This is the concluding part of a story about Slay, a strategy game first released in 1995. The first part explained the rules and how I came to take on the last level Rauk.

I installed Slay while at a motorway service station, a brief stop during an arduous drive home from a family holiday. The reasons for why this happened remain sketchy. I think I hoped to show the children a low-key mobile game that is addictive without having to turn to special effects or gem collecting meta-games. I explained I’d been sucked into Slay years ago. Here it is, I said while we shared low-nutrient motorway food, watch me play this game.

My wife had also played Slay before and she didn’t even wait before we’d finished eating to pull it off the Google Play shelf. My son then downloaded too. In the space of ten minutes I had managed to get most of the family hooked on Slay including myself.

Two months on, my wife is still playing daily. She is advancing through the levels in order. Between games, sometimes we talk.

My wife hasn’t taken on Rauk, but she’s heard me grumble about it. If you recall, I was on the verge of victory and it was pretty much a done deal. Next episode, Joel boasts about how he relished taking down the Rauk AI. And what he learnt about life and culture in general. Bonus points for a religious experience.

That is not what happened.

See, while I was excited I was also quite, quite terrified. I was fed up of moving little figures back and forth across the surface of Rauk; I needed release. Frightened that the same AI was cracking its knuckles to my North and my South, and knowing it wanted to link its territories, I decided I needed to punch South instead of just holding it back with a knight. I created a baron to do the dirty work.

I loved that first brutal move, when my baron ate one of AI’s delicious knights.

What I failed to realise was my territory was not earning enough to sustain a baron.

The territory bankrupted in turn 17 and immediately the AI territories went for my throat. I threw the phone into a wood chipper in disgust. And seeing as I had no idea how I had pulled this off in the first place, I was right back where I started.

I probably should have taken a break. This is not what happened.

Death in the Leaves

Maybe we should take a break, though, a detour through a verdant forest. Yes. I’m ready to talk about the trees.

Players will most commonly bankrupt an AI through bisection. If you can fire an army through the middle of an enemy territory, it divides in two and it’s highly likely – although not certain – that one of the two sides will hold expensive armies without enough income-generating hexes to carry them.

In the early game, you have to keep your eyes open not just for your bisection opportunities but your own bisection vulnerabilities because, trust me, the AI is all over that shit. If you’re playing a gentle level, you can recover from a few bisection mishaps. But when shit gets real, bisection attacks are often game-ending incidents. Tap the surrender button and start again.

Player territory bisected, pieces in smaller territory die

In the twilight phase of a Slay match, bisection becomes an invaluable tool to cripple a monster AI that has vast territory at its disposal. But there is a serious downside to bisection bankruptcies.

A bankruptcy transmutes people into gravestones. Gravestones then become trees and trees only do one thing. They grow. A bisection bankruptcy will sometimes dump a shit-ton of trees right beside your territory and they will bleed across your border like a stealthy ninja army. Each hex infected with a tree is unable to produce coin. In short, if you don’t stop rewilding, nature will fucking end you. Perhaps there’s some awful climate change interpretation of this mechanic but unfortunately I hate the ninja trees as much as the AI and I haven’t got time for metaphors today.

Player loses 17 coins to trees this turn

If you find yourself faced with a choice of fighting trees or fighting the AI, it often means you’ve already lost. You may even be winning against the AI, but the tree clock is ticking. Death waits in the leaves. Trees tend to be a bit of a wildcard and sometimes you actually have to plan for them. In theory, you could use the forests against the AI – instead of taking over tree-infested hexes on the AI side, just leave them to grow because they’re not giving the AI coin anyway. But it’s a very blunt tool. I don’t think I’ve ever won a game through tree herding, but perhaps it helped slow the advance of an AI in some unmeasurable way.

Fortunately, the front lines see-saw so much in Rauk, you don’t tend to see many trees. But they will haunt you on other islands and some cursed variations will put you up against tree problems from the get-go.

Monte Carlo

The more I locked horns with Rauk, the better I got. More advanced tactics emerged.

I realised the AI was slapping a castle on the map because it was scared of me as I had knitted two territories together on turn one. I had a counterintuitive breakthrough to avoid joining the two territories on the first turn. By the second turn, the bastard AI had already expended its resources and wouldn’t be able to buy a castle – and then it was safe to get the gang back together. This was great news and I settled on an arrow arrangement of the two territories in the first turn. This didn’t get me closer to victory, but it strengthened my opening position.

