💧 What the Score Measures
The liquidity score is a 0-100 rating of how quickly you could sell an item without cutting the price. It is not a rating of how good a skin is, how rare it is, or how much it is worth. It answers one question: if you listed this today at the going rate, would it be gone?
That question matters more than most people expect, because liquidity is a property of order flow, not of price. A knife worth $3,000 can change hands several times a day while a $4 skin sits for three weeks with no standing buy order behind it. Price tells you what an item is worth. Liquidity tells you whether you can get it.
You will see the score in three places: the ring on the liquidity card of every item page, the chip on browse and wishlist rows, and the full ranking on the CS2 liquidity board, where every rated item in the catalogue is sorted by trade flow.
Trade flow sets the tier, the tier owns a 25-point band of the 0-100 scale, and the evidence only decides where inside that band the item lands. A wide spread or a long listing queue costs a whole band. Nothing ever promotes an item above what its best measurement supports.
🏷️ The Five Tiers & Their Bands
Every rated item sits in one of five tiers, and each tier owns a fixed, non-overlapping slice of the 0-100 scale. The badge and the number can never disagree, because the number is derived from the badge rather than computed alongside it.
| Tier | Score | What it means for a seller | How to trade it |
|---|---|---|---|
| High Demand | 85 – 100 | Constant buyers at the going price | Treat it as cash. You can exit at the going rate whenever you want to. |
| Active Demand | 70 – 84 | Plenty of buyers — you can ask the going price | A normal position. List at market and let the queue carry it; undercut only if you are in a hurry. |
| Moderate Demand | 50 – 69 | Enough buyers, if you match the lowest price | Plan the exit before you buy. Fine to hold, slow to leave in a hurry. |
| Slow Mover | 20 – 49 | Buyers are not always there — expect to price below other sellers | Sizing matters. One copy is patience; five copies is a problem. |
| Rarely Traded | 0 – 19 | Sits on the market — selling means cutting the price | Assume the exit costs a discount, and price the discount in on the way in. |
The bands are deliberately uneven. Moderate Demand spans 20 points because most of the catalogue lives there, and a wide band keeps a busy item distinguishable from a barely-moving one. High Demand gets the top 16 points because past a certain volume the only useful statement left is "there is always a buyer". How long your listing waits is a separate question — that is the queue, and it has its own estimate on the time-to-sell board.
The thresholds behind each tier
Two measurements can place an item, and they are not interchangeable — one is a rate, the other is a rank. Both are shown here so you can see exactly where a boundary sits:
| Tier | Measured trade flow | BUFF liquidity percentile |
|---|---|---|
| High Demand | 380+ per day | 99.1 – 100 |
| Active Demand | 40 – 380 per day | 91.1 – 99.1 |
| Moderate Demand | 8 – 40 per day | 77.6 – 91.1 |
| Slow Mover | 1 – 8 per day | 54.7 – 77.6 |
| Rarely Traded | under 1 per day | under 54.7 |
The percentile boundaries sit at the same quantiles as the trade-rate boundaries — the top tier near the 99th percentile of the catalogue, the next near the 90th — so an item does not become easier or harder to rate depending on which source happened to see it.
💎 When "Rarely Traded" Is the Wrong Reading
The bottom of the scale measures turnover, and turnover is not demand. A Rarely Traded badge tells a seller to expect to cut their price, which is sound advice about a quiet $12 skin and flatly wrong about a $30,000 sticker. Both trade twice a year. Only one of them does so because nobody wants it — the other is thin because almost none exist, and its owner discounting is not how they exit, it is how they lose money.
So an item that is expensive and barely trades is shown as a Collector Market instead. Three conditions, all of them required:
| Condition | Why |
|---|---|
| Worth $20,000 or more | Scarcity is the only thing separating the two cases, and price is the evidence of it. An item with no reference price is rated exactly as it was before. |
| Already rated Slow Mover or Rarely Traded | Those are the two bands whose wording overreaches. A $50,000 knife clearing three trades a day is a liquid market that happens to be expensive, and its own label already says so. |
| Someone is bidding | A measured empty buy side is the market saying nobody is standing there at any price. That beats an inference drawn from the price tag, so it withholds the reading. |
It is a correction to why the number is small, not a promotion. The tier does not move, the 0-100 score does not move, and the item is no easier to sell than it was — the buyer pool is small, the wait is real, and finding the price is most of the work. What changes is only that the page stops telling the owner of the most sought-after item on it that nobody wants it. You can filter the board to these directly on the collector market view.
