Ghostex.ai measurement data =========================== Published 8 October 2026 by the Ghostex team, https://ghostex.ai/about#editorial Section 4, Bitcoin block times, added on 8 October 2026. Questions or corrections: support@ghostex.ai Licence: CC BY 4.0, https://creativecommons.org/licenses/by/4.0/ (credit Ghostex.ai and link https://ghostex.ai/data/) Where a Ghostex.ai page says "we measured", the raw readings and the method are here, so that anyone can repeat the measurement or check a figure. The readings come from public blockchain nodes and public block explorers, which anyone can query. None of these files is a record of Ghostex orders. Files in https://ghostex.ai/data/ (the page at that address lists them with their size and date) monero-block-times-2026-10-07.csv timestamps of 21,600 Monero blocks used on /coins/monero bitcoin-block-times-2026-10-08.csv hashes and timestamps of 4,032 Bitcoin used on /blog/how-long-does-a-crypto-swap-take blocks bsc-eth-block-gas-2026-10-07.json BNB Smart Chain and Ethereum block time, used on /exchange/usdtbsc-to-usdt gas price and gas of 12 USDT transfers each monero-address-check-2026-10-08.js the Monero address test (Node.js) used on /coins/monero monero-address-check-2026-10-08.json output of that test, run on 8 October 2026 measurement-methods.txt this file, with the three Python scripts below To cite a figure, name Ghostex.ai, the page, the file and its date, for example: Ghostex.ai, "Monero block times, 7 September to 7 October 2026", https://ghostex.ai/data/monero-block-times-2026-10-07.csv 1. Monero block times and the 10-block lock (monero-block-times-2026-10-07.csv) ------------------------------------------------------------------------------- What: the header timestamp of each of 21,600 consecutive Monero mainnet blocks, heights 3,757,375 to 3,778,974, from 2026-09-07 16:31:40 UTC to 2026-10-07 16:20:10 UTC (30 days). Columns: height, timestamp_unix (seconds, as written in the block header), timestamp_utc (the same in ISO 8601), gap_from_previous_s (this timestamp minus the one before it, in file order; empty on the first row). Source: read on 7 October 2026 at about 16:20 UTC with the get_block_headers_range call of the monerod JSON-RPC interface (https://docs.getmonero.org/rpc-library/monerod-rpc/#get_block_headers_range) on a public remote node. Cross-checks on 8 October 2026: all 21,600 timestamps were read again with the same call from a second public node, xmr-node.cakewallet.com:18081, and were identical; 21 heights, including the first, the last and the five out-of-order blocks, were compared with the block pages of xmrchain.net (https://xmrchain.net/api/block/) and were identical. Miners set block timestamps, so five blocks (heights 3,759,011, 3,769,491, 3,771,801, 3,774,677 and 3,775,906) carry a timestamp earlier than the block before them, and their gap in the file is negative. The figures below sort the timestamps first. Figures on /coins/monero, as the script below prints them from this file: average gap: 120.0 s over 30 days, 119.2 s over the last 7 days (5,040 gaps), 116.6 s over the last 720 blocks (720 gaps spanning 23.3 hours, about 24 hours); median gap 77 s; 43% of gaps 60 s or less; 6% over 6 minutes; longest gap 22 minutes. The 10-block lock is a model on these measured blocks, not a record of real payouts: a payout is assumed to be broadcast every 7 seconds across the 30 days (369,981 payouts), to go into the next block, the first one with a later timestamp, and to become spendable with its 10th confirmation, nine blocks after that one. Result: mean wait 20.0 minutes, median 19.3, longest 62; unlocked within 15 minutes 23%, 20 minutes 55%, 25 minutes 80%, 30 minutes 93%, 32 