The State of MM2 Trading — September 2026

A dated Murder Mystery 2 market report off 2,505,000 collected trade messages: most-traded items, what the community quotes versus what it trades at, and the open CSV behind every figure.

Updated 8 September 2026.

Published 8 September 2026. Every number on this page comes from our own trade corpus and can be recomputed from the CSV at the bottom.

What this report is built on

Two things about that table are the point of it. Coverage is asserted, not assumed — a hole in collection reads exactly like a quiet market, so the corpus manifest records every gap over 90 minutes and there were none. And most items do not get a published value: 453 are held because the evidence is too thin or the signals disagree. A value list that publishes a number for everything is telling you something about its editorial policy, not about the market.

The most-traded items

Counted as the number of parsed trade posts an item appears in, on either side.

The busiest items in MM2 are the cheap ones. Harvester, Icecream, Icepiercer, Bat and Candy are all under 300 in value and all in the top ten by volume, because they are the game's small change — the units bundles get padded with. Across the 161 published items, the rank correlation between how often an item appears and what it is worth is ρ = 0.24: essentially independent.

What traders want versus what they hold

Both tables use one gate — items appearing in at least 400 parsed trade posts, 69 of the 470 items seen — and both are emitted by the generator (demand in the stats file) rather than cut from the CSV by hand. Below that floor an eight-to-one ratio is a handful of posts.

Watergun is the single most one-sided item in the game right now: eight and a half offers for every request. Icepiercer is the mirror image at three requests per offer, and it is the third busiest item overall — a genuinely scarce cheap item, which is an unusual combination.

Trade volume by week

The MM2 corpus is a continuous collection with no gaps, so week-on-week here is like-for-like.

Four complete weeks, a range of 6,609 to 8,049, no trend. MM2's weekend effect is also mild: 1,134 parsed trades on an average weekend day against 1,046 on a weekday, an 8.4% uplift. For comparison, the same measurement on our Grand Piece Online panel gives 40%. MM2's trade channels run at a steady rate all week.

What the community quotes, and whether it moves

MM2 traders state prices outright — "HEARTWAND (340)", "snow dagger (250 value)" — which makes the community's consensus number directly measurable. We extracted 11,169 such quotes across 231 items. To compare across time an item needs at least 12 quotes from at least 5 distinct accounts in both the first 10 days and the last 10 days of the window, and each side is the median across authors of each author's own median, so one trader re-posting a shop list 40 times cannot outvote everyone else. 24 items qualified.

Fourteen of the 24 did not move at all — the same median to the decimal, five weeks apart. That is the real character of MM2 pricing: quoted values are sticky because traders copy them from each other and from lists, and they update in steps, not continuously. The two Beachy entries falling together is the clearest seasonal signal in the data, a summer set drifting after summer.

Why two MM2 value lists never agree

We can measure this without looking at anyone else's list, because our engine computes two independent estimates for the same item from the same corpus: what its barter chains imply, and what traders quote outright.

On the 90 published items that have both:

• median disagreement 20.9%
• 48 items disagree by more than 20%, one by more than 40%
• for comparison, the two barter-derived estimates (chain solve and simultaneous bundle fit) differ by a median of only 10.7% — agreement between two views of the same behaviour, which is why they are counted as one signal and not two

The widest disagreements:

This is the whole story of value-list disagreement, and it is not about who has better sources. Any list that leans on what traders say will land 20–40% away from a list that leans on what traders do, on the same day, with the same evidence. Both are defensible; they are answering different questions.

Where our two signals disagree beyond tolerance, we publish nothing: seven items are withheld on exactly that rule, including Sparkle Set, where the quoted "value" harvested from posts was 145,678,910 — a trader typing a sequence, not a price — and the barter evidence said 174.5. The gate caught it. That is what a gate is for, and it is why 453 items sit in "held" rather than being given a number to fill a table.

