Charts

• QRShipmentTrace is left out. Seven queries in the entire history and none at all since 1 August, so it would be a segment no bar could show.
• The bar is GeoCode + TrustSupply, and the shares are of that sum. Per-type figures are derived — each type’s queries that day times the day’s validators-per-query factor, rounded per type — so they can total a few more or fewer than the day’s headline Validations figure (3 Aug: 4,634 against 4,550).
• The last bar is today, still filling, so it sits low until the day closes.
• The months behind these days, and every day back to 24 May 2026, are in the Validations card above under Daily.

Three things shape what this chart can show

• No offline count. The API's active-validator figure disagrees with its own average nodes-per-validator, so none is shown rather than a number that contradicts its own source. Node health is on the Nodes chart, which counts nodes.
• Month-end sampling. Each point is the level on the last day. That steps over highs inside a month — 684 was reached in July 2024, not on a boundary — which is why the peak is quoted in the header.
• Churn is in the table. The same months carry the joiners and leavers behind each change, so a level that looks flat can still be hiding movement.

Three things to know when reading it

• The 2024 dip is an artefact. A short-lived experiment set a node at 900 MNW before being abandoned, so bars in that window read low — the divisor was right, the stake per node was not.
• The July 2024 reset is invisible. On the 25th every position was closed on one day (2,090 nodes) to mark the start of production that August; restaking began on the 29th, so the month ends near its normal level.
• Below uptime is carried forward from the last day the API answered. A run of identical values means the API skipped days, not that the network held steady.

• Closes come from the Uniswap V3 MNW/WETH 0.3% pool on Ethereum — the deepest and busiest MNW venue on either chain, and the pair DEXTools charts. Points before 29 Aug 2026 are a frozen CoinGecko copy, because the pool’s public history only reaches 180 days back. The pool runs a median 2.1% below CoinGecko over the 178 days both cover, but individual days range from -24% to +13%, so expect a step at that date rather than a clean join.
• All five periods come from the same 365 daily closes, so the high and low of a candle are the highest and lowest CLOSE in that bucket, not the intraday extremes.
• Candles rather than a line because a line of period-closing values throws the range away: the yearly view was two points on an axis topping out at $0.060 while the year traded to $0.19.
• A filled body means the period closed above where it opened, hollow means below. Direction is not carried by colour alone.
• Weeks start on Monday; the current period is still open and its close is the latest price.

• Balances are rebuilt from every MNW transfer in and out of the pool wallet, then walked forward day by day, so a month with no movement holds its level.
• The opening balance is back-derived: today’s balance minus the net over the window. The final day is checked against the token contract and the build refuses to write if they disagree.
• Asking for the balance at a past block directly does not work on the free tier: the block tag is ignored without saying so. Measured 19 Aug 2026, it returned one identical figure for 1 July and 1 August that matched neither real value.
• Movement here is not a trade count. Liquidity is added and removed as well as swapped, so a pool draining can mean a provider withdrew rather than that MNW was sold.

• The sum of the table above, not a separate measurement. Two lines rather than eight because the pools differ by five orders of magnitude: MetaMask holds nothing, so on a shared axis most of them sit on the floor. Uniswap V3 is drawn on its own because it is the pool the price and volume cards are built from.
• The last point is keyed "now" and not by its date. On the 1st of a month the newest dated column IS today, and two readings of one date collapse into one point: the start of the day and this moment are different measurements.
• The last point is today's figure, the same one the bars above carry, so the final step covers a part month rather than a whole one.
• Liquidity is added and removed as well as swapped. A fall means a provider withdrew as often as it means MNW was sold, and the two cannot be told apart from a balance alone.

• A running balance, not a series of snapshots: each day is the previous day plus that day's net transfers, so a day with no movement holds its level rather than disappearing.
• The last point is checked against the balance read straight from the token contract. If they disagree the build refuses to write — a chart of a balance that never existed is worse than no chart.
• That check earned its keep immediately: asking Etherscan for 10,000 transfers silently returns 1,000, so Gate's walk ended in July 2022 and reconstructed 62,037 against a real 55,810. KuCoin has 76 transfers and reconciled perfectly, so one wallet alone would have "proved" the broken version right.
• The axis does not start at zero. The balance moves about 10% around 2.7M, which is a flat line on a 0 to 4M scale, so the plot is cropped to the range the data occupies. Read the bottom gridline before reading the slope.
• History starts when the wallet first held MNW — Gate 18 Oct 2021, KuCoin 17 Apr 2026 — so the two series do not span the same period.