The $20,000 line is set high on purpose, and higher than "expensive" on its own would suggest. A $5,000 knife is a Ruby Doppler or a Pandora's Box — an expensive normal market with plenty of units and real turnover, not a scarce one. That price range is also where odd StatTrak and souvenir variants sit with optimistic asks and nobody bidding, which is exactly the population this must not fire on. Past roughly $20,000 what is left is genuinely scarce: Katowice 2014 holos, Case Hardened blue gems, souvenir grails.
Getting this wrong is not symmetric, which is why it errs toward silence. A genuine grail reading Rarely Traded is the state of things before this existed and costs its owner nothing new, while an unwanted expensive item reading Collector Market has been told not to discount — the one thing that would have sold it. A badge that fires on a few dozen real grails is worth more than one that fires on a few thousand items that are only dear and quiet. The line is still reasoned rather than measured against the catalogue.
📊 The Three Signals
Three things decide how liquid an item is, and a rating built on any one of them alone fails in a predictable direction:
| Signal | What it is | Why it fails on its own |
|---|---|---|
| Flow | Units actually traded per day | Overstates. An item can trade 20 a day and still leave your listing buried behind 600 others. |
| Queue | Listings on the sell side ÷ that venue's daily flow | Says nothing about demand. An empty queue on an item nobody buys is not liquidity. |
| Spread | Gap between the best buy order and the best listing | This is what selling now rather than eventually costs you. Wide spread means the flow is happening at prices you would not accept. |
The queue figure is measured against the trades at its own venue, not against pooled flow across every market. A queue only clears through the order book it sits in — the listings ahead of yours on one marketplace are untouched by sales happening on another. Dividing one venue's listings by everyone's trades would understate the wait, which is exactly the optimistic direction a selling signal must not fail in.
🔎 Where the Numbers Come From
Two independent measurements feed the rating, and each is blind where the other sees.
1. Measured trade flow
Units traded over a trailing 30-day window, combining Steam Market sales with the turnover we observe on the marketplaces we track. Days with no observation count as zero trades, not as missing data — thin sampling therefore drags an estimate down rather than inflating it.
2. BUFF's published liquidity percentile
BUFF163 is the largest CS2 marketplace by volume and publishes a 0-100 liquidity rank per item, built from fills it actually processed. That is strictly more than anyone can infer from listing snapshots, and it covers the part of the market Steam cannot see at all.
The Steam Community Market will not list anything above $1,800. Every knife, every pair of gloves and most high-tier covert skins therefore generate zero Steam sales data — not low data, none. Below the cap both sources usually speak. Above it, only BUFF does.
This is why the two numbers on a knife's card can look like they are arguing: a Karambit BUFF ranks in the high 99s may show a fraction of a trade per day beside it. The percentile is the measurement; the trades-per-day figure is only what we could see elsewhere, and a book that relists as fast as it sells barely moves in a snapshot.
Where both sources have an opinion, the more liquid of the two wins — within a band of what we measured. They read different order books and both are real, so an item that trades hard on Steam is liquid whatever BUFF's percentile says about BUFF. The reverse needs a limit, and here is why: a percentile is a position in BUFF's catalogue, not a rate. Ours is dominated by cases trading in the thousands, so our 99th percentile is 380 trades a day. BUFF ranks some 20,000 items, most of them barely moving, so their 99th percentile can be a skin that fills once every few days. Read one straight off the other and a skin selling eight times a month gets called "High Demand".
So below Steam's cap, where we hold confirmed sales, a high BUFF rank is treated as evidence that our own flow understates the book — not as a measurement of by how much. It lifts the rating by one band at most, and the board marks the rows where it did. Above the cap, where Steam sees nothing, there is no rate to hold it against and BUFF carries the rating outright.
🧮 How a Score Is Built
Four steps, in this order:
1. Place the item. Take the better of the flow tier and the BUFF tier — the BUFF tier held to one band above the flow tier wherever confirmed sales exist to hold it against.
