A game that launches badly does not recover on its own

At the end of January 2025, Metal Thunder was selling 5 copies a day, and three people in total wrote a review that month. That is where I made the decision to keep going, and I made it knowingly.

What I was looking at was not the sales graph. Whenever a video about the game went up, conversion was very high, whoever saw the game bought it. That told me the problem was not in the game, it was in exposure, and exposure is a solvable problem whenever it happens.

I took a gamble, but not a blind one, I had reasons to believe. The sentence most people build at that point is this, “the first month went badly but I will pull it together when I ship an update”. If there is no signal behind that sentence, it is only a wish.

Through January and February I shipped update after update. February brought 22 reviews, March 36, April 23. May brought 144. May was my first Daily Deal month, and in that single month I got more reviews than the previous six months put together.

So the signal I read turned out to be right. But among the games that came out in 2024, I am the exception. I only learned that recently, when I pulled my own reviews out date by date and set them next to one of Zukowski’s scans.

He looks at every game released on Steam in 2024. His measure is this, if you got fewer than 150 reviews in the first 30 days your launch counts as weak, and if you had passed 500 reviews by February 2026 you count as recovered. My first 30 days were 37 reviews. November 37, December 8, January 3, so 48 in three months. The game was no different from a ghost.

The scan’s second measurement day is 1 February 2026, and on that day I had 408 reviews. The threshold is 500. I missed the measurement by 92 reviews, and I passed 500 on 10 June 2026. So I recovered, but on the day it was measured I had not recovered yet.

The number of games that made the list is 28. Let us open that up, 28 of the roughly 18k games released in 2024, which is 0.156%.

I looked at those 28 one by one, and the turnaround of four of them is not real. Look at these four in particular, all of them collected their reviews with a free weekend. Rat Quest has 1,474 reviews, and only 17 of them come from someone who paid for the game. WHAT THE PAK?! 1,255 reviews, 48 paid. Caribbean Crashers 1,337 reviews, 37 paid. A Story About Farting 1,044 reviews, 29 paid. That is not recovery, that is handing the game out for free and piling up a review count.

And one of them is outright fake. SPY-der PIG got exactly 100 reviews every month, and the playtimes of the people writing them are identical to each other. This goes out to the people buying bots, every review and CCU bot you buy can be spotted by anyone who so much as looks at SteamDB. No game has a baseline and a peak that run in a dead straight line.

So the number of games that actually turned around is not 28, it is 23. One in roughly every 780 games released in a year.

The genres of the ones that turned around are not random either. Of those 28, 21% are horror, 14% friendslop, 11% adult. Almost half come from three genres, and what the three have in common is that people search for them independently of launch week. The player finds horror by seeing it from a creator, buys friendslop because a friend called them in, and on the adult side there is already an audience searching by tag. In the remaining genres you never find the player you missed at launch again, because they depend more on the Steam algorithm and hauling water in from outside rarely turns them around.

When you look at the conditions set for attempting a rescue, the heart of it comes out. If your Coming Soon page was up for less than 6 months. If you did not put out a demo at launch. If your median playtime is above 2.5 hours. If your genre is open to virality, like horror, adult or friendslop. If you reached fewer than 300 creators.

There is one item on that list about the quality of the game, and that is playtime. All the rest is work you did not do at launch. So a rescuable game is not a bad game, it is a game that was marketed short. If you did everything and it still did not sell, there is nothing there to rescue.

Do not read this as Steam punishing you either. I said the list of things people wrongly believe has an item on it, “if you launch badly Steam buries you”, and this scan confirms that. Steam gave the impressions, the player did not get excited. The result is the same but the reason is different, and the reason matters, because if you chase the wrong reason you lose a year.

There is also this. On that measurement day I had sold 40k copies. Today I am at 60k and I have 589 reviews, so one person in every 100 copies writes a review. Measuring recovery by review count counted me as late. It gets used because it is the only measurable thing, but when you are looking at your own game, take it into account.

For those 19 months the game was not left to itself for a single day. Under every jump I see in my review graph there is either an update or a discount, none of them came on their own. If I had chosen to wait in January, those three reviews would have been my last three reviews.

A rescue attempt costs you not only money but months, and your scarcest resource in this work is time. In a possible failure the biggest setback comes from exactly there, not the money you lost, but the game you could have made in that year.

My own criterion for the decision is this. If you got enough signal from players you keep going, and if you did not, cutting your losses at any point is a gain. Write this before launch, it does not get written during launch week. Set a date, and say now what number you will look at on that date. When that day comes, do not look at the sales graph, look at how many of the people who saw the game bought it. What happens when a video goes up tells you far more than the sales figure does.

If you do meet the conditions, you have one move. Ship a patch and put a small piece of free content next to it, but do not publish it right away. Spend a month reaching out to creators, put the patch on the same day as a discount, and pull the price at least 10% below your previous historical low. Look again a month later, and if it did not work, move on to the next game. As for a free weekend, do not think about it at all before the game is two years old.

Not one of those 23 games turned around on its own. They did the work they skipped at launch, a year late.