Method, and what this report cannot tell you

• Source. Public MM2 trading channels. A record is a trade post — what a player offers and what they want — resolved to canonical item names by the same resolver the value engine uses. An unresolved name becomes an explicit unknown and the trade is dropped rather than silently under-counted.
• Counting. "Appearances" counts parsed trade posts; "offered" and "wanted" count sides, so their sum can exceed appearances. Only 36,670 of 227,267 deduped messages state a parseable ratio; the rest are shop lists, chat and negotiation.
• Values. The published value set used here was solved on 24 August 2026 while the corpus runs to 7 September, because our daily solve has published nothing since then: a single malformed post in the corpus crashes the quantity parser, and it reproduces on demand. Every corpus figure on this page is current to 7 September; every figure taken from the value set carries its 24 August date. The fix is in hand and next month's report will not need this paragraph.
• Not a transaction log. MM2 has no public record of completed trades. Everything here is stated intent, at scale.
• Bots and duplicates. Messages are deduped by id and near-duplicate shop posts collapse, but repeated posting still weights appearance counts toward heavy posters. The quote comparison corrects for this explicitly (one vote per author); the appearance counts do not.
• Not in here. Real-money prices, competitor values, and anything about individual traders. Our value engine does hold a real-money reference signal internally, harvested elsewhere; no column in this CSV and no figure on this page is derived from it, and the generator strips those fields before it writes anything.

The data

• CSV, 470 items: mm2-trading-report-2026-09.csv — appearances, offered, wanted, quote count, quote medians at both ends of the window, published value, chain value, bundle value, quote value, observations, confidence, and the chain-versus-quote gap. The item column is the name used on this page and catalog_key is the join key, so every row here is findable in the file.
• Charts: the three SVGs on this page are free to embed with attribution.
• Every figure is regenerated by one script, seo/scripts/report_mm2.py, which re-reads the corpus with the value engine's own extractors and writes this page's CSV, charts and stats file. It runs monthly.

Source: PetMart / mm2index.com trade corpus, 2,505,000 messages, 2026-08-04 to 2026-09-07

Free to cite, quote and republish with a link to this page. If you want a custom cut — one item's full quote history, chromas only, a different window — ask and we will run it.

Where to buy the items in this report

We sell some of what is above. Availability changes daily, so check the page rather than this sentence:

• Harvester and Raygun are two of the five busiest items in the game.
• Celestial, Chroma Raygun, Gingerscope and Evergreen sit at the top end.
• The knife shelf is at MM2 knives; everything we stock for the game is on the Murder Mystery 2 hub.

MM2 items are handed over by a manual in-game trade — you meet our trader in a private server, the item moves through the trade window, and the completed order is photographed. Our delivery statistics are published live on /about rather than printed here where they would go stale.

If you only want values and not a purchase, mm2index.com publishes the same engine's output for free, with the observation count on every item.

PetMart is not affiliated with Roblox Corporation or with Nikilis.

Sources

• mm2-values/data/corpus.jsonl + corpus_manifest.json — 2,505,000 collected #trading messages, 2026-08-04T18:14Z to 2026-09-07T23:57Z, zero coverage gaps over 90 minutes, manifest generated 2026-09-08T00:13Z
• mm2-values/data/clean.jsonl — 227,267 deduped messages from 27,857 distinct accounts, parsed here with mm2-values' own extract/edges.py, extract/quotes.py and extract/resolve.py
• mm2-values/data/values.json — published values solved 2026-08-24T00:20Z: 161 published, 453 held, 7 flagged, minimum 20 observations
• seo/content/data/mm2-trading-report-2026-09.csv + .stats.json — this report's own output, regenerated by seo/scripts/report_mm2.py

Live prices on PetMart

• Knives — Murder Mystery 2
• Harvester — Murder Mystery 2 — $8.99, in stock
• Raygun — Murder Mystery 2 — $63.99, in stock
• Celestial — Murder Mystery 2 — $75.99, in stock
• Chroma Raygun — Murder Mystery 2 — $387.00, in stock
• Gingerscope — Murder Mystery 2 — $449.00, in stock
• Evergreen — Murder Mystery 2 — $69.99, in stock

More on PetMart

All Murder Mystery 2 items · Murder Mystery 2 guides · About PetMart · Proof of delivery

View on PetMart