2. Apply demotions. Spread and queue checks can each cost a full tier. Nothing promotes.
3. Find the position. How far into its band the winning measurement reached, 0 to 1.
4. Map to the band. Place that position inside whatever band the item ended up in.
Step 3 is log-scaled on trade flow and linear on BUFF's percentile. Flow is a rate that spans orders of magnitude — the step from 1 to 8 trades a day is the same kind of change as 40 to 380 — so a linear read would pin nearly every item to the floor of its band. A percentile is already a rank, so re-shaping it would double-count the distribution.
Step 4 is why relative position survives a demotion. An item sitting at the top of Active Demand that loses a tier to a 60% spread lands at the top of Moderate Demand — still the strongest thing in its class, just no longer in the class it measured into.
⬇️ What Pulls a Rating Down
Four rules can move an item down, and none can move one up. Every demotion that fired is listed on the item's liquidity card, so you always see why a rating is where it is.
| Rule | Trigger | Cost | Reasoning |
|---|---|---|---|
| Wide spread | Bid/ask gap of 40% or more | One full tier | The trades are real but they are clearing at a price well under the ask. Selling now costs you the gap. |
| Deep queue | 14+ days of listings ahead of you | One full tier | The flow is genuine, it is just not going to reach your listing any time soon. |
| No standing buy orders | Zero bids on the book | Capped at Moderate Demand | You can sell, but only by waiting for a buyer to appear. Nothing is standing there ready to take it. |
| Inferred flow | Trades implied by listings vanishing, with no BUFF rank | Capped at Active Demand, and below high confidence | A listing that disappears may have sold — or been delisted or repriced. That is evidence of movement, not proof of demand. |
A 40% spread threshold sounds high if you are used to equities. It is not, for this market: bids on CS2 venues sit far below asks as a rule, and the median tracked item runs around a 20% spread. A threshold that fired on the typical item would flag most of the catalogue and discriminate nothing. The raw spread is reported on the card regardless, so an item with a 25% gap still shows it — it just does not lose a tier for being ordinary.
🎯 Confidence vs Score
Alongside the score sits a confidence chip. It answers a different question — not how liquid, but how much evidence is behind that:
| Confidence | What produced it |
|---|---|
| High | Two or more independent venues, 20+ active days in the window, data no more than 2 days old — or two sources landing on the same tier. |
| Medium | At least 10 active days and data within 3 days, but short of the bar above. |
| Low | Thin or ageing coverage. The rating is the best read available, not a firm one. |
One extra check runs here: if the median day in the window is far below the mean, the series is being carried by one or two outlier days, and confidence drops. A single busy afternoon inside a quiet month is not a liquid market.
Confidence deliberately does not move the score. Averaging "how well it trades" and "how sure we are" into one number would hide both, so they stay separate signals you read together.
🚫 When We Refuse to Rate
Some items show "not enough data" rather than a low score. Three cases produce that:
- No data — nothing observed for the item at all.
- Thin coverage — fewer than 10 of the 30 days carried an observation. Below that floor, the average says more about our sampling than about the market.
- Stale — the newest observation is more than 7 days old.
An item we stopped seeing is not an item nobody wants. Rating it 0 would be a confident claim built on an absence of evidence, so no rating is shown at all. If a skin you own is missing from the board, the honest reading is "we cannot tell yet" — go and look at the live listing depth and buy orders yourself.
🛠️ How to Use It
When you are buying
- Check the score before the price, not after. A 15% discount on an item scoring in the 20s is not a discount — it is the market telling you what the exit costs.
- Match liquidity to your holding period. Flipping this week needs a 75+. Holding for a year, a 50 is fine.
- Read the demotion list. "No standing buy orders" on an item you planned to flip is a red flag the headline number alone would not have shown you.
When you are selling
- Use the queue figure to set expectations. If there are 30 days of listings ahead of you, listing at the going rate means waiting, not selling.
- Use the spread to price the hurry. A wide spread is the cost of instant sale — that is the discount an instant-sell buyout is charging you.
- Split large positions. Five copies of a Slow Mover item is a different problem from one copy; the queue you are joining includes your own other listings.