Arrow opening on Rauk

I also learnt it was possible to choose which capital became the joint capital when you merged two territories and that is a choice you should not leave to random chance – either because the capital affords protection or you want the capital to be out of harm’s way. Also, those tiny “unsalvageable territories” that are impossible to save? They can sometimes be used in suicide missions to harass or distract a dominant AI. And if you stick a castle on them, they can be even more effective for launching guerilla operations.

I think the biggest lesson, though, was to ignore Slay’s constant whining that I hadn’t moved all my pieces. Because you really really don’t want to do that every turn, particularly during the early game. If you move all your pieces to shut Slay up, you are going to lose because you are failing at defence.

Yet, despite all this, there remained no killer tactic reveal when it came to Rauk. It was still the same fundamental problem all the time. During the early game, as I expanded in one direction, an AI would nibble my rear, so I spent turn upon turn running to stand still, stuck chasing my tail. A typical session of Rauk felt like a bloody exercise in postponing defeat rather than a bloody grind to victory. More roguelike than strategy game.

And I was a Monte Carlo player, taking a random trajectory every match but always ending up before a brick wall. Rauk was not a game. It was a prison.

Prison Within A Prison

An hour and a half after I stole defeat from the jaws of victory, I conquered Rauk.


It didn’t really feel like a victory more like winning the lottery and claiming your lottery “skills” were responsible. The exhausting Rauk campaign came to an end with a whimper. I just rolled the dice enough times until I ended up with a midgame I could mount a winning campaign from.

It didn’t dampen my enthusiasm for Slay and I continued to play. I continue to play.

Was Rauk the only chaos level like this? No. It was not. “Must” was similar but somewhat worse as it gives you a lot more territories to set up, making for an acutely laborious Groundhog Day. In “Mipi”, I spent most of the match fighting from a weak position and every single turn was desperate edge-of-seat stuff. “Nouss” was a replay of Rauk, a snug island where your fledgling settlements are squeezed into a fine paste.

Nouss, turn 8: I actually won from this position

I was still learning. Tiresome though it was, I acquired a habit of counting hexes in small enemy territories, checking for the likelihood of bankruptcy when launching a risky attack that would leave me exposed to reprisal.

I also noticed that its sometimes better to cultivate rather than eliminate an AI; they may be the only thing slowing the approach of a bigger AI threat further away. Along similar lines, AI castles are just as useful protecting you from a formidable AI on the other side. If you spend effort taking down AI castles too early, you’re asking for trouble.

Another counterintuitive strategy that I’ve used just a couple of times is to sacrifice high-value pieces. If you push a knight out on a limb, the AI will take the bait and cut it off – but why would you do this? Instead of suffering from bloated expenses, you can start over developing a large army of peasants, perhaps the most deadly weapon at your disposal in the end game because they can swamp a large AI. The problem is you’ve usually got your cash tied up in knights to break through castles and thick enemy lines. It’s a very fine line though because losing a single knight means sacrificing 30 coins of investment plus the other AI might think you’re getting hammered and get a few knives in. You have to be confident it will serve the greater good.

Swamping the AI to clinch victory

Another sacrificial play is to concede hexes if it allows you a more manageable front line. Geography is king. It’s better to have the hard walls of a coastline about you than dormant, weak AI territories which are just itching to erupt. Remember, they love the scent of blood. They may not defeat you but they will tie up your resources.

Having learnt so much more, I thought, “Maybe I can do Rauk properly, now? Maybe I’ve learnt enough brilliant strategy to conquer it with smarts.”

The Wrong Question

A few days ago, I fired up Rauk again, to see what I was capable of.

I was capable of failure.


And more failure.

I’ve not managed to take down Rauk a second time. The early game feels just the same. Constant vulnerability, no room to move, no ability to bed in structure without surrendering progress to the AI. Am I now convinced that some of the levels are just chaos engines and there’s nothing you can do except roll the dice or take note of what happens and course correct at appropriate turns?

I quit Mosaic (Corey Hardt & sunil, 2022) because my scores were topping out around 100 but seeing the 200 at top of the leaderboard, I had to wonder if those were rare scores or some players had just figured out smarter strategies. When I played Loopy Wizard (Venbruk, 2022) I was initially building up better scores with every game but, eventually, those scores became a ceiling. Was I a victim of procedural generation serving up harsh levels or was I just not that good?

Loopy Wizard

It’s that fear, you know. Fear that your training is unfinished. Maybe there’s further to travel, you mistook a stopover for the destination. It’s this fear that makes game abandonment so difficult. That what you see as intractable difficulty is just dumb play.

A few weeks ago, I read on the Slay forums that someone pretty much knew what the AI was going to do all the time. And I wondered: is that where I am now? Or is that where I could be? All that post did was stir up doubt.

