How ETH Bubbles Scores Momentum

ETH Bubbles turns the Ethereum ecosystem into a single animated picture: hundreds of tokens, protocols, and networks drawn as bubbles whose size and color update as the market moves. This page explains exactly how that picture is computed — which data feeds it, what size and color literally encode, and where the visualization can mislead you if you read it as more than it is.

The Short Answer: There Is No Secret Score

Some dashboards blend indicators into a proprietary "momentum score." ETH Bubbles deliberately does not. The number a bubble displays — and the color and size derived from it — is the plain percentage change of a public metric over the window you selected: price change for tokens, total value locked (TVL) change for protocols and chains. We think a transparent map of real data is more useful than an opaque score, because you always know exactly what you are looking at and can verify every figure against the original source.

Why a Bubble Map? A Short Tour of Market Heatmaps

Visualizing a whole market as colored shapes is an old idea with two main families. Treemaps tile a rectangle with boxes whose area encodes market cap — the format used by CoinMarketCap's crypto heatmap and TradingView's crypto coins heatmap, and long familiar from stock-market tools. Bubble charts, popularized in crypto by Crypto Bubbles, trade the space-efficiency of tiles for free-floating circles that can animate, collide, and reorder — which makes changes over time feel physical rather than tabular. Both families speak the same visual language: color for direction, color intensity for magnitude, size for weight.

ETH Bubbles is a member of the bubble family with a few deliberate differences: it is scoped to the Ethereum ecosystem rather than the whole crypto market; it adds TVL-based views (DeFi, Restaking, L2s) next to price-based token views, which generic tools do not attempt; it defaults to sizing by the move (Performance mode) rather than by market cap, because "where is the action" is the question a momentum map should answer first; and every view has a crawlable, sortable table twin for exact figures. The comparison in brief:

None of these tools — ours included — adds information beyond the underlying data. What a good heatmap adds is bandwidth: a few hundred assets' worth of change data absorbed in a glance instead of a scroll.

Where the Data Comes From

CoinGecko: token prices and market data

Every token tab — Tokens, Meme, Stablecoins, RWAs, and AI — is powered by CoinGecko's markets API. For each asset we ingest the current price, market capitalization, 24-hour trading volume, a 7-day sparkline, and the percentage price change over five windows: 1 hour, 24 hours, 7 days, 30 days, and 1 year. Category membership also comes from CoinGecko: the main tokens dashboard uses its "Ethereum ecosystem" category, and the meme, stablecoin, RWA, and AI tabs use the corresponding CoinGecko categories, ranked by market cap. The home page lets you page through ranks 1–100 up to 401–500.

DefiLlama: protocol and chain TVL

The DeFi, Restaking, and L2s tabs are powered by DefiLlama, the open TVL aggregator. The DeFi view lists Ethereum protocols with more than $1 million TVL (excluding exchanges, chains, and bridges, which are not DeFi protocols in the usual sense), with TVL change over 1 hour, 24 hours, and 7 days. The Restaking view filters DefiLlama's Restaking, Liquid Restaking, and Liquid Staking categories. The L2 view tracks a curated set of major Ethereum Layer 2 networks — Arbitrum, Base, OP Mainnet, ZKsync Era, Linea, Scroll, and others — where the displayed change is the day-over-day movement of each network's TVL from DefiLlama's historical series.

Freshness and caching

Our server caches CoinGecko market responses for roughly 90 seconds, the DefiLlama protocol list for about 5 minutes, and L2 chain history for about 10 minutes; the global stats strip (total market cap, volume, ETH dominance from CoinGecko's global endpoint) refreshes about every 2 minutes. The gas figures on our gas page come from Etherscan's gas oracle with public RPC endpoints as fallback, cached for about 15 seconds. Add the upstream sources' own update cadences, and the honest description is: the map trails the market by a few minutes. That is fine for reading momentum; it is not an execution feed.

