The Stack

So, I’ve finished all five cases in CSI: Hard Evidence. All in all, I enjoyed this game more than I expected to. The chief thing to recognize is that it’s not at heart a mystery game, but a treasure hunt. This was clearest to me at the beginning of the fifth case, where you have to spot a number of bullets lodged in walls. I wish more of the scenes were as clue-rich as that, because it was one of the high points of the game.

I mentioned before that the game automatically tags with a green checkmark those scenes and clues that have been exhausted as information sources. This is just one of several “assists” that can be disabled from the options menu. Personally, I kept them all on, despite my complaints about the game being too easy: it seemed like disabling them would make the game harder in the wrong ways, extending the time spend searching fruitlessly in the wrong places and so forth. There was a point where keeping them on actually got me stuck for a while, though, when I didn’t yet understand that a DNA sample got its checkmark simply by being identified, and that this didn’t mean it was no longer of use in comparison to other samples.

So, yeah, it turns out that it is possible to get stuck after all, at least temporarily. Several forms of evidence processing involve comparing two pieces of evidence, and once you have many individual pieces of evidence, the combinatorial explosion makes it inconvenient to just cycle through all the pairs. So you do have to have some idea of what you’re looking for, some of the time.

There are some good things going on in the game’s UI. Like the navigation: I mentioned before that it’s based on clicking between nodes in a continuous 3D environment, but even more than that, the graph of these nodes is a tree, and clicking the right mouse button moves back up the tree towards your point of entry. The nice thing about this is that exactly the same interface is used for other tree-like UI elements, such as cancelling out of a menu. While I wouldn’t suggest that every game adopt an interface like this, it does seem like a good choice for games in settings where the player shouldn’t be able to get lost.

There are points that I definitely think could be improved, though. Nonstandard scrollbar behavior is my perennial gripe about homebrew GUIs, and while this game doesn’t have scrollbars per se, it does have button-based scrolling interfaces which don’t respond to the scrollwheel. It’s not even as if the game engine doesn’t have scrollwheel support: when viewing evidence, you can use the scrollwheel to zoom in and out. And speaking of viewing evidence, the controls for rotating items while inspecting them seem less than ideal. There are four buttons to the right of the view that can be clicked or held to rotate in two directions about two axes. But the axes are relative to the object, rather than the view, with sometimes unintuitive results. Plus, using buttons at all seems a little strained in a game that, in the scene views, normally handles rotation by moving the mouse to the edges of the screen. (Sometimes this even results in keeping an object at the center of your view and circling it, an effect similar to rotating an object in the evidence view.)

There’s also some stupidity in the way the game handles computers: your CSI toolkit contains a “USB data drive” that “detects encrypted data”, which is trivially decryptable by your lab equipment. Furthermore, people in this gameworld seem to be in the habit of encrypting their incriminating emails rather than deleting them. (Heck, just not encrypting them would be enough to escape detection here. It’s not like the game ever gives you the opportunity to read data that isn’t encrypted.) But I’m assuming that this is all inherited from the TV show. Plus, I may just be more sensitive to this than other simplifications made for the sake of gameplay. Goodness knows the fingerprint matching is greatly reduced from how it would work in real life, and the idea of getting a chemical analysis of a substance by sticking it in a chemical analysis machine is probably even more galling to chemists than anything done with computers here.

A certain amount of stupidity of content isn’t the only thing it inherits from the show. There’s the “bumpers”: when you go from scene to scene, you often get a brief montage of aerial views of Las Vegas, signifying “new scene” to the viewer. This is invariably followed by a “Loading” screen, which seems a little redundant, because it signifies the same thing. I suppose the limitations of the technology prevent it from displaying the bumper while loading the new data.

Another thing inherited: product placement. It’s not as blatant as in Lemmings 3D or Getting Up: Contents Under Pressure, but one of the cases has a subplot involving possible credit card fraud, and goes out of its way to mention how professionally those folks at Visa dealt with it, as well as just use the word “Visa” in preference to “credit card” wherever possible. (A print ad in the game’s documentation makes it clear that Visa is in fact sponsoring the game, or at least its documentation.)