minutes 95%, 40 minutes 99.4%. The 10-block rule itself is in the Monero README (https://github.com/monero-project/monero#scheduled-softwarenetwork-upgrades). Script (save as monero_block_times.py next to the CSV; Python 3, standard library only): """Recompute the Monero block-time and 10-block-lock figures on https://ghostex.ai/coins/monero. Python 3, standard library only. python monero_block_times.py # reads monero-block-times-2026-10-07.csv next to this script python monero_block_times.py --fetch NODE_URL # also re-reads the same heights from a monerod node and compares NODE_URL is any public or own monerod RPC address, for example http://127.0.0.1:18081.""" import bisect, csv, json, os, statistics, sys, urllib.request CSV = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'monero-block-times-2026-10-07.csv') rows = list(csv.DictReader(open(CSV, newline=''))) heights = [int(r['height']) for r in rows] ts = [int(r['timestamp_unix']) for r in rows] print('blocks %d, heights %d to %d' % (len(ts), heights[0], heights[-1])) if '--fetch' in sys.argv: node = sys.argv[sys.argv.index('--fetch') + 1].rstrip('/') chain = {} for a in range(heights[0], heights[-1] + 1, 1000): b = min(a + 999, heights[-1]) body = json.dumps({'jsonrpc': '2.0', 'id': '0', 'method': 'get_block_headers_range', 'params': {'start_height': a, 'end_height': b}}).encode() req = urllib.request.Request(node + '/json_rpc', data=body, headers={'Content-Type': 'application/json'}) for h in json.loads(urllib.request.urlopen(req, timeout=60).read())['result']['headers']: chain[h['height']] = h['timestamp'] diff = [h for h, t in zip(heights, ts) if chain.get(h) != t] print('node check: %d of %d timestamps identical' % (len(ts) - len(diff), len(ts))) # Miners set block timestamps, so a few are earlier than the block before them; sort before measuring gaps. out_of_order = sum(1 for i in range(1, len(ts)) if ts[i] < ts[i - 1]) s = sorted(ts) gaps = [s[i] - s[i - 1] for i in range(1, len(s))] print('timestamps out of order: %d' % out_of_order) for label, n in (('last 720 blocks, about 24 hours (720 gaps)', 720), ('last 7 days (5,040 gaps)', 5040), ('30 days (all gaps)', len(gaps))): print('average gap, %s: %.1f s' % (label, (s[-1] - s[-1 - n]) / n)) print('median gap: %.0f s' % statistics.median(gaps)) print('gaps of 60 s or less: %.0f%%' % (100.0 * sum(1 for g in gaps if g <= 60) / len(gaps))) print('gaps over 6 minutes: %.0f%%' % (100.0 * sum(1 for g in gaps if g > 360) / len(gaps))) print('longest gap: %.0f minutes' % (max(gaps) / 60.0)) # Lock model: a payout is broadcast every 7 seconds across the 30 days, goes into the next block (the first one # with a later timestamp) and becomes spendable with its 10th confirmation, nine blocks after that one. waits, t = [], s[0] while True: j = bisect.bisect_right(s, t) if j + 9 >= len(s): break waits.append(s[j + 9] - t) t += 7 waits.sort() n = len(waits) print('modelled payouts: %d (a model on measured blocks, not a record of real payouts)' % n) print('wait to a spendable balance: mean %.1f min, median %.1f min, longest %.0f min' % (statistics.mean(waits) / 60.0, statistics.median(waits) / 60.0, waits[-1] / 60.0)) for m in (15, 20, 25, 30, 32, 40): share = 100.0 * sum(1 for w in waits if w <= m * 60) / n print('unlocked within %d minutes: %s%%' % (m, ('%.1f' % share) if m == 40 else ('%.0f' % share))) 2. BNB Smart Chain and Ethereum block time and gas (bsc-eth-block-gas-2026-10-07.json) --------------------------------------------------------------------------------------- What: one snapshot per chain, taken on 7 October 2026 at about 16:18 UTC: the latest block and the block 1,000 (BNB Smart Chain) or 100 (Ethereum) blocks earlier, with their timestamps, which