How can I be scientifically sure that Rauk’s chaos cannot be understood nor managed? Maybe it all makes sense to the right player who plays a better game, whose peasant movements are decisive, confident.

Rauk had never been a battle over the nature of Slay. It was, and had always been, about me.

Footnote (added 13 June)

  • There are four difficulty levels – the lower levels are just too easy for me, so I’m playing against the toughest AI level. However, the game can accommodate if you want to enjoy a less stressful Slay.
  • This is something I wasn’t sure about originally but Slay is entirely deterministic. If you perform the exact same moves, the AI will rattle out the same responses as last time. This is why you can learn the AI responses from game to game and why I talk of “mathematical chaos” which is complexity that emerges from a few simple rules. Levels like Rauk seem highly sensitive to small movements while other levels are stable and perturbations in your choices don’t ultimately impact the shape of a game’s future.

Download my FREE eBook on the collapse of indie game prices an accessible and comprehensive explanation of what has happened to the market.

Sign up for the monthly Electron Dance Newsletter and follow on Twitter!

7 thoughts on “The Battle of Rauk’s Drift, Pt 2

  1. These were a lot of fun to read!

    I’m tempted to give the game a go myself. But on the other hand, I’m also not.

  2. Thanks Shaun. It’s a terrible time sink but, unlike some time sinks, I feel like my brain is being used rather than diverted. I’m not sure these skills will get me a job, though.

    I’ve killed off two long-time stumper islands today – Rytt and Runou. I’m definitely sharper at play now; I don’t make silly mistakes and I’m more sophisticated at encouraging AI-on-AI action. But Runou only seemed possible if I played it around three times, watching where problems materialise and countering them ahead of time in the next game. I guess I can live with that. I’ll see where we are with Rauk again in a couple of weeks 🙂

  3. Complete side note. Last Friday I decided to use Cheat Engine and enable House Rules (accessibility settings, allowed me to set enemy HP and damage dealt to half) in Rogue Legacy 2 to get through the last third of the game as quickly as possible.

    The reason was i was already ~70 hours into the game, I was tired of it, I wanted to stop playing but because of how my brain works I couldn’t just stop. I’d keep thinking about the game for the next few weeks, and the best option is to just get through it asap. So I did.

    Anyway, what I am getting at is that I think I am no longer interested in roguelite games. I had my fair share of fun with Zangband and Hengband and Binding of Isaac and Dwarf Fortress. But I no longer have the time for these games and they tend to occupy too much of my head space.

    And if you add meta-progress on top of that, like Rogue Legacy does, it becomes an actual negative experience overall in the long run. Though I think if the game allowed faster progress (rather than taking full 8 completions to get to the end of story) maybe my experience would’ve been different.

    Anyway that’s just a thought I had after reading this post. Slay looks fun but I am glad I am not into strategy games too much so that I don’t feel like playing it :).

  4. All Joel is Good Joel, but stuff like this is Best Joel. Part I had me hooked. Being of similar age we both have memories of games like this, and it’s nice to see they’re finding new life on mobile. It’s especially nice to see KIDS THESE DAYS engaging with them, finding them as fascinating as ever. It really puts a pin in the idea that older games might struggle to resonate with younger audiences whose members are used to modern production values, whether that be higher-quality visuals, more hand-holding, or whatever. Good games are good games.

    What really stood out for me was the message that Slay is like a master class in mechanics design. Everything works in a certain way for a reason, and the way everything works is carefully interconnected. Nothing exists as a mechanical island that stands apart from the rest. Trees do X, which affects Y, sometimes in unexpected way Z. To master the game you have to not just understand the rules, but develop a feel — almost instinctive — for how things will affect other things in fancy causality chains.

    That said, Joel, if I was reading correctly, it sounds like Rauk at least had an element of random luck associated with it. Is this so? I wouldn’t fault a game too much for a smidge of randomness; it can enhance replayability and encourage players to try different approaches and hone strategies. But it sounds like it could get pretty annoying if over-done.

    The AI sounds merciless in a way that AIs often aren’t programmed to be (I have an analogy involving Valkyria Chronicles that I’ll leave out since this comment is already too wordy for something that says nothing). Frustrating as ruthless AI can be, here it clearly drove you to evaluate new approaches and improve your game. The most gratifying victories are not just hard-won, but made possible by learning and expanding your arsenal of strategies.

    Thanks Joel!

  5. Maurycy You just reminded me of my reaction to Lone Survivor where it is only possible to understand the story if you see multiple endings and their only possible to reach if you play the entire game in a certain way. i.e. you have to replay. And I didn’t like the game enough to put myself through that. I’m always torn about whether you truly want “replayable” games and whether developers should actively reward replays because there’s a danger you’re not doing it out of personal enjoyment (Overjustification Effect/Into the Black again).