So, would I recommend this game to people who aren’t fans of the show? No, not really. But perhaps I would as a study of graphic adventure techniques. It’s working with a limited palette, but it does a few interesting things I hadn’t seen before.

Disclosure: I received this game for free from Telltale Games.

CSI: Hard Evidence was apparently made using the same development tools as the Sam and Max and Bone games, but it’s a real contrast in style. For that matter, Sam and Max is pretty different from Bone: wisecracking cynicism and urban decay vs. good-natured fantasy, as well as the contrast in puzzle style mentioned previously. But Sam and Max and Bone are both ultimately cartoons, rendered in a cartoony style. CSI, although pretty close to a cartoon in its exaggerated and stylized story content, tries to be realistic in its visual appearance, including human figures. And this, if you ask me, is one of its weak points. The regulars from the show are passable, but per-episode characters — the victim and suspects — live farther down in the uncanny valley. Mouths in particular seem troublesome, and tend to bunch up in odd ways when characters is talk.

More than that, though, Sam and Max and Bone are both based on the Sierra/Lucasarts paradigm: you have an avatar who walks where you click. CSI uses something more like a Myst-style interface. Movement between scenes is handled through a “locations” menu in your PDA (which also holds the inventory, options menu, and case details), but movement within a scene is handled through clickable hotspots. But the scene itself is rendered in 3D, and even without clicking, you can do some limited shifting around by means of the mouse — or, presumably, the right analog stick in the console versions, which seems like a better fit to the mode of interaction here. It’s not quite like any other game I’ve seen: the closest is Myst V in “panning” mode, but there, the panning was always just a matter of changing the camera’s orientation, not its position. Here, you can use the mouse to do things like circle a car and inspect it from all sides, if that’s the motion that’s scripted for that node.

Disclosure: I received this game for free from Telltale Games.

I have been presented with one more Telltale game, and have something of an obligation to give it a whirl. CSI: Hard Evidence is the fourth game based on the TV series CSI: Crime Scene Investigation, and the second produced by Telltale. Now, I’m not a fan of this show. I’ve seen only one episode, which struck me as cartoonishly over-the-top. But unfamiliarity with the source material doesn’t always stop me from playing adaptations. For example, I played the videogame adaptation of The Da Vinci Code specifically so that if anyone asked me if I had read the book or seen the movie, I could reply “No, but I’ve played the videogame.”

And, in fact, this CSI adaptation has a certain amount in common with the Da Vinci Code adaptation: an emphasis on hunting for clues, lots of special interfaces, and, most of all, easy puzzles.

My impression may be wrong here: I’ve only completed one of the five cases in the game so far, and you’d expect the first case to be the easiest. But a lot of the easiness comes from user interface features that guide you towards the right things to do. For example, once you’ve gotten every possible clue from an area, a checkbox appears on that area’s icon in the travel interface. Also, the crime lab contains various machines for things like DNA analysis or accessing a fingerprint database, and at any moment, those machines that can be usefully applied to evidence you’ve collected have a special exclamation-point icon on them.

I suppose this is because its target audience isn’t fans of adenture games, but fans of a TV show, and a police procedural, at that: the genre of mystery that’s most about following established procedure and least about brilliant deductions. I’ve talked before about how the payoff in adventure games is the pleasure of figuring things out, of the “moment of realization”. The problem is that this can only come with a certain amount of risk that the player doesn’t figure things out and winds up stuck. This game seems to want to avoid that more than anything else. It’s aiming at an experience similar to the TV show, and no one ever gets stuck watching a TV show. Any brilliant deductions that do occur will be spelled out to the viewer — and so it is here.

