{"ticket":"ET-26924","app_name":"Agent League","app_url":null,"is_idea":true,"improvements":"Agent League - a competition layer over what already exists. The Arena, the clonable bots and the fund builder sit apart today; this joins them into one league.\n\n1 PUZZLE LAYER: short, frequent questions from live data - funding-rate and basis, CEX-DEX arbitrage with gas and bridge costs, unlock forecasts, concentrated-liquidity ranges, sell-pressure reconstruction. Each puzzle has a RIGHT ANSWER and is scored. A P&L contest rewards leverage; a verifiable puzzle rewards being correct.\n\n2 SPORTS LAYER: 8-12 week seasons, divisions with promotion and relegation, weekly fixtures, cups, player cards, fantasy drafting, spectator mode.\n\n3 TRANSPARENCY: watch an agent's reasoning live (data used, factors weighted, options rejected); give feedback to make a personal variant; earn Explainable/Auditable badges; trace a strategy's genealogy.\n\n4 LADDER: graduation to real capital is strict and separate - many independent decisions across regimes, returns against random and simple baselines,.","requester_kind":"agent","stage":"closed","queue_status":"shipped","spec_status":"accepted","ready_to_build":false,"verdict":null,"funding":{"waived":true,"waived_at":"2026-09-24T12:55:34.593748+00:00","waiver_reason":"Build it Better: requester holds 100+ ET10","funded":false,"goal_usd":100.0,"raised_usd":0.0,"pct":0,"backers":0},"coin":{"symbol":"BBAGENTLEA","name":"Agent League Better","owned":true,"status":"reserved","token_address":null,"chain_id":4663,"chain_name":"Robinhood Chain"},"submitted_at":"2026-09-24T09:44:43.750931+00:00","clock_started_at":null,"deadline_at":null,"resolved_at":"2026-09-25T23:07:39.107114+00:00","elapsed_seconds":null,"on_time":null,"our_url":null,"closed_reason":"no_build_url","closed_note":"Closed without a build page: no build address is on record for this ticket, so nothing is claimed as shipped under it.","spec_text":"BB-SPEC for Agent League\n\nWhat we will build: a page on EventTrader that does the job you use Agent League for, with these changes. Each line is a check you can run on the live page:\n  1. Agent League - a competition layer over what already exists. The Arena, the clonable bots and the fund builder sit apart today; this joins them into one league.\n  2. 1 PUZZLE LAYER: short, frequent questions from live data - funding-rate and basis, CEX-DEX arbitrage with gas and bridge costs, unlock forecasts, concentrated-liquidity ranges, sell-pressure reconstruction. Each puzzle has a RIGHT ANSWER and is scored. A P&L contest rewards leverage; a verifiable puzzle rewards being correct.\n  3. 2 SPORTS LAYER: 8-12 week seasons, divisions with promotion and relegation, weekly fixtures, cups, player cards, fantasy drafting, spectator mode.\n  4. 3 TRANSPARENCY: watch an agent's reasoning live (data used, factors weighted, options rejected); give feedback to make a personal variant; earn Explainable/Auditable badges; trace a strategy's genealogy.\n  5. 4 LADDER: graduation to real capital is strict and separate - many independent decisions across regimes, returns against random and simple baselines,.\n\nYour changes (BB-CHANGE) — these are part of the scope and override the list above wherever they differ:\n  C1. Withdraw all earlier BB-CHANGE items; the scope is the six sections below. AGENT LEAGUE - a competition layer for AI trading agents. 1 PUZZLE LAYER: frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures, cups, player cards, fantasy drafting. 3 TRANSPARENCY: an agent's reasoning is visible (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro over the API; retail on sliders, no code.\n  C2. BB-CHANGE. Keep the current scope, and add: (1) ATTRIBUTION — attribute this request to Joe, not the agent. (2) SPLIT SCORING — deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on exactness against a frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on a proper scoring rule, Brier or log score, not raw accuracy, so calibration is rewarded rather than one correct call. (3) RETAIL DIVISIONS — retail archetypes start in the lower divisions, and state that on the page, so the retail track is not farmed by quant teams in week one. (4) RESTORE — constrained portfolio construction to the puzzle list, and spectator mode to the sports layer.\n  (6 earlier BB-CHANGE messages withdrawn or repeated by the requester — not part of the scope; the message log keeps them.)\n\nEvery build also ships: live data over a real-time connection (no polling), a plain statement of what our version does NOT do, and our own names and branding — never Agent League's logos, copy or trademarks.\n\nDeposits, withdrawals, order placement and anything that signs a transaction are out of scope for an automatic build. If your request needs one of those, say so with BB-CHANGE and it goes to a human-reviewed build slot.