    Steerpike Thanks! I find writing these days much more punishing than it used to be but I still love the high of reading my own words back to me. I made this! I love this sentence! etc. etc. And these journal-type affairs have to be done very quickly after the event otherwise the energy goes cold.

    Slay really is a bit of a marvel. There are aspects that I could criticise: the chaos levels; some of the UI; the AI never surrender early enough when it’s clear you’re past the point of no return. But the same game design for pulled me in during the late 90s still pops.

    And your comment reminds me there are two points I never clarified.

    • First, there are four difficulty levels and I’m playing the highest, Master. The lower levels are just too easy for me. But the game can accommodate the level you wish to play at.
    • Second, and this is something I wasn’t sure about originally, but Slay is completely deterministic. If you perform the exact same moves, the AI will respond the exact same way. This is why you can learn from game to game and is why I talk of “mathematical chaos” which is unexpected complexity borne of a few simple rules. Levels like Rauk seem highly sensitive to small movements while other levels are stable and perturbations in your choices don’t ultimately impact the shape of the future.

    I’m going to add these points to the article as a footnote.

  6. I forget I played this game after you wrote those great pieces about your experiences with Slay, got to play this scenario, and took me hours just to play it right. I know, I know, I jumped right to the last difficulty, but after It seemed to me too arbitrary, many lost battles later, I kinda lost my jam and got away, I wanted to play later to put some thoughts together to reply, but never more, I forgot completely, maybe someday. Also didn’t get a feeling it was the merit of the A.I., but rather the conditions previously set on the map, I guess someone can make a social commentary about that, won’t be me. I imagine based on this kind of chaos you mention, the more variable something is or more variables are in the “initial conditions”, more chaotic it is, i.e. we can’t predict that it is always have the same result, can’t say, because I don’t know anything about that, but If I would do a bet I would say the possibilities of winning in this case are more due to the chaos of reality itself from which we play this game, yes the same whole reality we live, where e.g. the script of the game will run some of its aspects randomly enough to give some room to breath for the player, not because of the script itself, let’s say that something else is letting this script be run one way or another. Also seems more as an error of the A.I, on purpose or not, since it feels so mechanical, like many A.I tend to play, possible with a potent A.I. like those last ones from the Chess or Go it would be almost impossible to beat, if the map isn’t absolute either. On the same time the opportunity to be mistaken, is an opportunity to creativity, one of the differences between humans and these A.I., we can lose even if we can win, for example just for psychological reasons, so to speak, but this creates a lot of opportunities to think and act outside a determinist scheme. If the world is not determinist, then this A.I. would have to act differently maybe one in trillions of times, no matter how powerful it is, maybe because of an electrical change in the circuits in the PC, maybe any other thing, maybe never at all. But if we do live in a determinist world, then there are no variables and it is all just an illusion, since it is all decided no matter how powerful we or the A.I. are. We just have to keep playing or don’t play at all, either way the result is the same, it’s just that we don’t have this knowledge yet, in this case the one way to achieve this or maybe the only way we can achieve this knowledge, is through experience. Therefore the experience of play would be how we play the game, in a certain way without ever playing it, because we don’t play and then receive experience, we receive experience and then play, or have the illusion of play, since we in fact receive the experience of playing. “God” doesn’t play dice with the universe like Einstein said or he not only plays dice, but alongside a variety of games, he’s a gambler too? Being the player playing the game the most important variable, the result will always be the same, until someone plays it? Is this a deterministic prison or a prisoners’ game? I have no idea which is which, but it is an interesting reflection. Game Design and Metaphysics, how about that.

  7. Pedro,

    It’s interesting to frame the difficulty of Slay just as “initial conditions” which proflagate chaos and others which are more predictable. It is true that the AI will play out the same way each time and, I guess, that’s part of why I think the brute force solving of the more chaotic levels feels less good. That you haven’t bested the machine – you just watched its moves and plotted some appropriate counter-moves because of prescience. Then again, I’ve had some more thoughts about Slay and there *may* be a third part, more of an appendix.

    Thinking of the bigger picture. Arthur C. Clarke said any sufficiently advanced technology would be indistiguishable from magic. Maybe any sufficiently complex deterministic system is indistinguishable from non-determinism. Maybe it’s only in these small games that we can feel the difference while, out there, it doesn’t matter. Especially if you feel trapped in a economic prison, earning a low wage with no potential for advancement…

Leave a Reply

Your email address will not be published.