So where is the pleasure in this game? I assume that there’s a certain amount of fantasy appeal, of joining the CSI family and having the people who you’ve come to know and love on the screen patting you on the back and saying “Great job!” whenever you follow procedure correctly. (The game is presented in first-person perspective with an unnamed protagonist, the better to aid player identification.) Obviously I’m missing out on that aspect; these people are strangers to me. The processing of the clues also provides an element of ergodic narrative, reminiscent of Portal (1986), but less linear and punctuated by little challenges such as finding a partial fingerprint in one of several possible matches. But the most interactive part — the part that seems most like a game — is simply finding the various clues and traces in the first place. This aspect of the game feels a lot like finding collectibles in an action game: it rewards being an obsessive completist and looking everywhere. The focus is on thoroughness.

In fact, there’s an interesting mechanism called “throughness points”. Every scene has various hotspots you can inspect, and not all of them actually contain clues. But whenever you inspect something that doesn’t hold a clue, you get a thoroughness point instead, and these are taken into account in your evaluation at the end of the case. So, with this mechanism, (a) finding new hotspots is never a waste, even if you discover nothing, and (b) you always know if the place you just clicked on contains a clue, because those are the spots without thoroughness points. More interestingly, your stats for the case indicate how many thoroughness points you haven’t found yet, which turns thoroughness points themselves into a kind of collectible — one that consists precisely of an absence of anything to collect.

Disclosure: I received this game for free from Telltale Games.

Now that I’ve completed all six episodes of season one, I’m wondering if mine was the best approach. Is it better to actually play the episodes episodically? Playing them as they come out undoubtedly lets you participate more in the Sam and Max fan community, speculating about things to come, even influencing the later content (as Merus points out in comments to my last post). But playing through the season all at once probably makes for a meatier experience. At one point in episode 6 (the last of the season), Max plays a videogame within the game, and complains “It was too short and not hard enough. I want my money back!”, an obvious dig at complaints in the forums. I haven’t followed the forums, but it’s inevtiable that people would make this complaint, as each episode takes just a few hours to play.

But I suspect I wouldn’t share that complaint anyway. I’m accustomed to short adventure games, thanks to the Interactive Fiction community and the annual comp in particular, so these episodes struck me as about the right length. Or possessing about the right amount of content, anyway. The episodes actually take longer to play than a typical comp game, but only because of the time spent walking Sam around from place to place — something I grew impatient with at times, and wished for a faster way to travel. (There’s a “warp drive” checkbox in the options menu, but apparently that’s just Telltale’s version of silly clowns.) So I may be one of the few people who wanted the episodes to take less time.

I wonder how much the folks at Telltale are aware of modern non-commercial IF? The Sam and Max games certainly show an awareness of their text-adventure heritage. Episode 5 features a whole scene set in a text-based environment, with Sam and Max themselves as the only graphical elements — a very stylish effect, I thought. It even uses that perennial only-possible-in-text gimmick, treating abstractions as tangible. Plus, there’s a sly shout-out to Zork in the beginning of Episode 4, subtle enough to pass unnoticed by the uninitiated. But the main influence on these games seems to be the classic Lucasarts games. Which may seem too obvious to point out — the first Sam and Max game was Lucasarts, after all, and Telltale seems to have quite a few Lucasarts refugees on staff. But what I mean here is the little touches, like the way responses to significant actions get shorter on repetition, and the way dialogue is used to provide hints disguised as jokes.

That last point reminds me a little of something John Cleese said about writing Fawlty Towers. The audience of a comedy show, according to Cleese, knows that anything that doesn’t lead into a joke immediately is a setup for a joke later on, and this robs the later joke of some of its impact. So he tried to make sure that all his setup material also yielded immediate humor, so that the viewer would be surprised at what was referenced again later. The principle is similar here, except that the goal isn’t (solely) an unexpected joke, but a moment of realization, when the player suddenly understands something’s significance without it having been shoved in their face.

Speaking of disguising your material, I notice that episode 5 keeps the whole business of doing things in threes (despite what I said before about episode 4 breaking the patterns), but tries to hide it by inflating numbers: there’s a group of four machines, of which one is useless, and a quest to obtain five gold coins, of which three are found together.