From feed to pixels: the full pipeline

For the curious, here is the complete path a number travels before it becomes a bubble. Our API server fetches from CoinGecko and DefiLlama and stores every response in a bounded in-memory cache with a stale-on-error policy: if an upstream source goes down or rate-limits us, we keep serving the last good data rather than a blank map. This is deliberately honest engineering with an honest side effect — during an upstream outage, figures can be older than the nominal cache windows above, with no visual indication beyond their staleness.

On a page request, our web server renders the entire page — including a plain, crawlable table of every bubble's name, value, and change — on the server, with its own short-lived HTML cache. Your browser then takes over, re-fetches fresh data on the same cadence, and swaps the static table for the animated canvas: each bubble is seeded at a random position and a physics loop applies gentle drift and collision resolution so bubbles spread out without overlapping. Two consequences are worth knowing. First, the animation is presentation, not information — a bubble's position and motion mean nothing; only size, color, and the printed number carry data. Second, because positions are random each load, the same market data produces a differently arranged map every time you visit. If you remember "the big green one was top-left," you remembered the physics, not the market.

What Bubble Size Means

Size is controlled by the metric selector, and the math differs by mode because the underlying data has very different shapes.

Performance mode (the default)

In Performance mode, each bubble's radius is proportional to the square root of its absolute percentage change relative to the largest mover currently on screen — which means a bubble's area scales with how big its move is compared to the biggest move in view. Direction is ignored for sizing: a −12% crash and a +12% pump produce equally large bubbles, and color carries the sign. This mode answers the question "where is the action?" — the eye lands on the biggest movers first, in either direction.

Market Cap, Volume, and TVL modes

Dollar metrics span many orders of magnitude — ETH's market cap is thousands of times larger than a rank-400 token's. Sizing bubbles linearly would collapse everything except ETH into invisible dots, so these modes use a logarithmic scale between the smallest and largest values in view. The consequence is important: size communicates order-of-magnitude differences, not proportions. A bubble twice the diameter of another is vastly more than twice its market cap. Token tabs offer Market Cap and Volume; the TVL-based tabs (DeFi, L2s, Restaking) offer TVL — mirroring what each source actually measures.

The packing budget

Two more rules shape every view. First, a floor: no bubble may shrink below about a third of the maximum relative size (with a hard minimum of a few pixels), so small assets stay visible and clickable. Second, a budget: all radii are rescaled together so the combined bubble area fills about half the canvas, whatever the timeframe or token count. Without this, a high-volatility window like 1Y — where nearly everything is a huge mover — would overflow the screen with overlapping giants. The trade-off is the single most important caveat about size: bubbles are sized relative to the current view only. Switching timeframe, tab, or even resizing your window rescales everything, so never compare a bubble's size against your memory of a different screen.

Why square roots and logarithms? The perception math

Both scaling choices exist because of how eyes read circles: you perceive a bubble's area, not its radius. If radius were made proportional to the value directly, area — which grows with the square of radius — would exaggerate every difference: a mover twice as big would look four times as big. Setting radius proportional to the square root of the value makes area proportional to the value itself, so the visual impression matches the data. Concretely: if the biggest mover on screen is +27% and another token moved 12%, that bubble's radius is √(12/27) ≈ 67% of the leader's — and its area is 12/27 ≈ 44%, exactly its share of the move.

The logarithm in the dollar modes solves a different problem: range. Consider three assets worth $100 billion, $1 billion, and $10 million. On a linear scale the second is 1% of the first and the third is 0.01% — both would vanish. On a log scale they sit at evenly spaced orders of magnitude (10¹¹, 10⁹, 10⁷), so the middle asset lands visually halfway between the extremes. That preserves ranking and rough magnitude at the cost of proportionality — the price of making five orders of magnitude legible on one screen.

