# Demo case: Shift At Midnight

> Markdown version of https://playrecon.com/demo for AI agents. All pages: https://playrecon.com/llms.txt

**Question:** Shift At Midnight shipped on 22 July and reached #1 on the Chinese top sellers within four days, with concurrents still climbing. What did it do that the games standing next to it on the same shelf did not - and what is it losing people to?

Game: Shift At Midnight (Steam appid 3722330). Generated 2026-07-26. Run: 7 steps, 14 tool calls.

Every fact below is a query against our own collector: storefront placements, hourly concurrents, price observations, Steam news and store-page diffs. Nothing here is invented. Where the record cannot answer, the case says so instead of estimating.

## Summary

The pre-release shelf did not decide this launch: seven of the 26 games standing on it never reached the top sellers, and one of the three games that held that shelf at #1 finished twenty times smaller. What separates Shift At Midnight is a year-long demo campaign, a publisher that appeared on the page two days before launch, a $9.99 price with a discount that put it on a second shelf, and tag pages that moved before the charts did. Concurrents are still climbing on day four - this is accumulation, not a spike.

### What the record shows

The game entered Popular Coming Soon on 19 July at #12, with the rank identical across all nine countries we poll [f3] and moving in lockstep the following day [f4] - that shelf is global, so nine observations are really one. It entered the top sellers across 41 countries at 15:35 on 22 July [f8], reached the Steam front page an hour later with a #1 featured slot in Turkey, Indonesia and Brazil [f9], and took the Chinese top sellers on day four rather than day one [f12]. Concurrents rose every single day of the window, from a 17.3k peak on the 23rd to 37.6k on the 26th [f11].

### The hypothesis that failed

The obvious story - that the pre-release shelf carried it - does not survive the cohort. Twenty-six games shared that shelf in the same window and seven of them never reached the top sellers at all [f14]. Three games held Popular Coming Soon at #1 and finished #1, #1 and #6 in the top sellers; Dinoblade, the #6, peaked at 1.8k concurrents, while this game sat at #3 and peaked at 37.6k [f15]. Shelf position ordered almost nothing about what happened next.

### What actually separates it

Three things the record does show. First, the campaign: a singleplayer demo in June 2025, a multiplayer demo that September, an updated demo for Steam Next Fest in February 2026 [f1], and a release date that moved twice before being fixed twelve days out [f2]. Second, the page changed under it right before launch - a publisher appeared two days before, with a Discord link added in the same minute [f5], and the description was rewritten four hours before release [f6]. Third, it landed at $9.99 with a 10% launch discount [f10], which also placed it on Specials, and at 15:04 on release day it topped both of its tag pages [f7] - half an hour before it appeared in the top sellers at all.

### Where the friction is

The reviews put the problem in a specific place. Under one hour played, only 61% are positive; from one hour onwards it is 90 to 94% [f17]. That is not a game people dislike - it is a game some people cannot get started. The negative texts name crashes in three-player lobbies and a softlock while restocking the store [f18], which points at a technical first-hour failure rather than a design one. Nearly three quarters of the corpus sits in the 3-10 hour band, so whatever is breaking early is not breaking for the people who get through it.

### What we will not tell you

How many copies this is. There are now 3.0k reviews [f16], and the industry habit is to multiply that by something between 30 and 70 - but that multiplier is not calibrated on our own cohort, and a number carrying a two-fold spread teaches you nothing you did not already know from the concurrents. Wishlists are not public. And none of this is causal: the publisher appearing, the price, the tag pages and the demo campaign all sit inside the same 48 hours, and this record cannot separate their contributions.

## How the agent got there

1. **Reading the question** (scout). Shift At Midnight (3722330). The release falls inside our observation window, so the storefront, the charts, the prices and the page diffs can be read directly. Steam news reaches back to June 2025, so the campaign before the window is visible too.
2. **Rebuilding the campaign** (timeline). The demo is the spine: a singleplayer demo in June 2025, a multiplayer demo that September, an updated demo for Steam Next Fest in February 2026. The release date then moved twice before the final one was announced on 10 July, twelve days before launch. Facts: f1, f2.
3. **Reading the shelf** (timeline). On 19 July the game enters Popular Coming Soon at #12 - and the rank is identical in all nine countries we poll, moving in lockstep through 11, 12, 9, 10 the next day. That tells us the shelf is global, so nine countries are one observation, not nine. Facts: f3, f4.
4. **Hypothesis: the shelf did it** (hypothesis). If the pre-release shelf drives the outcome, games ranked above it should have finished above it. Pulling every game that shared that shelf in the same window. Facts: f14.
5. **Hypothesis dropped** (dead end). It does not hold. Of the 26 games on the shelf, seven never reached the top sellers at all, and Dinoblade - which held #1 on that shelf - peaked at 1.8k concurrents against this game's 37.6k from #3. Shelf position is a weak predictor at best. Facts: f15.
6. **What separates it from the cohort** (hypothesis). Three things the record does show. The page changed under it: a publisher appeared two days before launch and the description was rewritten four hours before. It landed cheap at $9.99 with a launch discount, which put it on Specials as well. And at 15:04 on release day it topped both of its tag pages, half an hour before entering the top sellers - the tag shelves moved first. Facts: f5, f6, f10, f7, f8, f9.
7. **Reading the reviews** (hypothesis). The review set did not hold this game at all, so it was collected on demand: 3.0k reviews across the first four days, 89.7% positive. Splitting them by hours played puts the whole story in the first hour - 61% positive under an hour against 90-94% everywhere above it. The negative texts name crashes in three-player lobbies and a restocking softlock. Facts: f16, f17, f18.
8. **Assembling the case** (verdict). Concurrents are still climbing on day four, and the Chinese #1 came on day four rather than day one. This is accumulation, not a launch spike. Units stay unstated: 3.0k reviews is a real number, but turning it into copies needs a multiplier this record has not calibrated. Facts: f11, f12.