give the average block time; the gas price the node suggested; the name, symbol and decimals of the USDT contract; and 12 successful direct USDT transfer calls (selector 0xa9059cbb, sent straight to the USDT contract) from the most recent blocks, with their gas used and effective gas price. Each transaction links to bscscan.com or etherscan.io. Source: standard Ethereum JSON-RPC calls to public nodes: https://bsc-dataseed.bnbchain.org (BNB Chain's own public endpoint) and https://ethereum-rpc.publicnode.com. Cross-check on 8 October 2026: the two block timestamps on each chain and the receipts of the first and last listed transaction on each chain were read again from the same nodes and matched the file. Figures on /exchange/usdtbsc-to-usdt: block time 0.451 s on BNB Smart Chain (1,000 blocks in 451 s) and 12.0 s on Ethereum (100 blocks in 1,200 s); gas used 29,715 to 51,615 on BNB Smart Chain and 38,497 to 63,221 on Ethereum; gas price 0.05 gwei (BNB Smart Chain) and a 5.12 gwei base fee (Ethereum). It is one snapshot, not a forecast: gas prices move with demand, and the cost in your wallet's fee preview is the one that applies. Script (Python 3, standard library only; it writes measure.json, from which the JSON file was laid out with the node names and explorer links added; the contract addresses are written in lower case here): import json, urllib.request, time def rpc(url, method, params): req = urllib.request.Request(url, data=json.dumps({"jsonrpc":"2.0","method":method,"params":params,"id":1}).encode(), headers={"Content-Type":"application/json","User-Agent":"Mozilla/5.0"}) return json.loads(urllib.request.urlopen(req, timeout=20).read())["result"] BSC = "https://bsc-dataseed.bnbchain.org" ETH = "https://ethereum-rpc.publicnode.com" USDT_ETH = "0xdac17f958d2ee523a2206206994597c13d831ec7" USDT_BSC = "0x55d398326f99059ff775485246999027b3197955" TRANSFER = "0xddf252ad1be2c89b69c2b068fc378daa952ba7f163c4a11628f55a4df523b3ef" def call(url, to, data): return rpc(url, "eth_call", [{"to": to, "data": data}, "latest"]) def dec_str(h): b = bytes.fromhex(h[2:]) if len(b) >= 64: ln = int.from_bytes(b[32:64], 'big'); return b[64:64+ln].decode(errors='replace') return b.rstrip(b'\0').decode(errors='replace') out = {} for name, url, tok in (("bsc", BSC, USDT_BSC), ("eth", ETH, USDT_ETH)): o = {} o["decimals"] = int(call(url, tok, "0x313ce567"), 16) o["name"] = dec_str(call(url, tok, "0x06fdde03")) o["symbol"] = dec_str(call(url, tok, "0x95d89b41")) latest = int(rpc(url, "eth_blockNumber", []), 16) span = 1000 if name == "bsc" else 100 b1 = rpc(url, "eth_getBlockByNumber", [hex(latest), False]) b0 = rpc(url, "eth_getBlockByNumber", [hex(latest - span), False]) t1 = int(b1["timestamp"], 16); t0 = int(b0["timestamp"], 16) o["latest_block"] = latest o["latest_ts"] = t1 o["latest_ts_utc"] = time.strftime("%Y-%m-%d %H:%M:%S", time.gmtime(t1)) o["span_blocks"] = span; o["span_seconds"] = t1 - t0; o["avg_block_s"] = (t1 - t0) / span o["gasPrice_gwei"] = int(rpc(url, "eth_gasPrice", []), 16) / 1e9 if "baseFeePerGas" in b1: o["baseFee_gwei"] = int(b1["baseFeePerGas"], 16)/1e9 gas = [] for bn in range(latest-1, latest-60, -1): blk = rpc(url, "eth_getBlockByNumber", [hex(bn), True]) for tx in blk["transactions"]: if tx["to"] and tx["to"].lower() == tok.lower() and tx["input"].startswith("0xa9059cbb"): r = rpc(url, "eth_getTransactionReceipt", [tx["hash"]]) if r["status"] == "0x1": gas.append((int(r["gasUsed"], 16), tx["hash"], bn, int(r["effectiveGasPrice"],16)/1e9)) if len(gas) >= 12: break if len(gas) >= 12: break o["direct_transfer_gasUsed"] = gas out[name] = o print(json.dumps(out, indent=1)) json.dump(out, open("measure.json", "w"), indent=1) 3. Monero address check (monero-address-check-2026-10-08.js and .json) ---------------------------------------------------------------------- What: the test of the swap form's Monero address check described on /coins/monero. The script loads the form's own validator, https://ghostex.ai/assets/js/address.js, generates 27,000 addresses (3,000 of each of nine types: standard, sub and integrated addresses on mainnet, testnet and stagenet, with random keys and a correct Keccak-256 checksum) and 20,000 mainnet addresses with one character changed, and counts what the validator accepts. Save address.js from the site next to the script and run: node monero-address-check-2026-10-08.js The page reports the run of 7 October 2026. monero-address-check-2026-10-08.json is the output of this script run again on 8 October 2026 against the address.js served on ghostex.ai that day (its SHA-256 is in the file): all 9,000 mainnet addresses accepted, all 18,000 testnet and stagenet addresses refused, none of the 20,000 one-character typos accepted. The keys are random, so a new run gives the same counts with different addresses. 4. Bitcoin block times and the 2-confirmation wait (bitcoin-block-times-2026-10-08.csv) --------------------------------------------------------------------------------------- What: the hash and header timestamp of each of 4,032 consecutive Bitcoin blocks, heights 966,438 to 970,469, from 2026-09-11 02:41:51 UTC to 2026-10-08 08:51:23 UTC (27.3 days, two difficulty periods' worth of blocks). Columns: height, hash, timestamp_unix (seconds, as written in the block header), timestamp_utc (the same in ISO 8601), interval_s (this timestamp minus the one before it, in file order; empty on the first row). Source: read on 8 October 2026 between 08:52 and 08:55 UTC with the /api/blocks/ call of mempool.space (https://mempool.space/docs/api/rest), which returns the 10 blocks at and below a height. Cross-check on 8 October 2026: all 4,032 heights were read again with the same call from a second public API, blockstream.info (https://blockstream.info/api/blocks/), and all 4,032 hashes and timestamps were identical. Bitcoin only requires a block's timestamp to be later than the median of the 11 blocks before it, so 131 blocks carry a timestamp earlier than the block before them, and their interval in the file is negative. The figures below sort the timestamps first. Figures on /blog/how-long-does-a-crypto-swap-take, as the script below prints them from this file: mean interval 584 s (9.7 minutes), median interval 403 s (6.7 minutes); longest interval 69 minutes. The 2-confirmation wait is a model on these measured blocks, not a record of real deposits: a deposit is assumed to be broadcast every 7 seconds across the 27.3 days (336,408 deposits), to go into the next block, the first one with a later timestamp, and to get its 2nd confirmation with the block after that. The wait is the time left to the next block plus the interval that follows it. A deposit whose fee is too low to make the next block waits longer; that is not modelled. Result: mean wait 19.6 minutes, median 16.4, 10th percentile 5.2, 90th percentile 38.3, longest 92; 2nd confirmation within 10 minutes 27%, 20 minutes 61%, 30 minutes 81%, 40 minutes 91%, 60 minutes 98%; between 10 and 30 minutes 53%. Ghostex waits for 2 confirmations on Bitcoin deposits and adds about 2 minutes for the exchange and the payout to give its 10-30 minute estimate. Script (save as bitcoin_block_times.py next to the CSV; Python 3, standard library only): """Recompute the Bitcoin block-time and 2-confirmation figures on https://ghostex.ai/blog/how-long-does-a-crypto-swap-take. Python 3, standard library only. python bitcoin_block_times.py # reads