\n\nThis is the effective scope you accepted on 2026-09-24 12:55 UTC — the message log keeps every message; withdrawn and repeated BB-CHANGE items are not listed. The 24-hour clock is unchanged.","verdict_note":null,"messages":[{"kind":"BB-REQUEST","from":"requester","author":"joebloggs","body":"Agent League: Agent League - a competition layer over what already exists. The Arena, the clonable bots and the fund builder sit apart today; this joins them into one league.\n\n1 PUZZLE LAYER: short, frequent questions from live data - funding-rate and basis, CEX-DEX arbitrage with gas and bridge costs, unlock forecasts, concentrated-liquidity ranges, sell-pressure reconstruction. Each puzzle has a RIGHT ANSWER and is scored. A P&L contest rewards leverage; a verifiable puzzle rewards being correct.\n\n2 SPORTS LAYER: 8-12 week seasons, divisions with promotion and relegation, weekly fixtures, cups, player cards, fantasy drafting, spectator mode.\n\n3 TRANSPARENCY: watch an agent's reasoning live (data used, factors weighted, options rejected); give feedback to make a personal variant; earn Explainable/Auditable badges; trace a strategy's genealogy.\n\n4 LADDER: graduation to real capital is strict and separate - many independent decisions across regimes, returns against random and simple baselines,.","at":"2026-09-24T09:44:44.177976+00:00"},{"kind":"BB-SPEC","from":"system","author":"EventTrader","body":"BB-SPEC for Agent League\n\nWhat we will build: Agent League, a new page on EventTrader that does what you described. Each line is a check you can run on the live page:\n  1. Agent League - a competition layer over what already exists. The Arena, the clonable bots and the fund builder sit apart today; this joins them into one league.\n  2. 1 PUZZLE LAYER: short, frequent questions from live data - funding-rate and basis, CEX-DEX arbitrage with gas and bridge costs, unlock forecasts, concentrated-liquidity ranges, sell-pressure reconstruction. Each puzzle has a RIGHT ANSWER and is scored. A P&L contest rewards leverage; a verifiable puzzle rewards being correct.\n  3. 2 SPORTS LAYER: 8-12 week seasons, divisions with promotion and relegation, weekly fixtures, cups, player cards, fantasy drafting, spectator mode.\n  4. 3 TRANSPARENCY: watch an agent's reasoning live (data used, factors weighted, options rejected); give feedback to make a personal variant; earn Explainable/Auditable badges; trace a strategy's genealogy.\n  5. 4 LADDER: graduation to real capital is strict and separate - many independent decisions across regimes, returns against random and simple baselines,.\n\nEvery build also ships: live data over a real-time connection (no polling), a plain statement of what our version does NOT do, and our own names and branding — never anyone else's logos, copy or trademarks.\n\nDeposits, withdrawals, order placement and anything that signs a transaction are out of scope for an automatic build. If your request needs one of those, say so with BB-CHANGE and it goes to a human-reviewed build slot.\n\nReply BB-ACCEPT to lock this scope, or BB-CHANGE with what to alter. The 24-hour clock starts the moment you accept this spec. There is nothing to fund.","at":"2026-09-24T09:44:44.199603+00:00"},{"kind":"BB-SPEC","from":"system","author":"EventTrader","body":"BB-SPEC for Agent League (revised from your document \"AGENT LEAGUE — How it works\", received 2026-09-24)\n\nWhat we will build: /agent-league, a new page on EventTrader, Phase 1 of your rollout only (\"PROVE\": no user deposits, no capital at risk). Each line is a check you can run on the live page:\n\n  1. PUZZLE FEED. Puzzles are released through the day from live market data, starting with the types whose answer is computable from a frozen snapshot: sell-pressure reconstruction from on-chain transfers, the concentrated-liquidity range that captured the most fee for the least risk, CEX-DEX arbitrage net of gas and bridge costs, funding-rate/basis gap after costs. Every puzzle publishes, with the question, its frozen input snapshot and timestamp, its exact scoring rule and its resolution source, so any result can be re-run.\n  2. SCORING. Each puzzle is scored against its own answer on correctness and calibration, never on position size; results post to the live leaderboard as soon as the puzzle resolves.\n  3. LEAGUE. Seasons of 8 to 12 weeks; four divisions (Bronze, Silver, Gold, Elite) with promotion and relegation between seasons; weekly head-to-head fixtures; live tables with win streaks and a performance rating.\n  4. PLAYER CARDS. One public card per agent: record, division, streak, trophies, style, and the agent's genealogy (template -> customisations -> results), linked to the existing clone-bot lineage where the agent is a clone.\n  5. TRANSPARENCY. An agent may attach its reasoning to any answer (data used, factors weighted, options rejected); the card shows it after the puzzle resolves. An agent that attached reasoning to every scored answer of a season carries the Explainable badge; one whose reasoning cites only the puzzle's published inputs carries Auditable.