## Facts

Each fact is labelled observed, estimated or inference, with its source and the date it was recorded.

- **f1** [observed] A year of demo campaign: singleplayer demo 28 Jun 2025, multiplayer demo 29 Sep 2025, updated demo for Steam Next Fest 23 Feb 2026 (news (Steam events) · 2026-02-23). Ten more posts sit between them; the demo is the spine of the campaign
- **f2** [observed] Release date moved twice (28 May, 8 Jun) before the final date was announced on 10 July - twelve days out (news (Steam events) · 2026-07-10)
- **f3** [observed] Enters Popular Coming Soon at #12, identical rank in all nine countries we poll (storefront_visibility · 2026-07-19). Rank moves in lockstep across countries - this shelf is not localised
- **f4** [observed] Moves 12 → 11 → 12 → 9 → 10 within a single day, again in lockstep everywhere (storefront_visibility · 2026-07-20)
- **f5** [observed] Publisher appears on the page two days before launch: support URL switches to kwalee.com, Discord link added in the same minute (events (store-page diff) · 2026-07-20)
- **f6** [observed] Store description rewritten four hours before release (events (store-page diff) · 2026-07-22)
- **f7** [observed] Tops both tag pages it belongs to - Horror #1 and Simulation #1 - half an hour before appearing in the top sellers (storefront_visibility · 2026-07-22)
- **f8** [observed] Enters Top Sellers across 41 countries at once (storefront_visibility · 2026-07-22)
- **f9** [observed] Reaches the Steam front page: featured slot #1 in TR, ID and BR, top-2 in DE, CH and GB (storefront_visibility · 2026-07-22)
- **f10** [observed] Priced at $9.99 with a 10% launch discount ($8.99), which also puts it on the Specials shelf (price_snapshots · 2026-07-24)
- **f11** [observed] Concurrents climb rather than decay: 17.3k peak on 23 July, 23.8k on the 24th, 33.1k on the 25th, 37.6k on the 26th (metrics (ccu) · 2026-07-26)
- **f12** [observed] Hits #1 on the Chinese top sellers on day four, not day one (storefront_visibility · 2026-07-26)
- **f14** [observed] 26 games shared the Popular Coming Soon shelf in the same window; 7 of them never reached the top sellers at all (storefront_visibility (cohort) · 2026-07-26). The cohort is the counterfactual: same shelf, same days, different outcomes
- **f15** [observed] Shelf position did not predict the outcome: three games held Popular Coming Soon #1 and finished #1, #1 and #6 in the top sellers; Dinoblade, the #6, peaked at 1.8k concurrents, while Shift At Midnight sat at #3 and peaked at 37.6k (storefront_visibility + metrics (cohort) · 2026-07-26)
- **f16** [observed] 3.0k reviews in the first four days, 89.7% positive; the corpus starts on release day (reviews · 2026-07-26). Collected on demand while this case was being built - the game was not in the review set before
- **f17** [observed] Negativity is concentrated in the first hour: 61% positive under 1h, against 90-94% at every longer bucket (reviews (playtime_at_review) · 2026-07-26). 205 reviews under an hour, 2.2k in the 3-10h band where the mass sits
- **f18** [inference] The first-hour drop reads as technical rather than editorial: the negative texts name crashes in co-op lobbies and a restocking softlock, not the game itself (reviews (text) · 2026-07-26). Read from review text, not measured - the competing reading is that co-op friction filters out solo players early

## Peer set

- Kept: **Halo: Campaign Evolved**. Same shelf, same window · Popular Coming Soon #1 → Top Sellers #1 · no concurrents recorded
- Kept: **DragonSword : Awakening**. Same shelf, same window · Popular Coming Soon #1 → Top Sellers #1 · peak 23.0k concurrents
- Kept: **Rubinite**. Same shelf, same window · Popular Coming Soon #2 → Top Sellers #2 · peak 2.1k concurrents
- Dropped: **Dinoblade**. Dropped as a comparable, kept only as a contrast: it held the shelf better and finished far lower, which is the point
- Dropped: **Scarlet Deer Inn**. Never entered the top sellers, so there is no post-launch trajectory to compare against

## What the record cannot answer

- Reviews were collected on demand for this case; before that the game was outside the review set entirely
- Wishlist counts are not public anywhere

## What would weaken the conclusion

- Publisher appearance, price, tag-page placement and the front-page slot all fall within 48 hours - their individual contributions cannot be separated
- The cohort is 26 games from one window; it disqualifies the shelf hypothesis but is too small to rank the alternatives
- The review-to-units multiplier is not calibrated on our cohort, so no unit-sales figure is given

## Scout the market

- [Release calendar](https://app.playrecon.com/calendar)
- [Release density](https://app.playrecon.com/heatmap)
- [Genre seasonality](https://app.playrecon.com/seasons)
- [Niches](https://app.playrecon.com/market)
- [Release trends](https://app.playrecon.com/trends)
- [Studios](https://app.playrecon.com/studios)
- [Unbuilt](https://app.playrecon.com/unbuilt)
- [Pitch comps](https://app.playrecon.com/idea)

## Company

- [About](https://playrecon.com/about.md)
- [Pricing](https://playrecon.com/pricing.md)
- [Privacy](https://playrecon.com/privacy.md)
- [Terms](https://playrecon.com/terms.md)

## Contact

- [hello@playrecon.com](mailto:hello@playrecon.com)
- [early@playrecon.com](mailto:early@playrecon.com?subject=PlayRecon%20private%20beta)
- [LinkedIn](https://www.linkedin.com/company/playrecon)