What the Colors Encode

Color is computed from the percentage change over the selected timeframe:

• Direction: green for zero or positive change, red for negative.
• Intensity: the bubble's fill opacity, border strength, and glow all scale with the magnitude of the change, from barely tinted near 0% to fully saturated at ±18%, where the scale caps. Beyond that, a +25% and a +80% mover look identically bright — the percentage printed on the bubble (shown whenever the bubble is large enough to fit text) is what distinguishes them.
• Missing data: when a source provides no change figure for a window, the bubble renders as neutral (the dimmest state) rather than being hidden; the table view shows a dash for the same value.

The result is a heat map you can read at a glance: a sea of deep red on the 24H view is a broad drawdown; scattered bright green bubbles inside a grey field mean a few isolated movers in a flat market.

Timeframes and Category Tabs

The timeframe selector (1H / 24H / 7D / 30D / 1Y, defaulting to 24H) re-reads every bubble's change value for that window — size in Performance mode and all colors update instantly. The category tabs on the live bubble map switch datasets entirely. Because the two data sources cover different windows, not every tab supports every timeframe:

Selecting 30D or 1Y on a TVL tab therefore shows neutral bubbles — the data simply is not provided at those windows, and we would rather show nothing than fabricate it. Note also what "change" means per tab: on token tabs it is price change; on DeFi, Restaking, and L2 tabs it is TVL change, which moves both when assets are deposited or withdrawn and when the prices of already-deposited assets move — our guide to reading DeFi TVL unpacks that distinction, and ETH vs L2 tokens explains why an L2's TVL and its token's price are different things entirely.

Reading TVL Momentum vs Price Momentum

Because the two halves of the map measure different quantities, they reward different reading habits. A price change is one number with one meaning. A TVL change is a composite: it moves when users deposit or withdraw assets (real capital flow) and when the market price of already-deposited assets moves (pure revaluation). A lending protocol full of ETH collateral will show red TVL bubbles on a day ETH drops 8% even if not a single user withdrew anything — the same market event, refracted through a different metric.

Three heuristics help separate the components:

• Compare the TVL bubble to the token bubbles. If a protocol's TVL fell about as much as ETH's price did over the same window, the likeliest story is revaluation, not flight. A protocol whose TVL is flat while prices fall is quietly attracting deposits; one whose TVL falls far more than prices is seeing genuine withdrawals.
• Weight stablecoin-heavy protocols differently. Where the locked assets are mostly stablecoins, price revaluation is near zero and TVL change approximates true capital flow — the cleanest flow signal on the TVL tabs.
• Expect the 1H TVL window to be twitchy. Short-window TVL swings are frequently one large position entering or exiting, or an adapter update, rather than a trend. The 7D window is where TVL momentum starts meaning something.

This is also why the map deliberately keeps the two universes in separate tabs rather than blending price and TVL into one score: a composite would launder the very distinction you need in order to interpret either number correctly.

Beyond the Map: Table, Movers, Watchlist, Detail Pages

• Table view. Every dashboard has a toggle from bubbles to a sortable table of the same assets — the better tool when you want exact figures instead of relative impressions. A search box filters both views by name or symbol.
• Detail pages. Clicking a bubble opens a summary drawer, and each asset links to a full page — for example Ethereum for tokens or Uniswap for protocols — with price or TVL detail, supply data, chain breakdowns, and related assets.
• Market movers. The home page lists top gainers, top losers, and tokens trending by volume. One honest detail: this list is always computed over the last 24 hours, regardless of which timeframe the map above it is showing.
• Watchlist. Starring an asset pins it to a strip on the home page. The list lives entirely in your browser's local storage — no account, no server copy, no cross-device sync. It is a bookmark shelf, not a portfolio tracker.

