How CSPump tells a pump from noise.

Every CS2 skin's price moves a little, all the time. Most of that is nothing: a thin market, a single trade, someone relisting at a round number. CSPump's job is to tell the difference between that background noise and a real pump, dump, or accumulation phase, and to do it while the move is still happening. Here's exactly what the algorithm looks at and how it decides.

The three signals

Everything CSPump detects falls into one of these

Pump

A sharp, often coordinated price rise, with volume and price moving up together, faster than normal drift.

Dump

A sharp sell-off, usually right after a pump, as the people who bought early take profit and price gives it back.

Accumulation

Quiet, sustained buying below the radar: the phase that often precedes a pump, before price has moved enough for most people to notice.

What the algorithm looks at

Three inputs, no more

Always running

The algorithm scans the market continuously rather than on a fixed schedule, so a move gets caught while it's happening, not after.

Western and Chinese markets

Price and listing data is pulled from both Western and Chinese CS2 marketplaces, not a single feed, which is useful because a move sometimes shows up in one region before the other. That is the entire input; no off-market or unverified sources.

Shape, not just price

A single price change means little on its own. The algorithm weighs price movement together with volume and how long the move sustains before calling it a signal.

Confidence scoring

Not every signal is equally reliable, so here's how CSPump grades its own alerts

Low confidence

Thin market, low volume, or a short-lived spike. A handful of trades can move the displayed price without real demand behind it: treat these carefully.

Medium confidence

Meaningful volume with a price move that's held for more than a single trade, but without enough history yet to be certain it's not a fluke.

High confidence

Sustained volume and price movement together over time. This is the combination most likely to reflect a real pump, dump, or accumulation phase rather than noise.

Confidence scoring exists because no algorithm working from public market data is perfect. It's a way of telling you how much to trust a given alert, not a guarantee of what happens next.

What this is, and isn't

Keeping the claims honest

CSPump surfaces signals based on public data from the Western and Chinese CS2 marketplaces it tracks. It does not have access to private trade intent, off-market deals, or anything happening outside those markets, and it does not predict the future or guarantee an outcome. A high-confidence alert means the pattern strongly resembles past pumps, dumps, or accumulation phases, not that a specific price target will be hit. CSPump is not financial advice.

FAQ

The methodology questions people actually ask

Which markets does CSPump track?

Both Western and Chinese CS2 marketplaces. Tracking both matters because the two markets don't always move in lockstep: a pump or an accumulation phase can start showing up in one region before it's visible in the other, and single-market tools miss that entirely.

How is a confidence score calculated?

It's driven by trade volume, how long a price move sustains, and how unusual the activity is relative to that item's normal trading pattern. Thin, single-trade moves score low; sustained volume across multiple trades scores high.

How often is data checked?

The market scan runs continuously. Your own inventory prices are synced multiple times a day, and every item shows the exact timestamp of its last sync so you know how fresh the number is.

How accurate is the algorithm? What about false positives?

No algorithm is perfect. On thin markets, a small number of trades can move the displayed price without real demand behind it, which is exactly what the confidence score is for. Lower confidence means treat it carefully. Higher volume plus sustained movement is the more reliable signal.

See what's pumping right now.

Go to CSPump

How CSPump tells a pump from noise.