A log of the live cost of $1M orders on Binance and Coinbase, measured every minute, against the expected cost on OneBook.
A desk asks three things: what it costs against a named benchmark (Live cost, below, vs Binance by tier), how fast it fills (Time to fill), and fill quality (no live fills yet; every fill gets a receipt vs the best route, and markouts publish here when fills exist). 1 bps = 0.01%, so 10 bps on a $1M order is $1,000. Live = measured every minute from public order books. Expected = our cost model. VIP5+ plan = proposed pricing for VIP5 and VIP9 desks: no taker fee, and the house maker quotes its cost floor + 0.25 bps. 686,658 order-book samples since 2026-10-27 00:02 UTC.
All-in cost from the mid, in bps. OneBook is modeled. The rest is live.
Order size is on a log scale. Binance is a split order of $10K children, with and without a square-root impact estimate, and one instant order for reference. Coinbase is the book cost of one instant order.
Solver and intent routes are the cheapest of 0x, LI.FI, CoW Swap, NEAR Intents, Kyber and Jupiter, with buy and sell averaged against the Binance mid. The two OneBook bands are standard pricing and the proposed VIP5+ plan.
Split Binance all-in cost minus the expected OneBook cost, in bps. Each cell shows the high-cost case, then the low-cost case. VIP5 and VIP9 use the proposed VIP5+ plan.
OneBook cheaper in both casesclose to evenBinance cheaper
| pair | vs regular | vs VIP3 | vs VIP5 | vs VIP9 |
|---|---|---|---|---|
| BTC | +5.0 to +7.5 | +1.0 to +3.5 | +1.6 to +2.0 | +0.8 to +1.2 |
| ETH | +5.0 to +7.5 | +1.0 to +3.5 | +1.5 to +1.9 | +0.7 to +1.1 |
| SOL | +5.6 to +8.1 | +1.6 to +4.1 | +1.8 to +2.2 | +1.0 to +1.4 |
| pair | size | impact of the split order, bps | OneBook incl. hedge impact, bps | VIP5+ plan, bps | vs regular | vs VIP3 | vs VIP5 | vs VIP9 |
|---|---|---|---|---|---|---|---|---|
| BTC | $100,000 | 0.9 to 1.7 | 2.5 to 5.6 | 0.92 to 2.46 | +7.5 | +3.5 | +3.3 | +2.5 |
| BTC | $1,000,000 | 2.7 to 5.4 | 2.6 to 6.9 | 1.00 to 3.78 | +10.2 | +6.2 | +5.9 | +5.1 |
| ETH | $100,000 | 1.7 to 3.5 | 2.6 to 6.2 | 0.98 to 3.10 | +8.8 | +4.8 | +4.5 | +3.7 |
| ETH | $1,000,000 | 5.5 to 10.9 | 2.7 to 8.9 | 1.14 to 5.74 | +14.1 | +10.1 | +9.9 | +9.1 |
| SOL | $100,000 | 2.7 to 5.5 | 2.6 to 6.9 | 1.08 to 3.87 | +10.8 | +6.8 | +6.4 | +5.6 |
| SOL | $1,000,000 | 8.7 to 17.3 | 2.9 to 11.1 | 1.32 to 8.05 | +19.3 | +15.3 | +15.0 | +14.2 |
| pair | size | adverse selection | stale quotes | hedge | carry | margin | plan quote, 0 fee | vs VIP5 split | vs VIP9 split | vs VIP9 split + impact |
|---|---|---|---|---|---|---|---|---|---|---|
| BTC | $100,000 | 0.4 to 0.8 | 0.00 | 0.23 | 0.05 | 0.25 | 0.94 to 1.34 | +1.76 to +2.16 | +0.96 to +1.36 | +2.25 to +2.65 |
| BTC | $1,000,000 | 0.4 to 0.8 | 0.00 | 0.35 | 0.05 | 0.25 | 1.05 to 1.45 | +1.65 to +2.05 | +0.85 to +1.25 | +4.93 to +5.33 |
| ETH | $100,000 | 0.4 to 0.8 | 0.03 | 0.29 | 0.05 | 0.25 | 1.02 to 1.42 | +1.70 to +2.10 | +0.90 to +1.30 | +3.49 to +3.89 |
| ETH | $1,000,000 | 0.4 to 0.8 | 0.03 | 0.52 | 0.05 | 0.25 | 1.25 to 1.65 | +1.47 to +1.87 | +0.67 to +1.07 | +8.86 to +9.26 |
| SOL | $100,000 | 0.4 to 0.8 | 0.08 | 0.35 | 0.05 | 0.25 | 1.13 to 1.53 | +2.14 to +2.54 | +1.34 to +1.74 | +5.45 to +5.85 |
| SOL | $1,000,000 | 0.4 to 0.8 | 0.08 | 0.72 | 0.05 | 0.25 | 1.50 to 1.90 | +1.77 to +2.17 | +0.97 to +1.37 | +13.96 to +14.36 |
| pair | stale quotes, shock day | plan quote | vs VIP9 split | vs VIP9 split + impact |
|---|---|---|---|---|
| BTC | 0.78 | 1.83 to 2.23 | +0.07 to +0.47 | +4.16 to +4.56 |
| ETH | 1.58 | 2.80 to 3.20 | -0.87 to -0.47 | +7.31 to +7.71 |
| SOL | 1.78 | 3.20 to 3.60 | -0.73 to -0.33 | +12.26 to +12.66 |
VIP5+ plan. A split order costs a VIP5 or VIP9 desk about its fee, 2.3 to 3.1 bps, so OneBook cannot win them on a fee. The plan has no taker fee, and the house maker quotes at its cost floor plus 0.25 bps. The floor is adverse selection (0.4 to 0.8 bps), stale quotes over 2 s of exposure, hedging the netted 12% of flow passively at 1.5 bps plus its impact, and bond carry of 0.05 bps.