How to Actually Read the Map

A few patterns that make the visualization genuinely useful:

1. Compare timeframes to find fresh moves. A token bright green on 1H but neutral on 7D is moving now; one bright on 7D but grey on 1H already made its move. Divergence between windows is often more informative than any single one.
2. Use Performance size to cut through noise. In Performance mode the biggest bubbles are the biggest movers — a faster scan than reading colors across a hundred similar-sized circles.
3. Check breadth, not just leaders. If the whole Tokens tab is faint green, that is a market-wide drift; if three bubbles are bright while the rest are red, those moves are idiosyncratic and worth investigating individually — starting with who is buying, as covered in our whale tracking guide.
4. Expect the Stablecoins tab to be boring. A functioning stablecoin pins near 0% on every window; visible color there is itself the signal — see how stablecoins work for why a colored stablecoin bubble deserves attention.
5. Treat meme-coin green with suspicion. Extreme short-window moves in the meme sector say nothing about legitimacy — run our meme coin risk checklist before touching anything the map surfaces.

A Worked Example: One Screen, Four Bubbles

To make the whole pipeline concrete, take a hypothetical 24H Performance view containing four tokens: A +27%, B +12%, C −12%, and D +2%. Here is everything the renderer does with them:

1. Sizing. A is the biggest mover, so it anchors the scale. B's relative radius is √(12/27) ≈ 0.67 of A's — and so is C's, because Performance sizing ignores direction; a −12% crash and a +12% pump are equally "big news." D computes to √(2/27) ≈ 0.27, which falls below the 0.34 floor, so D is clamped to the minimum — small assets never vanish entirely.
2. Coloring. A's +27% exceeds the ±18% saturation cap, so it renders at full green intensity — the map can no longer distinguish it from a hypothetical +80% mover; only the printed percentage can. B renders green at 12/18 = two-thirds intensity; C renders red at exactly the same intensity as B; D is a barely tinted, near-neutral green.
3. Packing. All four radii are then rescaled together so their combined area fills about half your canvas, and the physics simulation spreads them out without overlaps. On a phone the same four tokens get smaller absolute radii than on a desktop — the ratios between them are what survives.

And here is what you should conclude from that screen: the action is in A (but its move is already made — check whether 1H is still bright or has gone grey); B and C are a symmetric pair worth investigating for a common cause (sector rotation? one protocol's incident?); D is noise. For exact figures, entries, or anything decision-grade, flip to the table view — the bubbles are for finding the question, not answering it.

What Momentum Research Actually Says

"Momentum" is not just chart-watcher folklore; it is one of the most studied effects in finance. The modern literature starts with Jegadeesh and Titman's 1993 finding that US stocks which outperformed over the prior 3–12 months tended to keep outperforming over similar horizons — an anomaly that, as a 30-year retrospective in Financial Markets and Portfolio Management reviews, has remained among the most persistent in asset pricing ever since. A related strand, time-series momentum, studies whether an asset's own past return predicts its next one — precisely the quantity a bubble's color displays.

Crypto researchers imported both questions quickly. Studies of time-series momentum in cryptocurrencies have documented exploitable trend effects in digital-asset returns, with the earliest work dating to 2017. But the caveats are just as well documented: momentum strategies suffer rare, violent reversals ("momentum crashes"), and recent crypto-specific research finds cryptocurrency momentum portfolios are subject to exactly such severe crashes, with results that are unstable across sample periods.

Where does ETH Bubbles stand in this? Deliberately upstream of all of it. The map displays trailing percentage changes — the raw ingredient every momentum study begins from — and stops there. We compute no signal, backtest no strategy, and make no claim that yesterday's winners will be tomorrow's. The research literature is the honest frame for what a bright green bubble means: a statistically interesting fact about the recent past, attached to a well-documented risk of sharp reversal.