When you are investing
- Liquidity is the risk you cannot see in a price chart. A price history looks equally smooth whether the item trades 500 times a day or twice a week.
- Percentage moves on thin volume are noise. One optimistic listing can move the average on an item that barely trades. Check the score before you act on a trending gainer.
- Size positions against the exit, not the entry. The number that matters is how many copies the market absorbs per day, not how cheap they look today.
💧 See every rated item ranked
The CS2 liquidity board ranks the whole tracked catalogue by real trade flow, updated every 15 minutes — filter by knives, gloves, rifles or cases, or sort by widest spread to find the items where exiting costs the most.
🧠 Five Common Misreadings
1. "Expensive means illiquid"
Not reliably. Popular knives and gloves in standard conditions trade constantly. What actually kills liquidity at the top end is not the price but the specificity — a rare pattern, a heavy sticker craft, an unusual float. See the most liquid CS2 skins for the breakdown.
2. "High trades per day means I can sell now"
Only if the queue is short. Flow measures the market, the queue measures your place in it. The rating already combines the two, which is why an item with impressive volume can still land in a lower band.
3. "The score dropped, so demand collapsed"
Check the demotion list first. A band-sized drop is more often a spread widening or a queue building than flow disappearing — and both of those can happen while the number of trades stays flat.
4. "Low confidence means low liquidity"
It means thin evidence. A recently released item can be genuinely liquid and still carry low confidence because the window has not filled up yet.
5. "Every condition of a skin has the same liquidity"
They do not, and the gap is often large. Field-Tested is usually the most traded condition of a popular skin simply because it is the most common; the same skin in Well-Worn can score two bands lower. Ratings are per condition for exactly this reason — float and wear change what you are actually selling.
❓ FAQ
Is the liquidity score the same as the price?
No, and the two are close to independent. Liquidity is a property of order flow, not of value. A $3,000 knife that changes hands every day is more liquid than a $4 skin nobody lists a buy order for. Price tells you what an item is worth; the liquidity score tells you whether you can actually get that number.
What does a score of 100 mean?
The top band is open-ended, so 100 is reserved for genuine extremes — items trading many multiples of the 380-per-day floor of the top tier. In practice a handful of cases and capsules reach it. Anything in the 90s already means "sells within hours at market price".
Why does a heavily traded skin sometimes score lower than I expect?
Because trade count alone flatters an item. Two checks can pull a rating down a full band: a bid/ask spread of 40% or more, which means the trades are happening at a price you probably would not accept, and a listing queue of 14 days or more, which means the flow is real but it is clearing other people's listings before yours. Nothing pushes a rating up — an item is never rated more liquid than its best measurement says.
Why does a knife show a high score but a fraction of a trade per day?
Because the two numbers come from different books. The Steam Market will not list anything above $1,800, so our Steam sales feed is silent on almost every knife and glove. Above that cap the rating comes from BUFF's published liquidity percentile, which is built from fills BUFF actually processed. The trades-per-day figure beside it only ever describes what we could measure elsewhere, and on a book that relists as fast as it sells that figure reads near zero.
What does the confidence chip mean, and why does it not change the score?
Confidence describes how much evidence sits behind the rating — how many days of the 30-day window carried observations, how many independent venues contributed, how recent the newest one is. How well an item trades and how sure we are of it are two different claims, and averaging them into a single number would hide both. So the score answers "how liquid", the chip answers "how sure", and you read them together.
Why do some items say "not enough data" instead of showing a low score?
Because an item we cannot measure is not the same as an item nobody wants. If fewer than 10 days inside the 30-day window carried an observation, or the newest observation is more than 7 days old, the rating is withheld. A badge people make selling decisions on should fail toward "we do not know", never toward "yes, it will sell".
How often does the score update?
The underlying view refreshes every 15 minutes, and the ratings move with it. The window itself is trailing, so a rating reflects the last 30 days of trading rather than the last hour — a single busy afternoon will not lift an item a band, and a single quiet one will not drop it.
Does a high score guarantee a fast sale?
No. It says the market has been absorbing that volume recently at the prices shown. Demand shifts, and your own asking price still decides whether you are at the front of the queue or the back. Treat it as a read on the market, not a promise about your sale.