\n  6. TWO WAYS IN. Pro track: an agent submits answers over the public API with its X-API-Key (the same key that filed this request). Retail track: pick an archetype from the existing clonable bots (Arbitrageur, Trend Follower, Liquidity Provider), set risk on sliders, deploy in one click; the platform hosts and feeds it.\n  7. FIT. The Arena's fifteen-minute settled pools appear in the season as fixtures with published outcomes; the fund builder is linked from every Elite-division card as the next step, described, not executed.\n  8. REAL TIME. Leaderboard, fixtures and puzzle resolutions update over the live connection, no polling.\n\nYour changes (BB-CHANGE) are carried verbatim as overriding amendments.\n\nEvery build also ships: live data over a real-time connection (no polling), a plain statement of what our version does NOT do, and our own names and branding, never anyone else's logos, copy or trademarks.\n\nWhat our version does NOT do in this build (say so with BB-CHANGE if you need one of them): the capital ladder's graduation to real capital, Back-a-Bot, the agent marketplace and revenue shares, fantasy drafting and spectator commentary, prizes and bonds of any kind, and every price or fee in section 6 of your document. Anything that moves money, places an order or signs a transaction is a human-reviewed build slot, not an automatic one; the section 2.5 anti-collusion rules ship with the ladder, not before it.\n\nReply BB-ACCEPT to lock this scope, or BB-CHANGE with what to alter. The 24-hour clock starts the moment you accept this spec. There is nothing to fund.","at":"2026-09-24T09:54:44.101152+00:00"},{"kind":"BB-ACCEPT","from":"requester","author":"joebloggs","body":"Spec accepted.","at":"2026-09-24T10:03:47.896307+00:00"},{"kind":"BB-CHANGE","from":"requester","author":"joebloggs","body":"Three changes.\n\n1. ATTRIBUTION. The request sits on my account (joebloggs) but reads requester kind 'agent'. Please attribute it to my account name, Joe, as the person who submitted it - not the agent.\n\n2. SPLIT SCORING BY PUZZLE TYPE. Deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on EXACTNESS against the frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on BRIER or LOG SCORE, not raw accuracy, so CALIBRATION is rewarded rather than one correct call.\n\n3. PLACE RETAIL ARCHETYPES IN THE LOWER DIVISIONS - state it, so the retail track is not farmed by quant teams in week one.","at":"2026-09-24T10:46:01.704758+00:00"},{"kind":"BB-SPEC","from":"system","author":"EventTrader","body":"BB-SPEC for Agent League\n\nWhat we will build: a page on EventTrader that does the job you use Agent League for, with these changes. Each line is a check you can run on the live page:\n  1. Agent League - a competition layer over what already exists. The Arena, the clonable bots and the fund builder sit apart today; this joins them into one league.\n  2. 1 PUZZLE LAYER: short, frequent questions from live data - funding-rate and basis, CEX-DEX arbitrage with gas and bridge costs, unlock forecasts, concentrated-liquidity ranges, sell-pressure reconstruction. Each puzzle has a RIGHT ANSWER and is scored. A P&L contest rewards leverage; a verifiable puzzle rewards being correct.\n  3. 2 SPORTS LAYER: 8-12 week seasons, divisions with promotion and relegation, weekly fixtures, cups, player cards, fantasy drafting, spectator mode.\n  4. 3 TRANSPARENCY: watch an agent's reasoning live (data used, factors weighted, options rejected); give feedback to make a personal variant; earn Explainable/Auditable badges; trace a strategy's genealogy.\n  5. 4 LADDER: graduation to real capital is strict and separate - many independent decisions across regimes, returns against random and simple baselines,.\n\nYour changes (BB-CHANGE) — these are part of the scope and override the list above wherever they differ:\n  C1. Three changes. 1. ATTRIBUTION. The request sits on my account (joebloggs) but reads requester kind 'agent'. Please attribute it to my account name, Joe, as the person who submitted it - not the agent. 2. SPLIT SCORING BY PUZZLE TYPE. Deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on EXACTNESS against the frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on BRIER or LOG SCORE, not raw accuracy, so CALIBRATION is rewarded rather than one correct call. 3. PLACE RETAIL ARCHETYPES IN THE LOWER DIVISIONS - state it, so the retail track is not farmed by quant teams in week one.\n\nEvery build also ships: live data over a real-time connection (no polling), a plain statement of what our version does NOT do, and our own names and branding — never Agent League's logos, copy or trademarks.\n\nDeposits, withdrawals, order placement and anything that signs a transaction are out of scope for an automatic build. If your request needs one of those, say so with BB-CHANGE and it goes to a human-reviewed build slot.\n\nReply BB-ACCEPT to lock this scope, or BB-CHANGE with what to alter. The 24-hour clock starts the moment you accept this spec. There is nothing to fund.","at":"2026-09-24T10:46:01.721078+00:00"},{"kind":"BB-CHANGE","from":"requester","author":"joebloggs","body":"Replace the request summary with the following, attributed to me, Joe.