A crypto spread screener is a tool that reads, in real time, the price of one asset across many exchanges and shows where the gap is wide enough to make money on. The key phrase is "make money on": a raw price gap almost always misleads, because it ignores fees, network transfer cost, and whether deposit and withdrawal are even open. A good tool shows not a raw percent but an executable spread - net of costs, with honest statuses. But there are dozens of these on the market, and most just paint big numbers. This article is about how to choose: the concrete signals that separate an honest scanner from a pretty aggregator, plus how it's built in Finder.

One level needs pinning down first, because everything else rests on it. The raw spread is (bid on exchange B − ask on exchange A) / ask - what a primitive "price aggregator" paints. The net (executable) spread is the same gap minus the taker fees on both legs and the network cost, with a check that withdrawal and deposit are open. The difference between them is the difference between "saw a percent" and "read a tradeable route." Exactly how a route is read layer by layer (gross → fees → network → D/W → depth) is a separate topic, covered in detail in «Crypto arbitrage routes: how to read them». We won't re-teach it here - instead we look at what a tool must do for you.

How to choose a screener: 7 signals of an honest tool

Most "scanners" differ not in whether they compute a net spread, but in details you only see in practice. Here are the concrete selection criteria - they weed out 90% of the junk tools.

PURR · same coin, two venues · what a row really pays price ↑ sell · bid on B 1.018 buy · ask on A 1.000 raw gap +1.8% − taker A 0.1% − taker B 0.1% − network (SOL) 0.2% net +1.4% tradeable an aggregator paints the +1.8% · a scanner shows the +1.4% you keep
The raw gap between two venues is the ceiling. Two takers and the network fee carve it down - here +1.8% becomes +1.4% net. Every criterion below exists to make that net number honest.

1. Data freshness and latency. The key technical parameter. A spread on a liquid asset lives seconds. If the tool polls exchanges every 30–60 seconds over REST, the route is already dead by the time it's shown. The mark of a good scanner is WebSocket data (a live bid/ask stream), not periodic polling, plus a visible quote-age stamp. If it never says when the price last updated, assume you're looking at the past.

2. Exchange and network coverage. A spread only exists between a specific pair of venues. A tool on 3–4 exchanges shows a fraction of the windows of one on 20+, and it misses exactly the illiquid alts where gaps are widest. What matters isn't only the CEX count but DEX and perp support, and - separately - network mapping: one coin on two exchanges may share no common network at all, in which case the route physically doesn't exist.

3. Source of D/W statuses. The crucial and most common failure. Ask yourself: where does the tool get the withdrawal/deposit status? An honest scanner pulls it straight from each exchange's API, per coin and per network - and if there's no data, it shows ❔ rather than guessing. A tool that infers "open" from indirect signals, or shows one flag per coin with no per-network breakdown, is more dangerous than no flag at all. Why a fake 🟢 = stuck capital is covered in withdrawal windows and fees.

4. False-positive handling. A big percent is almost always non-working. A strong tool hides or flags routes with closed D/W, dead liquidity and a stale price by default. A weak one proudly shows "+47%" at the top of the list, because that looks "more profitable." Red flag: a long list of huge percentages with not a single flag - that's not a scanner, it's a bait generator.

5. Depth-aware spread, not top price. The top-of-book bid/ask may hold $200, but at $2000 the price is already worse. Ask whether the tool computes executable size (summing order-book levels) or shows a spread "on one coin at the top." Without depth, the percent is honest only up to the first real order.

6. Network fee in the math. The fixed withdrawal fee ($1 on Tron, $5–30 on Ethereum) eats the whole spread on small sizes: $5 on a $200 transfer is −2.5%. A good scanner subtracts it from the net spread and suggests a cheap network. A bad one computes "as if the transfer were free."

7. Alerts, not just a table. Windows appear and vanish continuously - staring at a screen 24/7 is impossible. A useful tool can push a notification (push, Telegram, a sound) on your filter: "net > 1%, size > $1k, both legs 🟢." Without alerts, a scanner makes you do the polling instead of it.

A 30-second honesty test: find the route with the biggest percent in the tool and look at its D/W, depth and quote age. If +20% is shown with no closed-withdrawal flag, no size and no update time - you're looking at a bait aggregator, not a scanner. A big percent with no context isn't a find, it's a red flag.

Why you need a tool at all, not just hands

A fair question: why choose a scanner if you can compute the spread yourself? The answer is scale. Dissecting one route by hand is doable in a couple of minutes (exactly how - in how to read a route). But there are thousands of windows on the market, across different assets and exchanges at once, and each lives from seconds to minutes. Monitoring 20+ venues by hand is physically impossible - while you check one pair, a dozen others open and vanish. Manual work is great for learning to read the layers and verifying a specific route before entry. Keeping the whole market in view without a tool simply can't be done. So the question isn't "do I need a scanner" but "by which signals do I pick an honest one" - the seven above.

scan 20+ venues × thousands of pairs → rank by NET spread filter: net > 1% PURR Bybit→OKX +1.4% ~$6k · tradeable WIF MEXC→Gate +1.1% ~$4k · tradeable ARB Bitget→Bybit +0.7% below filter PURR Bitget→Bybit +0.1% ~$900 · ~zero net PURR HTX→OKX raw +9.2% — net not computed withdraw closed · HIDDEN MEW Gate→MEXC raw +6.5% — dead book no depth · HIDDEN TIA OKX→Bybit raw +4.8% — quote 51s old stale · HIDDEN 2 tradeable surface · the loud big-% rows sink net % →
What no pair of hands can do: rank thousands of windows by net spread every second. The loud "+9.2%" rows sink on a closed withdrawal, dead book or stale quote - the two honest ones float to the top.

