Thoughts before releasing the best game I've ever made

In November 2024 I started working on a roguelike game. It was supposed to take 2 months, I'd release it on Steam quickly and move onto the next project. In typical gamedev fashion, it took me x3 the time, but it's finally done and ready to release on May 1st.

The aim of this post is threefold:

1. Document data about my game before its release.
2. Provide some useful insights based on that data so other game developers can learn from my mistakes.
3. Present a case study for a game that has a marketing budget of $0 and zero external visibility.

It's easy to estimate initial sales when you have thousands of wishlists on Steam, but I have neither the skill nor the luxury of time for gathering those wishlists. Many other reasons for this, which are all beyond this post so I will not digress. My game is now hovering around 85 wishlists and the store page has been up for 4.5 months. Despite this poor performance, I estimate that the game will do ~1.5k sales in the first year at $5 per unit. The data and my reasoning for this conclusion are detailed further below, but before that, a description of the game itself.

The game is Glass Cannon, where you shoot different bullets which can have mods slotted in them. Each mod interacts with the environment and other mods in an interesting but simple manner, so ideally, players can make interesting builds based on combinations and emergent synergies to clear all 24 rounds. In between the rounds you get an upgradable shop with rerolls and a random weapon you can also roll for free.

My idea for the game is to allow for glass cannon builds (thus the name), to get players to invest in some crazy gimmick combination which they drive to its complete conclusion and win the game by making that gamble. There is some level of risk mitigation involved, where you have to make sure you don't die from your overpowered weapons yourself. Some bullets can knock you around on impact or the recoil from your shots can throw you across the arena and into enemies. In simple terms, the game's build construction consists of 20% risk mitigation and 80% YOLO theorycrafting. That's the exact amount I find the most fun.

Inspiration

Of course, I played a lot of Balatro last year and that inadvertantly impacted my game design (in ways that I'm not even cognizant of probably), but it's not exactly the kind of game I enjoy most. It seems to me that it's too much about risk mitigation. If you want to win, you have to reign in your YOLO build and make space for backup jokers which can get you chips if you happen to get an unlucky boss blind that debuffs your core combo. The runs I enjoyed the most were those where my gimmick didn't get countered by an unlucky boss blind and I was able to see the limits of the combo I came up with.

By far the biggest inspiration was playing gimmicky murloc decks in early Hearthstone, around 2018. Hearthstone is a card game by blizzard, which I think has valuable lessons for any game designer. Back in its early days, when players hadn't found the optimal way to play, all sorts of innovative card combos used to show up. My favorite cards were murlocs, which are small fish-like creatures with low stats.

Back then murlocs used to be weak on their own, but they would get powerful very quickly if you played them alongside each other. Almost every murloc has some kind of buff it can give only to other murlocs. When you have only one other murloc on the board, that buff is applied once, but when you have 6 murlocs, that +2/+1 buff becomes +12/+6, which was bonkers crazy back in the early Hearthstone days.

This deck was ineffective against some strategies that had early-game removal, but it was completely dominant if you could get it to start rolling. Inventing, crafting and then playing my shaman murlock deck was the kind of fun that I'm trying to replicate with Glass Cannon.

Redesigned the game, but still same results

I did a beta release of Glass Cannon on Februrary 2025 and got around 20 people to play it. I concluded from the playtesting that the game lacked content, balancing and variety, so I delayed the launch by 2 months to fix it.

In those 2 months, I completely redid the progression, added bosses, 120 different weapon mods, made interesting spawning patterns (enemies can be sparse, huddled, arranged in a line, layered in fortress-like structures etc.). I basically rewrote everything except the graphics, physics and UI so that the progression of the game was challenging and required the player to actually make interesting combos to win each round. You can't win the game if you just buy flat stat mods, you have to combine weapons or mods in an interesting way so that the result is "larger than the sum of its parts".

2 months later, on April 1st, I did a big beta v2 update and new players have had 20 days to come in and play the new version. The results show that I basically wasted my time. Even though I thought the game was significantly better, the redesign only yielded a slight improvement as you'll see in the playtime and feedback data further below.

One saving grace is that my data may be insufficient. For beta v1 I have 11 data points and for v2 I only have 9 because I can't seem to get the game into enough people's hands. Reddit keeps deleting my posts where I try to spread beta keys, I'm shadowbanned on X (my comments are literally immediately deleted and my tweets get 0 views) and my Youtube channel has 0 reach. The only 2 sources that worked were dumping keys on 4chan and sending free keys to some people in gaming Discords. Now even 4chan is gone.

My small reach can absolutely be remedied given enough time, but I'll save that for my next project. It should be my number 1 priority to get the game before launch into people's hands more frequently and earlier, before I decide to invest more time into it. Even if it's janky, unpolished and ugly, you should always be seeing people's reactions to your stuff because...

Players will put up with jank if your game is good

Steam drove lots of people to Gnomber, my first release, but the game's major design flaws prevented it from growing fast. The Steam reviews on Gnomber say it lags, but the actual reason why it has mixed reviews is that the game is not that fun. The lag was present for some time but I completely fixed it with a big patch. The reviewers never bothered to update their reviews, so the game stayed at mixed and was thus throttled by Steam's algorithm. I don't fault the reviewers at all, their feelings are correct - Gnomber is not a good game. If it had been a good game, they wouldn't mind the lag and would've kept coming back to the game even without the lag fix. They'd even go as far as to update their reviews after the fix.