Frequently Misread Patterns

Watching how people use the map has taught us where intuition goes wrong. Four patterns worth calibrating for:

• The 1Y base-effect trap. On the 1Y window, a token that collapsed 95% and then tripled from its low still shows deep red, while a steady grinder shows modest green — and a token that barely survived a near-death drawdown can print the brightest gain on screen purely because its starting point was rubble. Long windows reward checking the actual price chart on the asset's detail page before concluding anything.
• Volume-mode spikes are direction-blind. A bubble that balloons in Volume mode is being traded furiously — which happens in crashes at least as often as in rallies, and in the meme sector often accompanies wash-traded launches. Volume says "attention," never "demand."
• Sudden appearances are not launches. A bubble showing up for the first time usually means the token newly entered the source category or climbed into the visible rank range — the asset itself may be years old. Check its age before treating appearance as novelty.
• A colored stablecoin is a signal, not noise. The inverse of the meme tabs: stablecoins are designed to be grey, so even a faint sustained tint on a major stablecoin is more informative than a bright bubble anywhere else on the map. It is the one place where small percentages deserve big attention.

Honest Limitations

• Momentum is not prediction. Every number on the site describes the past. Bright green means a move already happened; historically, extreme short-term run-ups revert as often as they continue. Nothing here forecasts anything.
• Data is delayed and third-party. Between upstream update cadences and our caching (90 seconds to 10 minutes depending on the dataset), figures can lag the market by minutes — more during volatility spikes. We display CoinGecko's and DefiLlama's numbers faithfully, but we inherit any error they contain, and DeFi TVL itself partly relies on how protocols report their adapters.
• Market cap and TVL are not comparable. A token's market cap (price × circulating supply) and a protocol's TVL (assets deposited in its contracts) measure different things; the same protocol can have a small token and a huge TVL or vice versa. Cross-tab comparisons of bubble size are meaningless by construction.
• Sizes are relative and log-scaled. The packing budget rescales every view to fit the screen, and dollar modes are logarithmic — so size conveys ranking and rough magnitude, never exact ratios.
• Color saturates at ±18%. Beyond that, very different moves look identical; read the printed percentage or the table for exact values.
• Everything is denominated in US dollars. Percentage changes are USD-based, so in a market-wide move nearly everything colors together, and a token that stayed flat in dollars while ETH fell 10% — genuine relative strength — shows as neutral, not green. The map does not display ETH-relative or BTC-relative performance; breadth reading (how much of the screen shares one color) is the built-in workaround.
• Coverage has edges. Token tabs extend to CoinGecko's top 500 per category; the DeFi view cuts off below $1M TVL; the L2 list is curated rather than exhaustive. Absence from the map is a coverage decision, not a judgment.

Related Guides

• How to Read DeFi TVL — what the TVL metric behind our DeFi bubbles actually measures, and its failure modes.
• What Is the Ethereum Ecosystem? — the categories the map is organized around, explained from scratch.
• ETH vs L2 Tokens — why L2 network TVL and L2 token prices tell different stories.
• How to Track Ethereum Whales — the wallet-level view to pair with this map's market-level view.

Frequently Asked Questions

Sources & Data Attribution

• CoinGecko — token prices, market caps, volumes, category lists, and percentage changes (see the CoinGecko API documentation).
• DefiLlama — protocol TVL, restaking categories, and L2 chain TVL history (see the DefiLlama API documentation).
• Etherscan API — gas oracle data used on our gas tracker, with public Ethereum RPC endpoints as fallback.
• Crypto Bubbles and the CoinMarketCap crypto heatmap — the wider bubble-chart and treemap heatmap landscape discussed in the comparison section.
• Financial Markets and Portfolio Management — "Momentum: what do we know 30 years after Jegadeesh and Titman's seminal paper?" (retrospective on the momentum anomaly).
• North American Journal of Economics and Finance — "Dynamic time series momentum of cryptocurrencies" (time-series momentum evidence in crypto markets).
• Financial Markets and Portfolio Management — "Cryptocurrency momentum has (not) its moments" (crypto momentum crashes and instability across periods).
• Internal: the live bubble map, tokens, DeFi, L2s, and restaking dashboards described throughout this page.

Disclaimer: ETH Bubbles visualizes third-party market data for information only. Bubble sizes, colors, and percentage changes are not trading signals, endorsements, or financial advice, and data may be delayed or contain upstream errors. Always verify figures independently before making any decision.