The threes come back with a vengeance in episode 6, though, with a very satisfying pre-endgame that puts Max in the center of a story. It seems pretty important to me that this happens. Of the two main characters, Max is the more emblematic of what they are, more gleefully chaotic, more disarmingly cute. If you see one of the duo alone in any context, it’s pretty much always Max. But these qualities also make him a difficult player character, and so for most of the story told by these games, he plays the role of Sam’s wacky sidekick. Even after he becomes president of the United States in episode 4, he’s Sam’s wacky sidekick whose wacky features include the presidency. But in episode 6, he becomes for a while the focus of the player’s attention, the thing that the puzzles are about.

By now, I’ve played enough of Sam and Max, Season One to notice some overall patterns. Most obviously, all the episodes share a certain amount of content. There’s always Sam and Max’s office and its neighborhood, including Bosco’s Inconvenience Store and Sybil’s, a storefront that changes in purpose from episode to episode, but always has the same proprietor. Episode 1 takes place almost entirely in this environment, leaving it only at the very end, while the other episodes treat it as a kind of home base that you return to once in a while for help with the puzzles in the episode’s main area. Episode 4 seems at first to break the pattern by starting in a completely new place, but the player returns to the old neighborhood before things are far advanced.

If it sounds like I’m complaining, let me make it clear: re-using content like this is not a bad thing. I’ve written before about how adventure games benefit from establishing a sense of routine, and the same basic idea applies to series. Repetition establishes a theme for specific puzzles to be a variation on, and helps the player to follow the author’s thought processes. Every episode here contains a car chase puzzle of some kind, but no such puzzle is ever repeated. Bosco always has some ridiculously overpriced gadget behind the counter, and the player comes to anticipate discovering how to get the necessary money from the episode’s new content. Even outside of the puzzles, the shared environment is used as a way to illustrate the progress of the plot: the office closet accumulates trophies of every episode, and there are multiple changes to the decor and its descriptions in the aftermath of episode 4.

There are also some less concrete patterns at work. The first few episodes established a very strong pattern of subtasks that come in threes. Most of the time, you have one major obstacle to overcome and three explicitly-enumerated things you need to overcome that obstacle. It’s a strong enough element that when I was told, in episode 3, that there were two tasks Sam needed to perform to infiltrate the Toy Mafia, my first reaction was “Wait, two?” (A third task was added before the conversation was over.) There’s even a threeness in the architecture of the core neighborhood, with its three sub-areas (the office, Bosco’s, and Sybil’s). But this starts breaking down in episode 4: I noticed a three-part task there, but it wasn’t the chapter’s major goal, and the subtasks were far from explicit.

The season has an overarching plot involving hypnotic devices supplied by an unseen mastermind; all the episode villains are either victims of hypnosis or using the devices against others. Each episode links to the next with a little foreshadowing of the next villain at the very end, and links backward with continuity references and a few preserved inventory items. The first three episodes don’t make any large changes to the status quo, and end with Sam and Max in pretty much the position that they started in, but that ends with episode 4. By now, you’re probably noticing a pattern yourself: episode 4 is the point at which the established patterns break, because the escalating wackiness can no longer be constrained.

For all that the characters of Sam and Max are emblems of the later point-and-click adventure, they’re not a terribly natural fit to the genre. If Steve Purcell (the creator of Sam and Max) hadn’t been doing art for Lucasarts at the right time, I doubt that anyone would have seriously considered them as potential adventure heroes. The original comics are driven by randomness and non-sequiturs. True, adventure games can and do get away with this sort of thing in the situations they present to the player. But the actions of the player character have to make some kind of sense if the game is to be solvable by any means other than exhaustive guesswork, whereas Sam and Max in the comics were just as unpredictable as their surroundings, and seldom did anything that could be expected to help their situation. Coming up with situations that let Sam and Max act like Sam and Max but still provide motivations for the player must have been a challenge. Some of the puzzles in the 1993 game were criticized for being too arbitrary, which is to say, for being too close to the spirit of the comics and not adapting to the new medium enough.