The plan holds only if the hedge is passive, netting stays near the median, and the maker widens or pauses when volatility spikes. Revenue at 0.25 bps is $10,000 per $400M filled a month.
Market impact. Splitting does not avoid impact. A worked order still moves the price by about Y · σdaily · √(order / daily volume), the square-root law measured on Bitcoin by Donier and Bonart (2015), with Y = 0.5 to 1 (BTC $1.11B/day, σ 1.82%, ETH $0.55B/day, σ 2.57%, SOL $0.18B/day, σ 2.32%). OneBook pays impact only on the flow its maker hedges: 12% of gross flow in a median 30-minute window, 50% at the 99th percentile.
On Binance, the order is worked at 1% of volume, the pace that keeps the split price. The clock runs 8 seconds, and the line is one possible price path from live volatility. OneBook is the target design: one block. Finality under 1 s is measured on testnet.
Every maker posts a bond, its initial margin. Each settlement window pays variation margin. The risk is the adverse move on a maker's net position until close-out, not the size of each fill.
OneBook caps each maker's net fills per window at its bond, at admission. A crash then costs at most the bond. A router cannot do this, because it does not hold the book.
| model | maker capital per $1M of quoted depth | settles | who holds the funds | worst-case loss is bounded by | status |
|---|---|---|---|---|---|
| Off-exchange settlement (ClearLoop, MirrorX, Fireblocks) | $1,000,000 (1:1) | every 2, 4 or 24 h, or T+1 | a custodian | fully collateralized | ClearLoop, MirrorX, Fireblocks docs |
| Binance portfolio margin | $100,000 to $200,000 (5 to 10x on haircut collateral) | inside the exchange | the exchange | liquidation, then the insurance fund | Binance docs |
| Hyperliquid perpetuals | about $25,000 on BTC (up to 40x) | every block | on-chain | liquidation book, HLP vault, insurance fund, auto-deleveraging | public docs |
| OneBook, stage 1 (now) | $1,000,000 (k = 1) | every fill, in one block | the maker's on-chain account | fully funded: no credit risk | built, testnet |
| OneBook, stage 2 (after the gates) | BTC $123,457 (k 8.1), ETH $238,095 (k 4.2), SOL $169,492 (k 5.9), plus a 5 to 15% USDC buffer | variation margin every few seconds to minutes | the maker's on-chain account | net fills per window capped at the bond: zero breaches in the Jan to Oct 2025 replay | replay measured, ledger planned |
| parameter | stage 1 (now) | stage 2 (after the gates) |
|---|---|---|
| coverage k | 1: the company is the only maker | start at 4 to 8 per pair (zero-breach ceiling: BTC 8, ETH 4, SOL 6) |
| initial margin | 100% USDC; one balance backs all pairs | sized to a 30 to 60 min close-out at the year's worst move, correlation 1 across pairs and makers |
| variation margin | shadow ledger only, no settlement | cash USDC from an on-chain buffer every few seconds to minutes |
| quote life | 1 to 2 blocks | same, plus cancels ahead of takes in each block as a later step |
| fill cap | published lot-cap schedule; every house fill hedged or capped | net fills per window capped at the bond, at admission |
| mark price | n/a | OneBook's own trade mid plus an on-chain reference; no mark means no new fills |
| chain halts | n/a | positions mark at the resume price; the settlement interval is shorter than the halt-free run |
| who takes a loss | the company: market risk, no credit risk | the maker's bond first; a default fund only after 3 or more bonded makers and a legal read |
| gate | target |
|---|---|
| maker term sheet | 1 signed |
| live markouts | 14 days of per-window data, -1 bp or better |
| chain liveness | 14 halt-free days and a passed chaos-restart test |
| legal | counsel clears a fully collateralized vault |
| volume | median $2M a day |
Break-even volume = monthly cost / net take. Bond carry is 0.03 to 0.09 bps per $ filled, so capital cost is not the limit. The limit is how much bond backs the quoted depth.
