AlphaStar: Maybe?

Big news in gaming this past week is that Google DeepMind turned its attention to another game and had a dominating victory, this time StarCraft II. But are we seeing the same sort of thing we did with Go and Chess? I’m not sure, and I am probably not going to look deeply enough to make a judgment, but I wanted to pass along a couple of links.

For those who missed previous DeepMind adventures: Google has a machine learning setup that it is turning loose on games with amazing results. Go was long considered a space where humans would continue to beat computers because it has a very large decision space. As it turns out, no, modern computers can trounce that, and AlphaGo is the best player in the world, getting even better when it completely ignores the human history of playing Go and starts with only the rules, playing itself millions of times. AlphaGo was so good that human commentators could not see what it was doing and were laughing at its “mistakes” as it beat the best human players alive. Oh, it turns out humans have been playing Go suboptimally for hundreds of years. And then AlphaGo got better.

StarCraft II is a similarly large next jump, a game without perfect information, in real time, with an even larger decision space. I can just point you towards the DeepMind AlphaStar post on it. AlphaStar can beat professional StarCraft II players, apparently pretty consistently. You can see more discussion from the team at their Reddit AMA.

I found this counterpoint rather compelling. The argument is that AlphaStar is winning through superhuman clicking capabilities, not superhuman strategy as was the case in Go. AlphaStar was, so the argument goes, executing strategies dependent on perfect micro, telling individual units where to move and attack. AlphaStar will always win fights between even numbers because it can perceive and act quickly enough to target individual shots, and its strategic difference is favoring units where this advantage matters more in terms of perfect placement.

I am not qualified to comment on the extent to which AlphaStar is making human-imitating spam clicks versus effective clicks. I do find it compelling to point out that AlphaStar engages in 1500+ APM at critical moments, particularly the example of having three units teleport simultaneously to three different points for an attack. Even if you were capable of perfect precision, I cannot see a human using the mouse and keyboard fast enough to do that … while also engaging in 19 other clicks in the same second. Being able to act without an interface and off camera is really big, and it is not the sort of mastery intended here.

AlphaStar is intended to demonstrate human-equivalent or -superior decision-making. If it instead teaches itself to exploit human-superior clicking, that can lead to a win but not the goal, like the machine learning story about the computer who learned to win games by making the opposing computer crash. That is not winning within the game.

In the end, though, this is probably the most important comment. Yes, there is a lot to be considered in terms of what restrictions the computer needs in terms of speed and accuracy, but ultimately this was the proof of concept. With more time and practice, we should expect AlphaStar to do amazing things within the game.

And remember, AlphaGo got better than any human had ever been by learning from humans. It got better than that by learning on its own. Twice.

: Zubon

Customizable Difficulty

One feature I enjoy in Slay the Spire and Spirit Island is that both games have highly customizable difficulty, both in type and degree. This lets you pick how difficult (or easy) you want your game to be.

Slay the Spire has its base game and then Ascension. There are 20 levels of Ascension, each of which dials up the difficulty a little more, usually by adjusting a number or two. About half the levels are increasing the damage, health, and movesets of normal, elite, and boss enemies.

Slay the Spire also has custom runs, which let you pick one or many modifiers to the game. Want more control over your starting hand? Here are several options. Want more randomness? More options. Harder fights, easier fights, trading off one thing for another? Screens of options. Or try the daily mode for a random set.

Slay the Spire also supports mods now, so you can adjust the game to the limits of your competence.

Spirit Island has its base game and then adversaries. You can decide that your invaders come from one of four European nations, each of which has base modifiers and then levels 1-6 of increased difficulty. The “base” difficulty of each nation varies, so the game comes with a chart showing how each adversary & level ranks on a scale of 1-10. (They considered having the “levels” just be the difficulty level, but it caused more confusion trying to explain why the numbers might be “1, 2, 4, 6, 7, 9, 10,” and then people referred to the top level as “six” anyway.) The next expansion adds more adversaries and a way to combine them for stacking difficulty. And then there are two levels of lowered difficulty (which can stack). (I should try dialing difficulty all the way up and all the way down at once.)