Season One has fared better so far. To the extent that it keeps the protagonists zany, it does so in ways that don’t require player involvement. For example, at the beginning of the first episode, Max has filled a closet with cheese, for no reason other than “you can never have too much cheese”. This is something that happened before the game starts, so it’s part of the premise, something that the player reacts to. But it’s also something that a player character did, and revealed in response to a player action (opening the closet), so the effect is similar to a player-initiated non-sequitur.

The dialogue is another big risk for an adventure adaptations. A lot of the humor of Sam and Max comes from the off-kilter tone of the dialogue, and the contrast between the the plush-toy appearance of the title characters and the casual and cheerful way they discuss horrors and mayhem. Examining an electric fan, Sam comments “Max almost lost a finger in a fan like that once,” and Max replies “Yeah, but it wasn’t my own finger.” This would cross a line if it were depicted visually. We know that Max is a furry little psychopath, and Sam isn’t much better — for all that they’re “freelance police”, they recognize no law beyond their own whims, and there are puzzles that hinge on this. But mainly we know it from their words, not their actions.

The reason this is a risk is that it’s all too easy for reliance on dialogue to hurt gameplay. Far too many adventure games and RPGs devolve into click-on-everything-in-the-conversation-menu for large chunks of the experience. Breaking a large text dump into a bunch of menu options doesn’t make it less of a large text dump, and that’s not what people play adventures for. Fortunately, the folks at Telltale seem to be fairly sensitive to this, and keep the menus fairly trim and mostly optional, as well as using them for an unusually large number of actual dialogue-based puzzles, rather than just infodumps. In fact, most of the trademark Sam and Max dialogue doesn’t come from the conversation menus at all, but from responses to examining things, as in the electric fan example above. This is a point where having a sidekick unexpectedly helps: Max is an independent observer who can argue with Sam’s descriptions and comment on his actions even if you never talk to him explicitly.

Ah, Sam and Max: absurd and grotesque, snappy and cynical. Their humor is always at least partly grounded in their horribleness. I’m a fan of theirs from way back, even from before their first adventure game in 1993, or their cameos in various other Lucasarts titles. Back when they were an obscure indie comic book.

So you might think I’d be among the first to snap up the newer episodic Sam and Max games. But I didn’t, because I was wary of Telltale Games. I had played Telltale’s first adventure, an adaptation of the comic book Bone, and found it disappointing. The adventure content was minimal, as was the interactive detail: the bulk of the player’s time was spent on a series of lame mini-games shoehorned onto a story that didn’t really want them. And when I say “series of lame mini-games”, the part that bothers me the most isn’t the “lame”, but the “series”: the game was very linear, following the source material very closely. Most of the time, there was only one thing to do.

But the Sam and Max games aren’t adaptations of existing Sam and Max stories, and thus avoid a lot of the difficulties of adaptation. (Even the 1993 game, Sam and Max Hit the Road, which took a lot of its ideas from the “Surfing the Highway” comic, has an original story.) I’ve played through the first episode by now, and it’s got classical adventure game structure: after a brief prologue in a constrained area, it sets the player loose in the main game area with three major goals to pursue simultaneously and independently, followed by another, smaller set of three goals, followed by an endgame. And some of those subgoals provide good “Aha!” moments.

So, currently, I’m pleased. I’ll see if I can get through episode 2 tomorrow and try to spot common patterns. It really seems like each episode wants to be completed in a single sitting.

In my last post, I described a very effective combo using Kobold Elders and Kobold Shamans. By the end of map 4, I had two high-level heroes, one of whom was using this combo. The other, which had progressed around the map by a different route and had encountered different spells, was using a different technique, one that’s more elementary (so much so that I hesitate to call it a combo) but also quite effective: a deck made mostly out of rats and anger.