k = exposure per $1 of bond. The stage-2 k is the largest with zero breaches in a replay of every minute from 1 January to 31 October 2025, with the maker defaulting at its worst minute and a 30-minute close-out. That matches exchange portfolio margin, with no exchange holding the funds and every fill settled in one block.
Without the cap, netting is worth only 1.3 to 2.8x at the 99th percentile, because stress multiplies the dollar net about 10x while imbalance stays flat.
Your order on Binance, split into $10K child orders, against the expected OneBook cost. The live book walk fetches Binance from your browser.
Each lever we can pull, applied to the same live inputs. Binance numbers are live. OneBook numbers are expected.
| principle | Binance today | OneBook | status |
|---|---|---|---|
| Cost | VIP9 desk, BTC $1M split: 2.30 bps, 6.38 with market impact | 1.05 to 1.45 bps (VIP5+ plan) | Binance live, OneBook expected |
| Speed | 2.2 h to work it at 1% of volume, ±54 bps price risk (1σ) | one block, under 1 s, at a firm price | finality measured |
| Ease of access | 6 exchange accounts, each with its own KYC and funding, to reach 89% of top-10 volume | one account; makers connect the venues | planned |
| Control and risk visibility | the exchange holds your funds | funds stay in an on-chain account; each fill shows its price against the best route | built, testnet |
| Capital efficiency | funds parked at every exchange you use | makers cover up to 8.1x (BTC), 4.2x (ETH), 5.9x (SOL) their bond with zero breaches in the 1 Jan to 31 Oct 2025 replay; 88% of flow nets inside OneBook at the median | replay measured, ledger planned |
Snapshot 2026-11-09 08:55 UTC: one buy and one sell per route against the Binance mid, gas and fees included where the API reports them. Dry-run quotes; nothing is traded.
| route | chain | pair | size | buy | sell | cost, both sides averaged | status |
|---|---|---|---|---|---|---|---|
| CoW Swap | base | BTC | $100,000 | -8.32 | +16.99 | 4.33 | ok |
| NEAR Intents | native | BTC | $100,000 | +16.80 | +33.00 | 24.90 | ok |
| LI.FI | base | BTC | $100,000 | +13.78 | +38.85 | 26.31 | ok |
| 0x | base | BTC | $100,000 | 403 You cannot consume this service | |||
| CoW Swap | base | BTC | $1,000,000 | -1.41 | +29.36 | 13.97 | ok |
| NEAR Intents | native | BTC | $1,000,000 | +20.72 | +33.41 | 27.07 | ok |
| LI.FI | base | BTC | $1,000,000 | +15.94 | +100.97 | 58.45 | ok |
| 0x | base | BTC | $1,000,000 | 403 You cannot consume this service | |||
| CoW Swap | base | ETH | $100,000 | -12.10 | +17.86 | 2.88 | ok |
| Across | bridge | ETH | $100,000 | +9.50 | +0.75 | 5.13 | ok |
| NEAR Intents | native | ETH | $100,000 | +23.21 | +27.95 | 25.58 | ok |
| LI.FI | base | ETH | $100,000 | +8.24 | +43.14 | 25.69 | ok |
| deBridge | base→eth | ETH | $100,000 | +38.08 | +49.63 | 43.86 | ok |
| 0x | base | ETH | $100,000 | 403 You cannot consume this service | |||
