Valve erased six days of Dead by Daylight reviews: the math

Hero image: Behaviour Interactive, via the game's Steam store page.

On November 7 the Dead by Daylight store page showed Mostly Positive, with an asterisk under the score. The note says reviews posted between November 1 and November 6 have been excluded as off-topic activity. The game did not change that week. Valve changed which reviews the default number counts.

Most coverage stopped at the asterisk. Underneath it sits a six-day spike that can be counted, one day that carried almost all of it, and two percentages that reporters keep quoting as if they described the same thing.

Six days, four of them counted

Steam's own review history chart, read on November 7, gives the daily figure for four of the six days in the window.

Date	Negative reviews	Positive reviews	Negative share
Nov 2	262	not reported	n/a
Nov 3	448	not reported	n/a
Nov 4	1,285	147	89.7%
Nov 6	154	not reported	n/a
October, daily average	53	not reported	n/a

Only November 4 has both a negative and a positive count in the sources, so it is the only day where a share can be computed: 1,285 divided by 1,432 reviews is 89.7%. The other days give negative totals without a matching positive count, so that cell reads not reported.

The baseline needs two numbers, because the sources give two. Steam's chart shows 1,643 negative reviews for the whole of October, which is about 53 a day across 31 days. D/REZZED, reading the same chart, puts late October at roughly 70 a day. Both are correct and they answer different questions. 53 is a monthly average that includes a quiet first half; 70 is the run rate going into the spike. Against the monthly average, November 4 is 24 times normal. Against the late-October rate it is 18 times.

Add the four days we can count and you get 2,149 negative reviews. October's entire month was 1,643. Four days produced 31% more downvotes than the previous 31 days. November 1 and November 5 are missing from the chart we could read, so 2,149 is a floor, not a total.

Where the peak sits on the timeline

The dispute did not start on Steam. On October 30, a Behaviour staff member gave TheBurntPeanut the game's paid DLC on stream. The official account answered his offer to return it with "WE CAN TAKE THE HEAT! THEY ARE YOURS TO KEEP." Behaviour apologised on November 1, and said on the night of November 2 that the content had been pulled. Dexerto dates the removal announcement to November 3, when the studio published a longer statement admitting it had moved too fast. The two accounts differ by a day; the sequence does not.

The steepest day, November 4, falls after the reversal, not before it. By then two groups were posting: players angry that the DLC was handed over, and TheBurntPeanut's audience angry that it was taken back. A single count for that day mixes both, and the public chart cannot separate them. Dexerto notes the same split, and Kotaku's John Walker described the game on November 3 as "bombarded by negative reviews on Steam coming from every imaginable side."

Jake Lucky posted on November 6 that the score had been corrected. Dexerto and DualShockers published on November 7.

Two denominators, and a third

Reports on November 7 quoted 78% and 74% in the same breath. They are different measurements of different samples.

Figure	Sample	Reviews counted	As of
78% positive	all reviews, all languages	about 950,000	2026-11-07, SteamDB via Dexerto
74% positive	recent reviews, roughly the last 30 days	not stated	2026-11-07, DualShockers
79% positive	English-language reviews, all dates	269,393	2026-11-07, store page via D/REZZED

The 78% covers everything Valve counts for the game since 2016, in every language, with off-topic windows already removed. The 74% covers the recent window, the number that slid to Mixed and climbed back after the exclusion. One can move while the other stays still. The recent slice is a small part of 950,000 reviews, so a few thousand posts swing it and barely register on the lifetime figure.

The 79% adds a language filter on top. Counting only English reviews drops about 680,000 of the 950,000 that SteamDB sees. A reader who meets 79% and 78% in the same week is looking at two different queries, not two conflicting results. We are not averaging them, because the public pages do not expose enough to reconstruct either one exactly.

The score did not rise because the excluded reviews were judged wrong or fake. It rose because the default stopped counting a date range. Anyone who wants the unfiltered number can have it in two clicks, and the unfiltered recent score reads Mixed.

What the flag actually does

Valve's rule comes from a Steam blog post dated March 15, 2019. A tool watches for anomalous review activity across all games. It does not know why the activity is happening; it notifies a team at Valve, who investigate. If the team decides the activity is off-topic, Valve marks the date range and tells the developer. Off-topic means "a topic that we consider unrelated to the likelihood that future purchasers will be happy if they buy the game."

Nothing is deleted. Valve's own answer in that post is "No, we are only changing the way we calculate a game's Review Score." The November 1 to 6 reviews are still on the page. They sit outside the default calculation.

The window is also not selective. Every review posted between those dates drops out, including a sincere post about matchmaking or a killer's balance. Valve conceded the point in 2019 with the words "Unfortunately, this is correct."

A store setting puts off-topic periods back into the score, so the exclusion is a default that any reader can switch off. Turn the setting on for Dead by Daylight and the recent rating falls back to Mixed, which is what IGN and TheBadGamer reported in their write-ups of the change.

What we could not check

We wanted to line the review curve up against concurrent players and could not do it cleanly. SteamDB's chart pages returned a Cloudflare block from our network on November 9, and this desk will not quote daily peaks it has not read. SteamCharts, which we could reach, publishes monthly averages: October 2026 averaged 47,884 concurrent players, and the trailing 30 days averaged 49,032 with a peak of 79,512. Those are monthly and rolling figures. They say nothing about November 1 against November 5, so no claim about a player drop appears in this article.

Valve has not published a statement about this specific decision. D/REZZED found none as of November 7 and we found none either. The 2019 post describes the process and leaves the trigger threshold unstated, and Valve has not said how the system has changed since.

Every review figure above is stamped November 7, 2026, the date the sources read the page. Review totals move daily, and the four daily counts come from a chart that only one outlet published in full. The 89.7% figure is our division of their two numbers, not a figure either outlet printed.

For how we treat platform rules as data in general, see our refund window audit and why we stopped scoring games out of ten.