Spirit Island also has scenarios, which let you pick a set of modifiers to the game. Most of these make the game more difficult. All of them make the game different. One makes everything faster, one focuses on guarding the island’s center and another the island’s coast, and another adds a chance for random buffs to the spirits or invaders. Several add and remove win and loss conditions. I am particularly fond of one named “Second Wave” that lets you keep the same island after completing a game, reset some variables, then keep playing with new spirits.

Sometimes you want to try new things. Sometimes you want to challenge your limits. Sometimes you want to faceroll.

: Zubon

Mod the Spire: So Far No Good

I thought I would try the modded classes for Slay the Spire. I’m only a few in, but so far nothing has been great. All of them, even the one I’ve found that I like (the Slimebound), seem to suffer from wild swings in card quality. There are overpoweringly great cards and fairly awful cards, without a lot of “pretty good” in between. When I want a meaningful decision in selecting cards, I want tension between which option would be better for me; right now, I am only seeing that when I have multiple “I win” cards to choose from, otherwise it seems like “all crap” or “obviously pick that one.” This might average out to balanced, but at least as likely it leads to wild swings between “easy wins” and “I didn’t get any good cards.”

Player mods may not come out looking good compared to the base game, which has been tested against thousands of players for millions of hours. They may do neat things, but they have had less than scrupulous balancing applied.

Evaluating player mods is difficult and time consuming in a rogue-like. A few playthroughs is not enough to see everything, less evaluate it. Did I see only the lousy cards or the overpowered ones? Did I get the wombo combo early or never? Am I really willing to spend hours digging through poop in hopes there is a pony in there?

Special shout out to mods with cards that are straight-up better or worse versions of existing cards. The exact same effect with a different cost, or the same effect with a different number, is either an argument against the original game’s balance or a sign they didn’t even notice it. Maybe the different numbers make sense in the context of the new class, given its other mechanics and cards, but probably not.

: Zubon


We played through the campaign of Charterstone. It was OK?

Charterstone is a legacy worker placement game. “Worker placement” is a game type where you place workers around the board to take actions like harvesting clay, building buildings, and turning in completed quests. “Legacy” means that the game changes over the course of the campaign, rather than being a series of isolated games. In a non-legacy worker placement game like Lords of Waterdeep, you expand the buildings available over the course of the game then wipe it out when you’re done. In Charterstone, your village grows over time as you add and replace buildings. There is also a storyline across the games, with rules and complexity added over time as well as changing the rules to simulate story events.

I like the idea of it more than the execution. I have several worker placement games already, including a better one from the same creators, so the gameplay is not the big draw. The storyline, surprises, and legacy aspect must be. There are a few of those that are rather good, but those are one-time surprises that lack the replayability one expects in a boardgame. (The game has rules for playing with Your Unique Village after the campaign, but that is again just the novelty of having Your Unique Village.)

The storyline is OK. There are several points with delightful surprises, particularly the end of game 4. More or less anytime a new gameplay element was introduced, especially a new component from the sealed boxes, that was great. Those combine the joys of a storyline twist with opening a wrapped present. (Spoilers are allowed in the comments, if you have played and want to discuss.)

The game is not a massive, branching path, so player decisions have almost no effect on the storyline. There is a cumulative effect, and there is a climactic battle determined by earlier decisions, but you make those decisions with little information and no indication of future impacts. You are familiar with this in games: make a decision right now that will have permanent effects, with little context and no explanation of its implications. This is usually considered bad design. The quick tip here is that making the mechanically disadvantageous decision is usually the one that gets you closer to the “good” ending. And the entire difference between the “good” and “bad” endings is the flavor text on two cards.

We played Charterstone with two players. It is a 12-game campaign, and you want people to commit to the whole thing before playing. That’s kind of rough, especially if you want a full game of 6 players and have trouble getting them to commit for a regular gaming event. Even if they want to commit, life happens. 7th Continent has rules to drop in or out; Charterstone has them only for dropping out.