| Across | bridge | ETH | $1,000,000 | +9.80 | +0.75 | 5.28 | ok |
| CoW Swap | base | ETH | $1,000,000 | +1.28 | +22.28 | 11.78 | ok |
| NEAR Intents | native | ETH | $1,000,000 | +31.07 | +39.73 | 35.40 | ok |
| LI.FI | base | ETH | $1,000,000 | +17.88 | +60.47 | 39.18 | ok |
| deBridge | base→eth | ETH | $1,000,000 | +44.65 | +62.77 | 53.71 | ok |
| 0x | base | ETH | $1,000,000 | 403 You cannot consume this service | |||
| Jupiter | solana | SOL | $100,000 | -10.63 | +12.13 | 0.75 | ok |
| CCTP + Jupiter | base→sol | SOL | $100,000 | -9.33 | +13.13 | 1.90 | ok |
| NEAR Intents | native | SOL | $100,000 | +20.36 | +29.87 | 25.11 | ok |
| LI.FI | solana | SOL | $100,000 | +14.85 | +36.51 | 25.68 | ok |
| deBridge | base→sol | SOL | $100,000 | +35.46 | +47.83 | 41.64 | ok |
| Jupiter | solana | SOL | $1,000,000 | -1.95 | +33.40 | 15.73 | ok |
| CCTP + Jupiter | base→sol | SOL | $1,000,000 | -0.65 | +34.40 | 16.88 | ok |
| LI.FI | solana | SOL | $1,000,000 | +92.40 | +75.77 | 84.09 | ok |
| deBridge | base→sol | SOL | $1,000,000 | +55.24 | +125.86 | 90.55 | ok |
| NEAR Intents | native | SOL | $1,000,000 | 400 No liquidity available |
| pair | size | best solver route | buy / sell cost | solver cost, both sides averaged | house maker plan | best of the two | vs VIP9 split |
|---|---|---|---|---|---|---|---|
| BTC | $100,000 | kyber on base | +1.83 / -0.69 | 0.57 | 0.94 to 1.34 | 0.57 | +1.73 |
| BTC | $1,000,000 | kyber on base | +8.32 / +4.11 | 6.22 | 1.05 to 1.45 | 1.05 to 1.45 | +0.85 to +1.25 |
| ETH | $100,000 | kyber on base | +2.29 / +0.28 | 1.29 | 1.02 to 1.42 | 1.02 to 1.29 | +1.03 to +1.30 |
| ETH | $1,000,000 | cow on ethereum | +5.60 / +3.27 | 4.43 | 1.25 to 1.65 | 1.25 to 1.65 | +0.67 to +1.07 |
| SOL | $100,000 | jupiter on solana | +2.41 / -0.32 | 1.05 | 1.13 to 1.53 | 1.05 | +1.83 |
| SOL | $1,000,000 | jupiter on solana | +7.59 / +4.74 | 6.17 | 1.50 to 1.90 | 1.50 to 1.90 | +0.97 to +1.37 |
Hollow = $100K, filled = $1M, worked at 1% and 5% of volume. Time = order / (share of volume × 30-day daily volume); risk = σ√t. Exact values in the fill-time table.
| pair | size | time at 1% of volume | price risk while it works (1σ) | time at 5% of volume | price risk (1σ) | OneBook |
|---|---|---|---|---|---|---|
| BTC | $100,000 | 13 min | ±17 bps | 3 min | ±8 bps | one block, under 1 s, firm price |
| BTC | $1,000,000 | 2.2 h | ±54 bps | 26 min | ±24 bps | one block, under 1 s, firm price |
| ETH | $100,000 | 26 min | ±35 bps | 5 min | ±15 bps | one block, under 1 s, firm price |
| ETH | $1,000,000 | 4.3 h | ±109 bps | 52 min | ±49 bps | one block, under 1 s, firm price |
| SOL | $100,000 | 80 min | ±55 bps | 16 min | ±24 bps | one block, under 1 s, firm price |
| SOL | $1,000,000 | 13.3 h | ±173 bps | 2.7 h | ±77 bps | one block, under 1 s, firm price |
Exact values in the netting table and data/netting_sensitivity.csv; drag the curve tab to explore.