To illustrate my point further, look at some of the early reviews on Nubby's Number Factory, a game that's sitting at 5k reviews as of writing. This is when the game had crashes and major bugs that on the surface seem like would significantly degrade the experience:

Also note the playtime of the reiewers: 17 and 44 hours. You can also see some of the first few streams of NorthernLion playing the game and experiencing bugs and UX issues, but he kept playing the game regardless. Clearly, people are willing to overlook jank if the game is fun. And Nubby is a lot of fun.

I internalized this maxim a long time ago because I used to be involved with the startup world building SaaS and mobile apps for a few years. The mainstream general consensus among tech startup people reflects this discovery: if your MVP is useful, it doesn't matter if it's janky and ugly. I would say that this rule generalizes for many things including businesses, governments, friendships... It's basically the same discovery that evolution selects for good enough as opposed to perfect.

In simpler terms, people dislike the jank in Gnomber because the core of the game itself sucks. People didn't mind the jank in Nubby becuase the core of the game is really good. A 100% improvement in Gnomber's performance did not remedy the situation at all.

How do I know the game is good?

I don't. I have no grounds to believe that Glass Cannon is better than Gnomber. In fact, it may perform worse, leading me to believe that I've basically wasted dev time on this. Still, I don't know if that is reasonable to conclude with data from ~50 mildly interested players as opposed to Gnomber's 1000+ players who were into it enough to pay $3 for it.

Honestly, I don't enjoy playing Glass Cannon anymore as much despite the fact that it was designed according to my tastes. Having worked on the game, I already know basically every possible viable build in it (around 300 of them), so the YOLO theorycrafting part of it is completely spoiled for me. Because of this, it's really hard to tell if the game's fun anymore. I didn't have this problem with Gnomber as much, as it relied more on mechanical skill than Glass Cannon, so I could tell that there was some semblance of fun in the game which proved to be true judging by its sales.

Right now, only ~25 people have played the final version, 9 of which have left feedback. The feedback itself is mostly positive, but not great either. Here you can see the self-reported and Steam-reported playtimes for beta v1 and beta v2 for comparison.

As you can see, a slight improvement in Steam's playtime, which could realistically be attributed to just more players and therefore more diversity in playtime. Even the self-reported playtime reflects this: more people who play either a short amount of time (<1 hour) or a long amount of time (5+ hours). Compared to Gnomber's numbers, it's still lagging behind:

Expectations

Gnomber's playtimes are better in general because more than 1000 people played it and a big percentage of them got recommended the game by Steam, which means they were the types of people who enjoy similar games. By comparison, Glass Cannon's players were just random people from the internet, so it's likely that a big portion of them was only mildly curious and willing to burn 5 minutes for a free beta version of a game.

Rougly 50 people have played Glass Cannon and only 4-5 of them have played for longer than 5 hours. If the game sells worse than Gnomber, I can reasonably conclude that 50 completely random players are enough to test the viability of the game, so I won't actually need to improve my reach. I will add a link to the post discussing the conclusion here at a later time.

On a more positive note, I got this from the feedback:

Future plans

If the game sells well, I'll keep updating it with content, but seeing that that's unlikely due to the stats, I'll probably just jump onto the next project directly, but this time I'll start marketing it from day 1 and really limit myself to 2 months. The goal for me right now is to grow as a game designer so I can consistently deliver good games. I believe I still achieved growth with Glass Cannon, as I did some of the deepest design work of my life on it, but I could've shrunk the dev time and released earlier with lesser content and still gotten the same amount of learning out of it (and probably sales too).

I still have 10 days until release, so I'll probably sneak in a small challenge mode just to test the waters. In case it sells, the players can see what future updates may hold.

I should have tested the game with more frequency and with more people so I could've come to the conclusion that this game would do a measly 1.5k sales sooner. Had I done that at the 2nd month, it would've taken me 1 more month to release, so I would've gotten

 $5 * 1500 * 0.7 = $5250 

If you are a game developer, I think your goal should be to have around 100 people willing and ready to play anything you make. That way, you have a good sample size to judge if the alpha version of your game is good enough to invest more time into. In addition, you'll have 100 people willing to do word-of-mouth marketing for you once you do release, which I believe is completely enough for making even the biggest hits. Just look at Notch's post history on TIGForums. The game took off immediately because its core was so compelling. The forum definitely had under 100 people discussing the game in its early days.

I should also get into the habit of documenting the pre-release stats of all of my games. That way I can compare and judge to see if anything new I'm making is worse or better than my previous stuff. This post is a step in the right direction.

For now I want to see what Glass Cannon can do with zero advertising. I believe it's better than Gnomber and I want to see if it manages to surpass it. I will post updates on this blog when I release it and do follow-ups in the first 2-3 months, if it'll still be relevant to do so.

Here is my post documenting the game's release stats:
Glass Cannon Release Log