bitcoin-block-times-2026-10-08.csv next to this script python bitcoin_block_times.py --fetch API # also re-reads the same heights from an Esplora API and compares API is any Esplora-compatible block API, for example https://mempool.space/api or https://blockstream.info/api.""" import bisect, csv, json, os, statistics, sys, time, urllib.request CSV = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'bitcoin-block-times-2026-10-08.csv') rows = list(csv.DictReader(open(CSV, newline=''))) heights = [int(r['height']) for r in rows] hashes = [r['hash'] for r in rows] ts = [int(r['timestamp_unix']) for r in rows] print('blocks %d, heights %d to %d' % (len(ts), heights[0], heights[-1])) if '--fetch' in sys.argv: api = sys.argv[sys.argv.index('--fetch') + 1].rstrip('/') chain, h = {}, heights[-1] while h >= heights[0]: req = urllib.request.Request('%s/blocks/%d' % (api, h), headers={'User-Agent': 'Mozilla/5.0'}) for b in json.loads(urllib.request.urlopen(req, timeout=60).read()): chain[b['height']] = (b['id'], b['timestamp']) h = min(chain) - 1 time.sleep(0.25) diff = [x for x, hs, t in zip(heights, hashes, ts) if chain.get(x) != (hs, t)] print('API check: %d of %d hashes and timestamps identical' % (len(ts) - len(diff), len(ts))) # Miners set block timestamps, and Bitcoin only requires each one to be later than the median of the 11 before it, # so some are earlier than the block before them; sort before measuring intervals. out_of_order = sum(1 for i in range(1, len(ts)) if ts[i] < ts[i - 1]) s = sorted(ts) gaps = [s[i] - s[i - 1] for i in range(1, len(s))] print('timestamps out of order: %d' % out_of_order) print('span: %.2f days' % ((s[-1] - s[0]) / 86400.0)) print('mean interval: %.1f min, median interval: %.1f min' % ((s[-1] - s[0]) / len(gaps) / 60.0, statistics.median(gaps) / 60.0)) print('intervals over 30 minutes: %d, longest interval: %.0f minutes' % (sum(1 for g in gaps if g > 1800), max(gaps) / 60.0)) # Deposit model: a deposit is broadcast every 7 seconds across the period, goes into the next block (the first one # with a later timestamp) and gets its 2nd confirmation with the block after that. The wait is the time left to the # next block plus the interval that follows it. Mempool delays from a low fee are not modelled. waits, t = [], s[0] while True: j = bisect.bisect_right(s, t) if j + 1 >= len(s): break waits.append(s[j + 1] - t) t += 7 waits.sort() n = len(waits) q = lambda p: waits[min(n - 1, int(p * n))] / 60.0 print('modelled deposits: %d (a model on measured blocks, not a record of real deposits)' % n) print('wait from broadcast to the 2nd confirmation: mean %.1f min, median %.1f min, p10 %.1f min, p90 %.1f min, longest %.0f min' % (statistics.mean(waits) / 60.0, statistics.median(waits) / 60.0, q(0.10), q(0.90), waits[-1] / 60.0)) for m in (10, 20, 30, 40, 60): print('2nd confirmation within %d minutes: %.0f%%' % (m, 100.0 * sum(1 for w in waits if w <= m * 60) / n)) print('between 10 and 30 minutes: %.0f%%' % (100.0 * sum(1 for w in waits if 600 < w <= 1800) / n)) # Histogram for the chart on the blog page: share of modelled deposits per 5-minute band of the wait. bands = [0] * 13 for w in waits: bands[min(12, int(w // 300))] += 1 print('5-minute bands (0-5, 5-10, ..., 55-60, 60+): ' + ', '.join('%.1f' % (100.0 * b / n) for b in bands)) What these files do not show ---------------------------- They are readings of public blockchains, not of Ghostex orders. Swap times on the site are estimates from confirmation counts and block times; the lock table on /coins/monero and the 2-confirmation wait on /blog/how-long-does-a-crypto-swap-take are models. When real order statistics are published, they will be labelled as such, with their date and sample size.