| pair | 100% one-way | 50% net | 25% net | 12% net (2025 median) | break-even net share vs VIP9 |
|---|---|---|---|---|---|
| BTC | 6.49 (-4.19) | 3.10 (-0.80) | 1.79 (+0.51) | 1.25 (+1.05) | 36% |
| ETH | 10.62 (-8.29) | 4.58 (-2.25) | 2.33 (-0.01) | 1.45 (+0.87) | 25% |
| SOL | 15.47 (-12.59) | 6.32 (-3.45) | 2.98 (-0.11) | 1.70 (+1.17) | 24% |
| tier | BTC: saving before fee → fee at half | ETH: saving before fee → fee at half | SOL: saving before fee → fee at half |
|---|---|---|---|
| regular | 8.75 → fee 4.37 | 8.57 → fee 4.29 | 8.87 → fee 4.44 |
| VIP3 | 4.75 → fee 2.37 | 4.57 → fee 2.29 | 4.87 → fee 2.44 |
| VIP5 | 1.85 → fee 0.92 | 1.67 → fee 0.84 | 1.97 → fee 0.99 |
| VIP9 | 1.05 → fee 0.52 | 0.87 → fee 0.44 | 1.17 → fee 0.59 |
Captured quotes. OneBook fills each order at the cheaper of its house maker quote and the best solver quote, from 7,384 cycles of quotes (not fills), buy and sell medians averaged.
Time to fill. Time = order / (share of volume × Binance 30-day daily volume). Price risk = σdaily · √(days). A OneBook fill is a firm price, known before the trade.
Netting. The house maker hedges only the flow that does not net inside OneBook. At launch flow is one-way and the plan loses to VIP9 at $1M; it wins once net flow falls below the break-even share. Competing makers push the 0.25 bps margin toward 0.
Fees. Saving before any OneBook fee = split Binance cost minus the house maker cost at $1M. A fee of half the saving leaves the desk the other half.
A split Binance order is $10K children, each paying the live book cost and the taker fee. Sampled every minute.
Binance, one orderBinance, splitOneBook expected
Exact values in the cost table tab and data/*.csv.
| pair | size | split as | Binance book, split | Binance book, instant | Coinbase book, instant | split all-in, regular | VIP3 | VIP5 | VIP9 | OneBook expected | OneBook, VIP5+ plan |
|---|---|---|---|---|---|---|---|---|---|---|---|
| BTC | $10,000 | 1 x $10K | 0.00 | 0.00 | 0.21 | 10.00 | 6.00 | 3.10 | 2.30 | 2.5 to 5.0 | 0.90 to 1.30 |
| BTC | $100,000 | 10 x $10K | 0.00 | 0.00 | 0.96 | 10.00 | 6.00 | 3.10 | 2.30 | 2.5 to 5.0 | 0.94 to 1.34 |
| BTC | $1,000,000 | 100 x $10K | 0.00 | 0.71 | 4.01 | 10.00 | 6.00 | 3.10 | 2.30 | 2.5 to 5.0 | 1.05 to 1.45 |
| ETH | $10,000 | 1 x $10K | 0.02 | 0.02 | 0.46 | 10.02 | 6.02 | 3.12 | 2.32 | 2.5 to 5.0 | 0.95 to 1.35 |
| ETH | $100,000 | 10 x $10K | 0.02 | 0.16 | 1.53 | 10.02 | 6.02 | 3.12 | 2.32 | 2.5 to 5.0 | 1.02 to 1.42 |
| ETH | $1,000,000 | 100 x $10K | 0.02 | 2.67 | 6.76 | 10.02 | 6.02 | 3.12 | 2.32 | 2.5 to 5.0 | 1.25 to 1.65 |
| SOL | $10,000 | 1 x $10K | 0.58 | 0.58 | 1.37 | 10.58 | 6.58 | 3.68 | 2.88 | 2.5 to 5.0 | 1.02 to 1.42 |
| SOL | $100,000 | 10 x $10K | 0.58 | 0.90 | 3.88 | 10.58 | 6.58 | 3.68 | 2.88 | 2.5 to 5.0 | 1.13 to 1.53 |
| SOL | $1,000,000 | 100 x $10K | 0.58 | 9.37 | 12.70 | 10.58 | 6.58 | 3.68 | 2.88 | 2.5 to 5.0 | 1.50 to 1.90 |
There are no live OneBook fills yet, so there is no fill-quality number to show, and we will not invent one.
What exists today: every testnet fill carries a receipt listing the alternatives at fill time, the route chosen, the all-in cost in bps from the mid, and our charge. When live fills exist, this tab becomes their record: per-window markouts against the arrival mid, the share of fills inside the expected band, and the worst fill of each day, good or bad.