The common stink rat is the beginner’s monster for team red: it’s 1/1 and costs 1 mana to cast. Anger is an enchantment that gives all your creatures +1 to their attack power. This bonus stacks, so your damage potential increases with both the number of rats and the number of Angers. Of course, the same bonus applies to non-rats, and that’s important sometimes — I put a couple of bats into this deck as well, because a couple of the enemies had a spell called Flood that disables anything that can’t fly — but the rats were cheap and disposable, and the strategem works better with lots of creatures than with a few powerful ones.

Both of these decks have the advantage that they start damaging the opponent immediately: kobold shamans and stink rats both cost 1 mana, so you can summon them on your very first turn. But more importantly, they’re both O(n²). That is, your potential to do damage on a turn, barring interference, is roughly proportional to the square of the number of turns that have passed. Even though the amount of mana available to you increases with every turn, you still only get to draw cards at a constant rate per turn. So in the long run, the number of instances of a spell active at any moment is going to be linear on the more limiting factor, time. But the damage potential of these strategems is determined by the product of the number of instances of two different spells.

There are other combos with this property; it may in fact be a feature of every deck that wins at high levels. In fact, there’s one instance I’ve observed of a spell that’s O(n²) without a combo: Grass Snakes. Every time a grass snake hits the opponent hero for damage, its attack power (and health, but that’s not what I’m considering here) increases by 1. I suppose this means that a snakes-and-anger deck would be O(n³). [Edit: Not really, see comments.] (Actually, that combo is impossible: Anger is red, and snakes are green, and never the twain shall meet. But apparently there are a few rare green spells that have a similar buff-all-friendlies effect.)

But I doubt that such a deck would actually function as O(n³) in practice, because the bonus on the snakes is per-snake, which makes it vulnerable. Every time a snake dies, any bonus it built up dies with it. The other decks I described are more robust: if you kill my rat, the next rat I summon will be just as angry. Catching up to where you were is linear (that is, O(n)) on the number of rats killed. For snakes, in the worst case it’s linear on the number of turns the oldest snake was alive… which, now that I think about it, makes it also linear on the number of snakes killed, because both the maximum age and the number of snakes are linear on the number of turns played so far. I guess big-O notation doesn’t tell us everything.

Here’s a better analysis: If you can kill my (oldest) snakes at the same rate as I can summon replacements, my damage potential from snakes will never increase. Whereas in the rat/anger deck, a rat equilibrium will just slow me down from O(n²) to O(n), because I’ll still be casting Anger at a constant rate. Unless you have spells that remove enchantments and we have equilibrium there too. I suppose what I mean by “robust” is that disrupting it completely requires more things.

Probably the biggest joy in CCG-style play is coming up with combos. By “combo”, I don’t mean the kind you get in, say, tile-matching games, where the rules explicitly grant bonuses for things that are not valuable in themselves. I mean emergent synergy, the special properties of different cards combining in a way that makes them more effective. This is the essence of deck construction.

For example, in Magic: the Gathering, there is a card called “Lure”. (Or at least, there was when I was playing it. A lot of cards were removed in subsequent revisions.) Lure is an enchantment that you cast on a creature. When a creature with Lure makes an attack, anything that can block it must do so. By itself, this would typically mean that the creature with Lure is met with overwhelming force and dies, but in the process, it lets your stronger attackers go unblocked for that turn, because blockers can only block one thing at a time. But if you put that Lure on a creature with Regeneration, it can survive the onslaught and play its part over and over, or at least until the opponent kills it with direct damage spells. Alternately, you can put it on a Thicket Basilisk, which has the special power that anything blocking it dies at the end of the combat round, and wipe out all the opponent’s creatures in one turn. Put all three cards (Lure, Regeneration, Thicket Basilisk) in your deck and you have a three-card combo with the potential to wipe out your opponent’s army every turn — but only if you happen to get them all in your hand and manage to cast both of the enchantments on the basilisk before the opponent kills it. With a minimum deck size of 40 cards, three-card combos can be hard to pull off.