How Finder answers these criteria

Finder is built around exactly those seven signals - not as a universal "aggregator of everything" but as a scanner that doesn't lie about tradeability. Briefly, point by point:

  • Freshness. A live bid/ask stream over WebSocket, not periodic REST polling - quotes update in real time.
  • Coverage. 20+ venues: CEXs, DEXs and perp exchanges, with network mapping so routes are built only on networks that actually overlap.
  • D/W source. Statuses pulled straight from exchange APIs, per coin and network. No data → ❔, not a made-up 🟢. The principle is strict: an honest "don't know" beats a fake "open."
  • False positives. Routes with closed D/W, dead liquidity and stale prices sink/hide rather than headline the list.
  • Depth. The spread is computed against book size, not a single top price.
  • Network in the math. Net spread = gross minus two takers minus the fixed withdrawal fee of the chosen network.
  • Alerts. A notification on your filter, not "sit and refresh the table."

Alongside spot spreads, the dashboard has a funding table (perp funding-rate divergence), liquidity walls and DEX dumps - for the arbitrage types where the gap lives outside the spot book.

Example: how to read a row in the table

The criteria show up best on the list row itself. Here's how an honest scanner displays three different routes for token PURR - and what the operator's eye checks in each row. Columns: gross / net / network / quote age / depth / withdraw·deposit.

ASSET  ROUTE            GROSS   NET     NET'K  AGE    DEPTH     W · D      VERDICT
─────────────────────────────────────────────────────────────────────────────────
PURR   Bybit→OKX        +1.8%   +1.4%   SOL    2s     ~$6k      🟢 · 🟢    tradeable
PURR   HTX→OKX          +9.2%   —       —      48s    🔴dead    🔴 · 🟢    hidden
PURR   Bitget→Bybit     +0.7%   +0.1%   BSC    1s     ~$900     🟢 · 🟢    ~zero

Read against the criteria above:

  • Row 1 passes all seven: net survives fees and network, the quote is fresh (2s), the book holds ~$6k, both legs 🟢. This is the working window.
  • Row 2 is the bait a weak tool would put first for "+9.2%." An honest scanner doesn't even compute net: PURR withdrawal on HTX is 🔴 closed, liquidity is dead, and the quote is stale (48s). The row is hidden or sunk - not worth a second of attention.
  • Row 3 is nominally 🟢/🟢 and fresh, but after two takers +0.7% leaves +0.1%, and depth is only ~$900. Technically not a phantom, but not a trade either.

The tool's whole value is in the columns after GROSS: the percent alone says nothing - net, age, depth and D/W together decide. The same layer-by-layer read on one route with full arithmetic is in how to read a route.

FAQ - how to choose a spread screener

What should a good spread screener do?

At a minimum: compute the net spread (after two takers and the network fee), pull honest D/W statuses per coin and network, estimate book depth, filter out non-working gaps, and update data in real time. If a tool shows only a raw percent, it's an aggregator, not a scanner.

How is a scanner different from a price aggregator?

An aggregator shows the raw price gap - a pretty but nearly useless number. An arbitrage scanner subtracts fees and network, checks that withdrawal/deposit are open and the book has depth, and hides bait, i.e., it answers "is this tradeable right now," not "where's the bigger number."

What should I look at first when choosing?

The three things that get faked most: the source of D/W statuses (pulled from exchange APIs per network, or guessed), quote freshness (WebSocket vs once-a-minute polling), and big-percent handling (does it hide closed D/W, or proudly put "+40%" at the top). Those three signals weed out most empties.

What net spread counts as workable?

After two takers (~0.2%) and transfer, ~0.5–1% net is workable on liquid assets. On illiquid ones windows are wider (2–10%+), but convergence/slippage risk is higher too. A big percent on its own, with no open D/W and no depth, is a red flag, not a find.

How many exchanges should it cover?

The more, the more windows and the fewer missed illiquid routes: 3–4 exchanges give an order of magnitude less than 20+. Finder covers 20+ CEX/DEX/perp venues: spot routes in the spreads view, rate divergence in the funding table.

Not financial advice. Even an honest scanner shows a point-in-time estimate, not a promise: while you read the row, the route can converge, and statuses and depth change within seconds. The tool narrows the search and removes the bait - but the decision and the risk stay yours.

Where to next: the broad context - the pillar Crypto arbitrage guide. What a scanner computes for you - how to read a route. Where honest D/W comes from - withdrawal windows and fees. When auto-execution is bolted onto a scanner - arbitrage bots. Check the seven signals on live data - in the web dashboard.