In a two-player game, we missed most of the gameplay conflict, as there is lots of space on the board, and the range of options was somewhat limited as the other parts of the village filled up with random buildings over time. There were always lots of options, but “fill in the empty charters with random buildings” gives you a mix of poorly developed options, while your own part of the village has the custom engine you have built up over the course of the game.

The rules explanation is just bad. In an attempt to keep the new mechanics a surprise, the rulebook starts mostly empty, and you fill it in over time. That can lead to references to things that do not exist. The first game explanation seems inadequate, so watch the recommended video. There is an official FAQ, but by reputation its “mild spoilers” are not all that mild, which makes sense given the thinness of the story. Also, just looking at the questions will give away things being asked about.

I really wanted to like Charterstone. If nothing else, it seems like good discipline to treat a game as something temporary, consumable, and alterable, rather than something I must keep pristine and preserve. You probably have many games you have never played a dozen times, so if you run through the campaign, you are getting your money’s worth. I just think you would get better time value out of better gameplay, and you can read the storyline in about five minutes.

: Zubon

Mod the Spire

As a last big thing before the official release in two weeks, Slay the Spire now supports mods. There have been mods for months, but it is now supported through Steam Workshop.

At a friend’s suggestion, I am going to start trying out mods, particularly modded characters. When you have played 400+ hours, you may need something new to keep it fresh. I expect some to be well conceived, some to be poorly balanced, and generally to enjoy fumbling around with “is this bad or am I doing it badly?” Because Slay the Spire was my first real “git gud” game.

One nice thing about mods, rather than the base game, is that they can be ridiculous and that’s OK. Like the modifiers you can add in custom mode, you can intentionally make the game much harder, easier, more random, etc. That is both adjusting the game to fit your own needs and doing something silly you might enjoy occasionally but not want every time. God mode is fun for a little while, even if it is toxic to the game in the long run.

If you have mods you like and want to recommend, the comments are open.

: Zubon

Victory and Loss, In Parallel or as Opposites

Also note that the win condition (fear, kill invaders) is totally decoupled from the loss condition (too much blight, no presence). That means you can be very close to winning *and* very close to losing at the same time. This is not a coincidence, by the way. It was a very intentional part of the design. The result is that players can be much closer to winning than they thought because they were only thinking about whether they will lose. In most games, being close to losing means far from winning, so it feels like you’re further than you are.

The upshot is that you can often win at higher difficulties than you’d otherwise expect. One of the things Spirit Island does well is making easy games feel challenging. Players like feeling challenged and they like winning. Spirit Island delivers both.
Ted Vessenes, designer on Spirit Island

A common feature in Eurogames is a scoring mechanic (sometimes hidden) that keeps all players in the game until the end. You can recognize that someone is ahead of you, but you do not face a loss condition that eliminates you from the game. Contrast this with games like Monopoly or Risk, where eliminating opponents is the point of the game. Contrast this with games that do not have win conditions as such, just that you can lose The Game and the goal is to not-lose.

When we play a game with a loss condition, losing tends to mean being far from winning. The losing player has few playing pieces, little power, etc. Losing is far from winning. A mechanic more common in survival games is a race. You are trying to win before you lose, which seems like a tautology but plays differently from games where you are not racing against a loss condition but rather against other players’ attempts to win first or against other players who win by making you lose.

Spirit Island victories tend to feel either really amazing or kind of trivial. Racing against a loss condition can feel like kind of a cheat, as you steal victory from the jaws of defeat. A game like Blood Rage makes that a central mechanic, where it is Loki’s strategy is to win by losing. American games and stories tend to favor action heroes over guile heroes, so we are culturally primed to see “a trick” as generally a bad thing. You are supposed to smash your enemies, not win by getting a ninth fear card as they are on the verge of overwhelming the island. But that can also be an amazing feeling, as you see the race happening and roll under the closing door like Indiana Jones.