The dynamic is a little different in Etherlords. For one thing, the 15-card deck size makes it almost certain that any combos you put in will come up. Also, some combos are more explicit than what I’ve just described. Each type of creature comes in multiple subtypes, with different levels of strength and different special abilities, and sometimes those special abilities are pretty clearly linked. Take the Kobold Elder: its power is to untap all Kobolds in play other than Kobold Elders. This is an ability that can only be used in combination with a rather small set of other cards. If you ask me, the only other kind of Kobold worth having in your hand is the Kobold Shaman, which is the Etherlords version of M:tG‘s Prodigal Sorcerer: you tap it to do 1 point of damage to any player or creature. Thus, if you have both in your deck, you get to do (number of Kobold Elders times number of Kobold Shamans) points of damage every turn, without even making an attack or casting any spells. This is potent. In one encounter late in map 4, I was routinely killing things with 8 or 9 health (that is, the hardiest creatures I had yet encountered) the moment they appeared.

Both of these effects — the smaller deck size and the signalling of likely combos through creature type — serve to make combos easier. The former makes them easier to execute, the latter makes them easier to discover. I can’t know if this was the result of a deliberate policy of combo-friendliness on the part of the designers or just a happy accident reinforced by playtesting, but that’s how it turned out. The downside, I suppose, is that it makes the combos seem very planned, and leaves little scope for the player to make genuinely new discoveries. But this is OK for a computer game with fixed content. Real CCGs take advantage of the combinatorial explosion to create an impression of infinitely variable gameplay. But campaign modes in strategy games are more about introducing you to gameplay elements one by one, and typically end when they run out of new things to show you. I expect that’s how it is here.

Map 4 seems to be the point at which Etherlords really starts to be about non-combat spells, or, as it calls them, “Global” and “World” spells. The distinction between Global and World is unclear to me; it may have been clearer in the game’s original language (Russian, apparently). But the player starts the the scenario with a whole bunch of them, and the player pretty much has to figure out what they do by experiment, because the docs are so dismal. You can get pretty far in the level without needing any overland spells beyond the ones to summon new heroes, which are familiar by this point. I’ve explored nearly half the map by now, and only just started running into a need for more. This is because the opponent just started attacking my castle ethereally.

Normal, physical attacks on castles are pretty simple: you march a hero up to a castle, and every turn that the hero is there, the castle receives an amount of damage equal to the hero’s level, possibly with a bonus from skills or magic items. Eventually either the castle’s owner manages to kill or drive away the attacker(s), or the castle is destroyed. Ethereal attacks work more or less the same way, except for the “march a hero up” part. Any hero can attack any castle ethereally, provided you have enough mana to keep the spell going. Heroes can also be assigned to ethereal defense of a castle by means of another spell. The effect is that the attacker has to defeat the defender in combat before doing any damage.

The thing is, no one actually dies in ethereal combat. Whoever wins, both heroes just go back to their bodies at the end, without even gaining any experience points from the exercise. I can imagine attacking an opponent’s castle ethereally just to keep their most powerful hero tied up defending it instead of killing your guys. In fact, that’s almost what’s happening to me: the enemy is attacking my castle with a level 6 hero, and I have one level 6 hero that I want to keep leveling up, but can I afford to? It seems like it might be a good strategy to go on a counteroffensive here, send one of my weaker heroes against his castle to see if he switches the strong guy to defending it.

Come to think of it, ethereal combat is a lot like a strategic version of what normally goes on at the tactical level. In combat mode, you basically choose every turn whether each creature under your control should attack or hang back to block attacks from the enemy. There’s no movement, no map to wander around, just a basic attack/defend option and a hero (in combat mode) or castle (in map mode) with a bunch of hit points, which it’ll lose if you don’t defend it. I wonder if this analogy was deliberate, or if the designers just felt it was the natural way to design things after they had spent so much time on M:tG-style play.