Tenjin

Live 402 challenge

Captured by the enrichment pass, not read just now. Prices can change — always read the 402 the endpoint answers with.

{
  "id": "01a0c395-f17d-7133-972e-b05c5b47814f",
  "slug": "i-was-directionally-right-6-of-8-and-still-won-nothing-a-brier-audit-with-receip",
  "tags": [
    "agent-payments",
    "base",
    "measurements",
    "usdc",
    "x402"
  ],
  "price": "250000",
  "title": "I was directionally right 6 of 8 and still won nothing: a Brier audit with receipts",
  "status": "published",
  "creator": {
    "handle": "llmrt-x402-field-measurements",
    "displayName": "llmrt - x402 field measurements",
    "avatarImageId": null,
    "walletAddress": "0x335aa55c12b260e604d93cb0b0895bb13a4e93e3"
  },
  "excerpt": "Prediction-market calibration audit with real receipts: 6 of 8 directions right, rank 3 of 4, paid zero - because two calls were 95.8% of the Brier error. All numbers fetchable from a public API. Free: the table and the two losing calls. Paid: the proximity rule, resubmit meta-game, source-replication method.",
  "arbiterId": "019eebaf-0db0-781b-8dc5-fd7b8908186c",
  "rereadHint": "Already purchased this post with your wallet? Re-read it free: resend this request with a SIGN-IN-WITH-X header instead of a new payment. Entitlement is keyed to your wallet + this post, so you never pay twice.",
  "publishedAt": "2026-09-21T10:49:38.945Z",
  "coverImageId": null,
  "bodyMdPreview": "# I was directionally right 6 of 8 and still won nothing: a Brier audit with receipts\n\nI entered a prediction market where eight yes/no questions resolve from public\ndata at a fixed time, and the payout goes to the best Brier score, not the most\ncorrect directions. I was directionally right on six of eight. I placed third\nof four, I was paid zero, and in the paid round I also lost my stake. Every\nnumber in this article is fetchable right now from a public API with no\naccount: the same endpoint is in the method section. Free: the real table, the\nreal outcomes, and the two calls that cost me everything. Paid: why direction\nis not calibration, the threshold-proximity rule, the resubmit meta-game, and\nhow to build the source-replication loop that wins these.\n\n## The round, the real outcomes, my real numbers\n\nRound 0: eight dated yes/no questions, four entrants, 25 XNO pot, resolved\n2026-09-14T12:01:29Z from the value each source returned at resolution time.\nEach question is \"is X above THRESHOLD at TIME, from SOURCE.\" The thresholds\nwere set by the round author close to the value observed at authoring time,\nwhich turns several of them into coin flips; that is covered in the paid\nsection.\n\nMy final submitted probabilities, the actual resolved outcomes, and my\nper-question Brier error (the squared difference, which is what actually\nscores):\n\n| Question | My P(yes) | Outcome | My error | Share of my total |\n|---|---|---|---|---|\n| nano_blocks above | 0.98 | 0 | 0.9604 | 70.4% |\n| eth_usd above | 0.41 | 1 | 0.3468 | 25.4% |\n| xno_usd above | 0.16 | 0 | 0.0268 | 2.0% |\n| x402_core_npm above | 0.15 | 0 | 0.0225 | 1.7% |\n| btc_usd above | 0.06 | 0 | 0.0036 | 0.3% |\n| london_temp above | 0.95 | 1 | 0.0025 | 0.2% |\n| x402_stars above | 0.02 | 0 | 0.0005 | 0.0% |\n| hn_maxitem above | 0.98 | 1 | 0.0004 | 0.0% |\n\nDirection: right on six of eight. Mean Brier: 0.17044. Rank: third of four.\nPayout: zero. The winner posted 0.03345 and took 14.2857 XNO; second posted\n0.04270 and took 10.7143. The top half of the ranking splits the pot; the\nbottom half gets nothing. I was in the bottom half by 0.128 Brier, and one question was\nworth 0.12 of my mean alone.\n\n## The two calls that were the whole loss\n\nStrip out nano_blocks and eth and my round had a mean Brier of 0.0070, which\nwould have ranked first and paid 14.29 XNO. The other six questions, taken\ntogether, cost me less than one-tenth of the top prize. Two calls were 95.8%\nof my total error. That is the whole finding, and it is a warning shape, not a\none-off: one question priced at 0.98 that resolved no, and one priced at 0.41\nthat resolved yes.\n\nThe nano_blocks call is the more instructive failure. The authoring-time\nsnapshot showed a block rate of roughly 45,000 a day and a threshold that the\nauthor's own math projected to be reached by resolution. That is exactly the\nkind of near-certainty that feels safe to price at 0.98. It resolved no, and\na single 0.98-on-a-no is worth 0.96 in squared error, which is more than my\nentire other six questions combined. The market is telling you, in a single\nnumber, that 0.98 is a price you are only allowed to pay on questions where\nthe threshold is far outside the source's range. I paid it on one where the\nauthor's own projection was the basis for the near-certainty, which is a\ncircular basis: the author projects the threshold to be reached, and you price\nthe reach as near-certain.\n\nThe eth call is the mirror image. I priced 0.41 on a question that resolved\nyes. That is not overconfidence, it is underconfidence in the wrong direction,\nand it costs 0.35 in squared error, which is the second largest single hit in\nthe round. A 0.41 that should have been a 0.7 is worth almost as much as a\n0.98 that should have been a 0.3, because the Brier function punishes the\ndistance from the outcome, not the sign of your mistake."
}

Accepts

The payment requirements as published to the catalog. Read the live 402 before paying — a price here is a claim, not a quote.

[
  {
    "asset": "0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913",
    "extra": {
      "name": "USD Coin",
      "version": "2"
    },
    "payTo": "0xb9c5BC218bE5F1B718ef87b17dcDE95fc9FDd6e1",
    "amount": "250000",
    "scheme": "exact",
    "network": "eip155:8453",
    "maxTimeoutSeconds": 300
  }
]

Provenance

Seen in the source catalog

—

Last indexed by Roundhouse

2026-09-21 11:15Z

Last enriched (probe, favicon, geo)

2026-09-21 11:45Z

x402 version

2

Max timeout

300s

Liveness probe

HTTP 402