Your competition is not the games launching this month

Steam grossed $11.1 billion in the first half of 2026, and only 21% of it came from games released that year. All of the rest came from the backlog. So your game is not competing with this year’s games, it is competing with a discounted decade.

People usually think the real competition is the games coming out this month. They look at it as “these are releasing, I will squeeze in among them”. But the game releasing in the same week as you is not your problem. The money goes to hundreds of old games the player still has not bought and that land in front of them again at every sale.

The bad part is that new games are shrinking in total. In the first half of last year new releases did $2.79 billion, and this year it came down to $2.40 billion. So it is not only their share that fell, the money they made fell too.

The reason is the number of games coming out. 9,265 new games were released in the first five months of 2026, against 7,800 in the same period last year. Zukowski calculates the year will close at 25,799 games, where 2025 was 20,853. That is 23.7% in one year.

Do not let that number scare you though. Around 75% of what comes out on Steam comes from people releasing a game for the first time or with no commercial expectation, and that pile is not your competition. Look at this distribution in particular: in the first quarter of 2026 there are 96 games that passed 1,000 reviews, and 3,686 games that got between 0 and 9 reviews. For every game with 1,000 reviews there are 38 that are never seen at all.

The money is coming down from the top too, and the conclusion here is not that indie is finished. While the top 10’s share of the platform fell from 41% to 28%, the revenue of the band between 101 and 1,000 went from $1.1 billion to $4.5 billion. More than four times.

But the number of releases outrunning the revenue growth has a price, and that price is charged to the long tail. The sentence “if it does not land in the first week it will recover in the long run” collapses right here.

In GameDiscoverCo’s July data, the copies a game sells in its first year are 2.46 times what it sold in its first week. In 2021 it was 3.05. It has shortened by 20% in five years.

Do not look at the average, look at the bands. A game that sells between 1k and 10k copies in the first week does 1.96 times in the first year. One that sells between 10k and 100k does 3.16 times. One that sells over 100k does 4.64 times. So the long tail does not arrive the same for everyone, it multiplies whatever you did in the first week. And a game that does not behave well at launch has almost no long tail at all. 1.96 times does not sound bad, but for a game that sells 3k in its first week it means 3k copies spread across a whole year.

There is one more number on the price side, but read it carefully. Games under $10 stay at 2.32 while the ones above it do 2.7. Do not tie this to the price decision. Cheap games are usually shorter and smaller games, and the shortness of the tail comes from there. Also all of this is in copies, and the revenue multiplier is lower still, because you sell that tail on discount.

People buy those old games but do not play them. 16,565 pre-2024 games were looked at, and while backlog spending rose 26%, the average daily players per game fell 10%. I built Dustpile on top of this pile in the first place, a free tool that shuffles the backlog you bought and never played like Tinder. So the decision to buy has come loose from the intention to play. The old game competing with you does not even have to promise the player time, being on discount is enough.

Even so, Zukowski called this a golden age last week, and he is not wrong. Sir, We Have an Orc Problem, which came out in July, was made by two people in four months. $361k in the first 24 hours, 200k copies in 20 days. The total they spent on paid marketing was $200. Two weeks after the announcement it had 3k wishlists, and by launch it was at 150k.

The two pictures do not contradict each other. The backlog is not eating new games, it is eating new games that cannot be told apart. A middling production at $70 cannot compete with last year’s Game of the Year at $25. But an idea you understand in one sentence never enters that race.

Build your plan on the first week. Set your first week copy target and multiply it by your own band’s multiplier. Do not use the average. If you are in the 1k-10k band your multiplier is 1.96, not 2.46.

I have done this calculation before, and I will keep doing it until everyone reading has learned it. Let us open it up. Your game is $7 and you sold 5k copies in the first week. The launch discount is mandatory, say 10%, so $6.3 per copy and $31.5k in the first week. Your band is 1.96, so your first year total is 9.8k copies, 4.8k of them coming from the tail. That tail goes on discount too, say 33% on average, so $4.7 per copy, another $22.5k. Gross at the end of the year, $54k. Steam takes 30% of it, you are left with $38k, and tax has not come out yet.

Now look at where that 5k is going to come from. In GameDiscoverCo’s scan a median 22% of launch wishlists convert in the first week, and the band I have seen across 20+ games is 15-30%. Going from 22%, to sell 5k in the first week you need to go into launch with roughly 23k wishlists.

Put that number next to the Next Fest picture. The median game takes 806 wishlists from the fest, and the top 5% pass 13k. So you cannot collect 23k from a single fest, that figure is a job spread across a year.

Now put those two next to the cost of the project. If it does not cover it, your problem is not the long tail, it is the launch. Most people learn this in their second year unfortunately, when it can be worked out today.

The long tail is not a safety net, it is a multiplier. However much you multiply zero, it stays zero.