Don’t mistake bug fixes for content

In April I shipped ten patches and Metal Thunder’s average CCU was the same after that run as it was before it. A year earlier I had shipped eleven patches, and after that run the same number went up and never came back down to where it was. The difference is not how often I patched, it is what was inside them.

From the outside the two runs look the same. Same guy, same game, almost the same number of patches, and the second one squeezed into a much shorter stretch of time. On the Steam page both of them sit there as announcements lined up one after another.

So do not use the sentence “I am keeping the game active, I ship something every week” as proof to yourself. That only tells you how much you worked, not whether the player came back.

Over nine weeks I put in things that changed how the game plays. I reworked the difficulty modifiers completely and handed all of them to the player, from enemy count to spawn rate, from health to armor. Endless Mode came in, God Mode came in, a mode where they can turn the UI off entirely came in. In Endless I made the difficulty climb over time and added modifiers with no upgrades and no leveling. I put a prestige indicator on every weapon and opened up picking your weapon when you go into a sortie. I did most of it because players asked for it, and in the patch notes I write “you asked for it”.

In April 2026 I shipped the big update and then spent ten days closing the holes I had opened myself. Helicopters had ended up immortal, I had forgotten an old armor flag. Because of the save refactor the player’s lifetime progress had been wiped from the menu, they could not earn money, winning the game softlocked it, the squad would not take orders, accuracy showed 120%, and there were people who could not fire on Steam Deck. Some of it was things I broke while adding new things, and some of it, like the save refactor, was maintenance work anyway.

Separate the two by their content, not by their count. Four of the eleven patches in 2025 put a new system into the game. Two of the ten patches in April did, and one of those two was the big update that caused the other eight in the first place.

The point is this, a bug fix brings the game back to where it should have been and stops there. It gives the player no new reason to come in, because it promises the thing that should have been there anyway.

The number I call baseline is the average CCU of a quiet day, a day when nothing is happening. You are not on sale, nobody is talking about you, and you have not put anything new in. However many people are inside on that day is your game’s real size, because what you are measuring is retention.

I saw the same thing twice with discounts. Last May, on my first Daily Deal, the peak was 431 and that month’s average went up to 81, but by August it was down to 8. This year the Daily Deal in June was the cheapest the game has ever been, my peak that day hit 198 and the highest of the previous month was 18. My quiet day average in the month before the discount was 7.78, and two months later, today, it is 7.24.

The discount in May came right after the patch run, but that is not what carried the floor up. My average was 20 in February and 19 in March, so it was already there while the discount was nowhere in sight. A discount gives people a reason to buy, not a reason to come back.

On DriveWave I see the opposite of this. The game has not had a single spike to date, its highest peak is 37, and since it is free to play you should not compare its CCU level with Metal Thunder. The thing to look at is not the level, it is the direction the curve is going.

In October when it came out the quiet day average was around 19. In December I took my hands off the game, all my time was on Metal Thunder, and no update went out in January. That month the average fell to 7.72. In February I came back, the handbrake came in and the cars got easier to drive. In April the per-car upgrade system and the garage hub came in. In May I wrote the controller pipeline so that it works even when another game is in front. In July wheel drops actually started dropping into Steam Inventory. And in August I gave people the most requested thing since launch, Skydeck went fullscreen. Today the average is around 19 again.

Every time I left the game alone the floor collapsed, and every month I put something in it came back. And this is not even a growth story, the game only returned to its own launch level. Getting back what you lose when you take your hands off sometimes takes you months.

Putting one single new thing inside the fix run you had no choice about usually changes this. There was one in Metal Thunder’s April run, in 713 I reworked direct hits and put in a new prestige feature. That was the only thing in that run I could tell the player about, the rest was the news that I had fixed what I broke.

Every net positive contribution you make raises the baseline in a compounding way. But the net positive part is the thing you added to the game, not the part where you repaired what you broke. The two sit side by side in the same changelog and behave differently on the graph.

Ask one question before you send an update. Is this something that will make somebody who has finished the game open it again, or does it remove a problem they would run into when they open it. Both will get done, but only the first one goes onto the baseline.

When you measure, compare a quiet week with a quiet week, not the discounted week. Take the average of the month before the update and look at the average fifteen days after it. If it settled above the old one you gained players, and if it came back to the same place you got the players you already had to play for one more week.