The gate we hold ourselves to before scaling a pilot: 14 days of per-window markouts at −1 bp or better.
Book cost is the price impact of a market order, both sides, median of the last 15 minutes, in bps from the mid price. The split cost leaves out price drift while the order works. The instant column is one market order for the full size, for reference.
What the system has demonstrated, under what conditions, and the market it sits in.
| component | what it does | status |
|---|---|---|
| Matching capability | 690M+ messages/s | benchmarked |
| Finality | under 1 s, in one hop | measured, wall clock |
| Executor throughput | ~232,898 tx/s on 1 node, 1M tx, full Ed25519 checks. Excludes consensus and networking. | measured |
| Virtual machines | 5 on one shared state: SVM, EVM, WASM, Move, ZK | built |
| Token standard | one 84-byte header that all 5 VMs read | built |
| Hilsa | checks other chains directly, peer-to-peer, with no relayers. 30+ chains in the catalog. Solana in the default build. Polkadot, Aptos, and Cosmos on a local network. Ethereum in progress. | built |
| OneBook | the order book. Orders match and settle in the same block. Quotes expire after 1 to 2 blocks. | built, testnet |
| Fill receipts | every fill shows its price against the best route, in bps | built, testnet |
| Deposits | open after a scheduled network upgrade | planned |
| Clearing ledger | tracks each maker's net exposure against its bond | planned |
| metric | value | source |
|---|---|---|
| Total value of digital assets | $2.73T | CoinGecko |
| 24 h traded volume | $62.3B | CoinGecko |
| Assets actively traded | 22,521 | CoinGecko |
| Exchanges listed | 1,513 | CoinGecko |
| BTC share of total value | 59.4% | CoinGecko |
| Blockchains tracked | 421 | DefiLlama |
| Blockchains holding value (TVL > 0) | 294 | DefiLlama |
| Value locked on-chain (DeFi TVL) | $88.7B | DefiLlama |
| Stablecoins / supply | 425 / $303.3B | DefiLlama |
| On-chain trading protocols (DEXs) | 1,573 | DefiLlama |
| DEX volume, 24 h / 30 d | $8.4B / $253.3B | DefiLlama |
| exchange | share of top-10 volume, % |
|---|---|
| Binance | 45.7 |
| Kraken | 10.0 |
| Gate | 9.4 |
| OKX | 9.0 |
| Coinbase Exchange | 7.7 |
| MEXC | 7.1 |
| Bitget | 4.8 |
| Crypto.com Exchange | 3.9 |
| Bitstamp by Robinhood | 2.1 |
| HashKey Exchange | 0.4 |
| holds your funds | moves capital to the price | settles atomically | sees other chains directly | |
|---|---|---|---|---|
| Routers / execution systems | no, you pre-fund each venue | no | no, each venue settles alone | no |
| Centralized exchanges | yes, internal ledger | inside one exchange | inside one exchange | no |
| Single-chain order books | on-chain | no, assets bridged in | on their own chain | no |
| The Singularity | on-chain accounts | yes, bonded makers | yes, one block | yes, Hilsa |
How much exposure $1 of bond covers with no loss, and what out-of-date quotes cost a maker. From Binance public candles and trades.
Every minute, 1 Jan to 31 Oct 2025, maker defaulting at its worst minute. Breach counts: bond tables tab and data/exp3_k_safe.csv.
Worst 30-minute adverse move each day, 1 January to 31 October 2025, in bps. A maker at coverage k loses more than its bond on any day its line crosses the k line. BTC crosses k = 10 once, on 10 October.
