natural selection so much more immediately apparent, transparent, and, for me, entertaining.
As in the classic shooter, you move your ship left and right while firing at advancing waves of enemy spawns, possibly utilizing the cover of various optional, player-selectable defense barriers (more of which can be unlocked.)
At the end of each battle (called "generations" here,) the game chooses enemy types based on fitness - measured by speed, movement types, aggression, dominance, and projectile speeds, among other under the hood calculations. The most successful designs, coupled with the potential for mutation and genetic anomalies, then get "inherited" into the next generation. Thus the game becomes progressively more difficult over time.
The player can likewise earn a form of currency by playing skillfully and dispatching waves quickly, with which to unlock their own ship, weapon, health, and defense upgrades. You are therefore essentially in a race against time, trying to establish an equilibrium that allows survival for another generation in the face of an increasingly deadly AI.
[h1]Evolution and Emergent Behavior In Response To Playstyle[/h1]
The brilliant thing about this design, is that the way you choose to play can influence how the AI evolves over time. Some examples I noticed while playing:
[list]
[*]Unbeknownst to me, there was some slight leftward drift in my left analog stick due to calibration error. It was almost undetectable to me due to being engrossed in the game, but the AI "noticed." (That's an anthropomorphism of the natural selection system in the game, of course. The enemies which survived relative to that flaw in my own behavior simply became dominant in the evolutionary process.) Enemies began bunching up and rapidly advancing on one side of the screen, exploiting my slight bias, as if to account for my apparent aggression toward that side. When I finally realized what was happening, I recalibrated my controller, and over successive generations this shift to the left gradually diminished in the AI ships. However, it then began to spam enemies towards the right, perhaps "learning" (to be clear, no actual learning happens in this game - it's not using reinforcement learning, just pure natural selection with dynamic fitness and mutation) that I had been paying attention to the left more than the right even after I consciously corrected for the issue. That ended up being how I lost, as I was so used to focusing on the other side, I missed an enemy slipping through my defenses on the right!
[*]In Experiment Mode (more on that later,) I decided to make myself all-powerful. I set myself up for rapid advancement through upgrades, and maximum payoff for victories. But I also wanted to see what the AI could do when pushed to its limits. As such I set up its evolution to advance rapidly, and its potential for useful anomalies to be greater than normal. The game surprised me, despite what I expected to be an easier challenge than the main Arcade mode, by ending my run only 19 generations in. It adapted to my overwhelming firepower, by selecting the enemies which survived longest: in this case, those whose projectiles formed defensive clusters which protected the ships from my auto-targeting Needle lasers. This filled the screen after a few generations with so many projectiles my system could no longer render them at a smooth frame rate. Eventually, unable to compensate for the input lag this introduced, they whittled my lives down and killed me. (Their firepower also quickly became so great, that my fully upgraded defense barriers were even ripped to shreds!)
[*]I decided to experiment by intentionally taking hits from weaker enemies, and quickly destroying the greatest threats. I was pleasantly surprised when the game began to produce weaker enemies, as it interpreted their fitness as higher than the real dangers. You can fool the system!
[/list]
These are just a few examples. I doubt whether they would be so clearly communicated, so fun, or so immediately apparent if this were a more sophisticated game. Somehow the simplicity of a Space Invaders-like design is a perfect fit not only for demonstration purposes, but also for the most important thing in any game: FUN.
[h1]Experiment Mode[/h1]
In addition to the main Arcade mode, which has multiple difficulty settings, there is also the Experiment Mode. Here you can adjust all sorts of modifiers to make the game harder or easier for yourself, change how the AI evolves, its biases, how it measures fitness, player rewards, and more. This allows you to do exactly what it says: experiment.
You can use it effectively like a cheat mode if you want to, or you can genuinely make the game more challenging than it could ever be in Arcade mode even on Insane difficulty. The choice is yours.
[h1]Recommendation[/h1]
This is a surprisingly clear & direct demonstration of user-interactive natural selection at work in a software environment. More importantly, unlike more advanced evolution sims, it's a functional, intuitive, fun gaming experience. If you like Space Invaders style SHMUPs, you'll likely enjoy this one. Even more so if you like rogue-lite style progression, and have any fascination with truly adaptive difficulty.
It's Space Invaders, evolved... literally, and pun intended.
[h1]PROS[/h1]
Tons of potential mutations, resulting in many different ship types, attack patterns, weapon varieties, etc.
Genuine adaptive difficulty - one of the few "real" implementations of it, in fact.
Clearly communicated fitness measures, demonstrating basic natural selection understandably and enjoyably.
A fun SHMUP in addition to all of this.
Plenty of player unlocks, and an empowering sense of progression balanced against increasing challenge.
Simple but pleasing visuals.
Catchy mixture of electronic & metal tunes.
[h1]CONS[/h1]
Evolution always wins in the end
Potentially repetitive
Lags with heavy particle rendering