Other times, you are clearly ahead in the race but will take a while to claim technical victory. There are rules to bring about the end one way or the other, but if you get an early lead and just need to check off one box to win, it can become a dull waiting game as you try to whack that last mole. This is rarer and usually a sign you should dial up the difficulty.

But sometimes this creates the most amazing crescendo, where the only anticlimax is that your big turn at the end is so overwhelming that you do not get to play it all out. Your power accelerates as the enemy’s does, which gives you more targets and more ways to crush them. It is not uncommon to have a problem area festering on the map throughout the game, only to wipe it out in a turn or two, sweeping the invaders from the land while generating large amounts of fear, potentially hitting both win conditions on the same turn. The enemy buildup creates tension, and big tension leads to big explosions.

Finally, as the developer says, the games tend to feel closer than they are. The enemy is building up and spreading as you are, so even if you are way ahead of the curve, you can still get these hordes of enemies you are trying to contain. Unless you are very safe, you do not feel safe, and even then the next wave is coming.

As a gameplay note, I recommend having someone actively monitoring the win condition and calling it out at the start of the turn. It feels less like winning “just happened” if you can see that you are building towards it. “We are at terror level 2. We have 2 cities and 5 towns to clear out. We win if we eliminate all the buildings, or if we can clear those two cities and get 11 more fear. Invaders ravage in mountains this round then build in sands.” This is especially helpful for new players, because it focuses them on what is more important in the game right now.

: Zubon

Steam Sale: Slay the Spire

The Steam winter sale is going on now. You already know this.

Slay the Spire is on sale. In January, the game is leaving Early Access and getting a permanent price increase. This is the best time to buy it unless you are going to wait for it to be 75% off the new price.

I’m a guy who waits until everything is 75% off and/or $5 or less, and half the time decides I don’t really want it by the time the price gets that low. And I bought Slay the Spire for full price. And it would have received good value for four times that.

: Zubon

Games of the Year

Computer game: Slay the Spire. There is nothing I have played or enjoyed more. It is a great rogue-like deckbuilder at an indie price, with lots of variation and replayability. It has a very high skill ceiling, and you can also dial up the randomness if you prefer that.

Board game: Spirit Island. It is a deep strategy cooperative game with variable powers and strong theming. It places you in the opposite of the usual game role, driving off the settlers. This can also be played in modes that are almost purely predictable or with increasing randomness. It also plays excellently as a solo game. I have played Spirit Island every two days since I learned the game. It is great, although difficult to introduce to large groups.

What are your top games this year?

: Zubon

Necessary But Not Sufficient

Honestly the thing that makes me saddest about the AAA Video Game Clusterfuck du Jour is knowing that there are probably a ton of people who spend shitloads of crunch time lovingly working on the one tiny part of the game they had control over only for it to be hopelessly fucked up at the executive level I don’t have the interest to get involved in the latest kerfuffle, and by the time you read this it could be an entirely different kerfuffle, but this remains the same.

Ravious and I had the chance to see WildStar pre-release. There were some great people making great systems, props, etc. The game as a whole had a feeling of “let’s throw everything into the mix and hope for emergent gameplay.” And it just shut down permanently. People worked for years, and you know how hard folks work in the game development industry. And at this point, the best I can say is, “I hope it was fun while it lasted.” It wasn’t a lot of fun when I tried it.

A recurring thought I have at work is that it is necessary but not sufficient for our office to do a great job. That, for the system as a whole to succeed, we must succeed, but our success is not enough to carry the rest of the system. We can do great things that immediately get trashed by the next step in the process.

So it goes.

: Zubon

Slay the Spire: The Final Act

The actual ending of Slay the Spire is now available, an unlockable 4th act. It is not much of a 4th act, but it gives you a final boss fight for the game. I don’t think it adds much except for a sense of finality. Except that you can immediately start another run, so final isn’t final.

Slay the Spire has clearly been my game of the year, providing me with more playtime than most MMOs I have tried. It’s great, worth buying even without a sale. It feels a little anticlimactic that adding the final boss seems less significant than some of the regular weekly updates. The baseline for the game sets the bar pretty high.

: Zubon