Largest k with zero breaches
| pair | close-out, min | worst adverse move, bps | when (UTC) | p99, bps | p99.9, bps | largest k with zero breaches |
|---|---|---|---|---|---|---|
| BTC | 5 | 818.8 | 2025-10-10 21:15 | 61.7 | 130.4 | 12.2 |
| BTC | 30 | 1239.8 | 2025-10-10 20:50 | 157.5 | 317.4 | 8.1 |
| BTC | 60 | 1290.9 | 2025-10-10 20:27 | 224.1 | 426.0 | 7.7 |
| ETH | 5 | 1541.2 | 2025-02-03 02:03 | 103.3 | 231.9 | 6.5 |
| ETH | 30 | 2361.5 | 2025-02-03 01:40 | 253.6 | 504.2 | 4.2 |
| ETH | 60 | 2554.2 | 2025-10-10 21:20 | 374.3 | 704.7 | 3.9 |
| SOL | 5 | 1186.9 | 2025-10-10 21:39 | 119.6 | 259.3 | 8.4 |
| SOL | 30 | 1687.2 | 2025-10-10 21:27 | 295.3 | 604.6 | 5.9 |
| SOL | 60 | 2194.9 | 2025-03-02 15:22 | 417.6 | 853.9 | 4.6 |
Breaches at a 30-minute close-out
| pair | k | breach windows | share of minutes, % | episodes | of them on 10 Oct | worst loss, in bonds |
|---|---|---|---|---|---|---|
| BTC | 1 | 0 | 0.0 | 0 | 0 | 0.12 |
| BTC | 5 | 0 | 0.0 | 0 | 0 | 0.62 |
| BTC | 10 | 23 | 0.005 | 1 | 23 | 1.24 |
| BTC | 20 | 85 | 0.019 | 6 | 46 | 2.48 |
| ETH | 1 | 0 | 0.0 | 0 | 0 | 0.24 |
| ETH | 5 | 18 | 0.004 | 2 | 5 | 1.18 |
| ETH | 10 | 127 | 0.029 | 3 | 61 | 2.36 |
| ETH | 20 | 451 | 0.103 | 27 | 82 | 4.72 |
| SOL | 1 | 0 | 0.0 | 0 | 0 | 0.17 |
| SOL | 5 | 0 | 0.0 | 0 | 0 | 0.84 |
| SOL | 10 | 81 | 0.019 | 4 | 46 | 1.69 |
| SOL | 20 | 824 | 0.188 | 37 | 83 | 3.37 |
Quote life on each x axis; one shared log y axis. Exact values in the stale table tab and data/exp4_staleness.csv.
| pair | day | quote life, s | windows | avg excursion, bps | pick-off loss per side per refresh, bps | stale-hit probability |
|---|---|---|---|---|---|---|
| BTC | calm control (Sep 16) | 1 | 70296 | 0.042 | 0.0016 | 0.0018 |
| BTC | calm control (Sep 16) | 2 | 41902 | 0.074 | 0.0043 | 0.0051 |
| BTC | calm control (Sep 16) | 4 | 21581 | 0.142 | 0.0131 | 0.0133 |
| BTC | crash (Oct 10) | 1 | 79334 | 0.832 | 0.5793 | 0.0712 |
| BTC | crash (Oct 10) | 2 | 42774 | 1.115 | 0.7778 | 0.0976 |
| BTC | crash (Oct 10) | 4 | 21597 | 1.614 | 1.1485 | 0.142 |
| BTC | tariff shock (Apr 7) | 1 | 84577 | 0.62 | 0.1628 | 0.1039 |
| BTC | tariff shock (Apr 7) | 2 | 43150 | 1.07 | 0.3936 | 0.1933 |
| BTC | tariff shock (Apr 7) | 4 | 21597 | 1.801 | 0.8746 | 0.3136 |
| ETH | calm control (Sep 16) | 1 | 56300 | 0.137 | 0.0157 | 0.0108 |
| ETH | calm control (Sep 16) | 2 | 38377 | 0.214 | 0.032 | 0.0224 |
| ETH | calm control (Sep 16) | 4 | 21371 | 0.375 | 0.0769 | 0.0525 |
| ETH | crash (Oct 10) | 1 | 70809 | 1.685 | 1.2278 | 0.1373 |
| ETH | crash (Oct 10) | 2 | 41556 | 2.128 | 1.5784 | 0.171 |
| ETH | crash (Oct 10) | 4 | 21547 | 3.049 | 2.3341 | 0.2391 |
| ETH | tariff shock (Apr 7) | 1 | 82181 | 1.224 | 0.4989 | 0.2184 |
| ETH | tariff shock (Apr 7) | 2 | 42809 | 1.969 | 1.007 | 0.3338 |
| ETH | tariff shock (Apr 7) | 4 | 21583 | 3.124 | 1.9248 | 0.468 |
| SOL | calm control (Sep 16) | 1 | 55767 | 0.28 | 0.0392 | 0.0359 |
| SOL | calm control (Sep 16) | 2 | 38591 | 0.449 | 0.0827 | 0.07 |
| SOL | calm control (Sep 16) | 4 | 21358 | 0.767 | 0.1992 | 0.1384 |
| SOL | crash (Oct 10) | 1 | 65853 | 1.999 | 1.4474 | 0.1771 |
| SOL | crash (Oct 10) | 2 | 40412 | 2.458 | 1.7824 | 0.2147 |
| SOL | crash (Oct 10) | 4 | 21475 | 3.426 | 2.5548 | 0.2924 |
| SOL | tariff shock (Apr 7) | 1 | 77283 | 1.375 | 0.6782 | 0.2312 |
| SOL | tariff shock (Apr 7) | 2 | 42382 | 2.224 | 1.2807 | 0.3392 |
| SOL | tariff shock (Apr 7) | 4 | 21569 | 3.587 | 2.3889 | 0.4697 |
| pair | size | best solver route | cost, bps | vs regular | vs VIP3 | vs VIP5 | vs VIP9 |
|---|---|---|---|---|---|---|---|
| BTC | $100,000 | kyber on base | 0.57 | +10.25 | +6.25 | +3.35 | +2.55 |
| BTC | $1,000,000 | kyber on base | 6.22 | +6.77 | +2.77 | -0.13 | -0.93 |
| ETH | $100,000 | kyber on base | 1.29 | +10.09 | +6.09 | +3.19 | +2.39 |
| ETH | $1,000,000 | cow on ethereum | 4.43 | +11.48 | +7.48 | +4.58 | +3.78 |
| SOL | $100,000 | jupiter on solana | 1.05 | +12.65 | +8.65 | +5.75 | +4.95 |
| SOL | $1,000,000 | jupiter on solana | 6.17 | +17.34 | +13.34 | +10.44 | +9.64 |
| item | value | source |
|---|---|---|
| Binance spot taker fee | 10 bps regular, 6 VIP3, 3.1 VIP5, 2.3 VIP9 | Binance fee schedule |
| Coinbase executable cost, $100K / $1M | BTC 0.8 / 3.0 bps, ETH 1.4 / 5.9, SOL 3.7 / 13.6 | Coinbase order book |
| Off-exchange settlement (ClearLoop, MirrorX, Fireblocks) | 1:1 collateral, net settlement every 2, 4 or 24 h, or T+1 | ClearLoop, MirrorX, Fireblocks docs |
| ClearLoop notional, July 2026 | $74.8B | Copper ClearLoop |
| CEX portfolio margin for makers | 5 to 10x on haircut collateral | Binance portfolio margin docs |
| FX netting (CLS) | members fund ~1% of >$8T/day gross | CLS Group |
| Equity netting (NSCC CNS) | nets 98.6%: $2.2T/day to $33.5B | DTCC NSCC |
| Default fund as share of initial margin | 13.1% global, 2.9% CME | CCP public disclosures |
| Market price of a maker | 0 maker fee + 0.1 to 0.8 bps rebate | venue maker fee schedules |
| Adverse selection, passive fill at 10 s | 0.4 to 0.8 bps | market microstructure studies |
| Binance 30-min taker flow imbalance, 2025 | p50 0.11 to 0.14, p99 0.43 to 0.53 | computed, Binance 2025 klines |
| BTC / ETH / SOL correlation in stress | 0.87 to 0.90 | computed, Binance 2025 klines |
Every live figure, its endpoint and how often it refreshes. Every research table is a file.
| figure | endpoint | refresh | latest |
|---|---|---|---|
| Binance book cost | data-api.binance.vision/api/v3/depth (limit 5000) | every 60 s, plus your browser on demand | 4s ago |
| Coinbase book cost | api.exchange.coinbase.com/products/<pair>-USD/book?level=2 | every 60 s | 4s ago |
| Binance volume and volatility | data-api.binance.vision/api/v3/klines, 1d, last 31 | hourly | cached |
| Binance price candles | data-api.binance.vision/api/v3/klines, 5m, last 288 | every 4 min | cached |
| Solver and intent routes | api.0x.org, li.quest/v1, api.cow.fi, 1click.chaindefuser.com | every 30 min | 2026-11-09 08:55 UTC |
| Market size | CoinGecko /global and /exchanges; DefiLlama chains, stablecoins, DEXs | read once | 2026-10-27 |
| Bond replay and stale quotes | Binance public 1-minute candles and trades, 1 Jan to 31 Oct 2025 | one-off | 2026-10-26 |