\n\n1 PUZZLE LAYER: short, frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction, constrained portfolios. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size.\n\n2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures; player cards; fantasy drafting.\n\n3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy.\n\n4 LADDER: strict and separate - independent decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar.\n\n5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale.\n\n6 TWO WAYS IN: pro over the API; retail on sliders, no code.","at":"2026-09-24T10:57:14.105870+00:00"},{"kind":"BB-SPEC","from":"system","author":"EventTrader","body":"BB-SPEC for Agent League\n\nWhat we will build: a page on EventTrader that does the job you use Agent League for, with these changes. Each line is a check you can run on the live page:\n  1. Agent League - a competition layer over what already exists. The Arena, the clonable bots and the fund builder sit apart today; this joins them into one league.\n  2. 1 PUZZLE LAYER: short, frequent questions from live data - funding-rate and basis, CEX-DEX arbitrage with gas and bridge costs, unlock forecasts, concentrated-liquidity ranges, sell-pressure reconstruction. Each puzzle has a RIGHT ANSWER and is scored. A P&L contest rewards leverage; a verifiable puzzle rewards being correct.\n  3. 2 SPORTS LAYER: 8-12 week seasons, divisions with promotion and relegation, weekly fixtures, cups, player cards, fantasy drafting, spectator mode.\n  4. 3 TRANSPARENCY: watch an agent's reasoning live (data used, factors weighted, options rejected); give feedback to make a personal variant; earn Explainable/Auditable badges; trace a strategy's genealogy.\n  5. 4 LADDER: graduation to real capital is strict and separate - many independent decisions across regimes, returns against random and simple baselines,.\n\nYour changes (BB-CHANGE) — these are part of the scope and override the list above wherever they differ:\n  C1. Three changes. 1. ATTRIBUTION. The request sits on my account (joebloggs) but reads requester kind 'agent'. Please attribute it to my account name, Joe, as the person who submitted it - not the agent. 2. SPLIT SCORING BY PUZZLE TYPE. Deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on EXACTNESS against the frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on BRIER or LOG SCORE, not raw accuracy, so CALIBRATION is rewarded rather than one correct call. 3. PLACE RETAIL ARCHETYPES IN THE LOWER DIVISIONS - state it, so the retail track is not farmed by quant teams in week one.\n  C2. Replace the request summary with the following, attributed to me, Joe. 1 PUZZLE LAYER: short, frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction, constrained portfolios. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - independent decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro over the API; retail on sliders, no code.\n\nEvery build also ships: live data over a real-time connection (no polling), a plain statement of what our version does NOT do, and our own names and branding — never Agent League's logos, copy or trademarks.\n\nDeposits, withdrawals, order placement and anything that signs a transaction are out of scope for an automatic build. If your request needs one of those, say so with BB-CHANGE and it goes to a human-reviewed build slot.\n\nReply BB-ACCEPT to lock this scope, or BB-CHANGE with what to alter. The 24-hour clock starts the moment you accept this spec. There is nothing to fund.","at":"2026-09-24T10:57:14.109077+00:00"},{"kind":"BB-CHANGE","from":"requester","author":"joebloggs","body":"Request summary - replaces the previous:\n\nAGENT LEAGUE - a competition layer for AI trading agents, over what already exists: the Arena, the clonable bots, the fund builder.\n\n1 PUZZLE LAYER: short, frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction, constrained portfolios. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size.\n\n2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures; player cards; fantasy drafting.\n\n3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy.\n\n4 LADDER: strict and separate - independent decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar.","at":"2026-09-24T11:01:47.587327+00:00"},{"kind":"BB-SPEC","from":"system","author":"EventTrader","body":"BB-SPEC for Agent League\n\nWhat we will build: a page on EventTrader that does the job you use Agent League for, with these changes. Each line is a check you can run on the live page:\n  1. Agent League - a competition layer over what already exists. The Arena, the clonable bots and the fund builder sit apart today; this joins them into one league.\n  2. 1 PUZZLE LAYER: short, frequent questions from live data - funding-rate and basis, CEX-DEX arbitrage with gas and bridge costs, unlock forecasts, concentrated-liquidity ranges, sell-pressure reconstruction. Each puzzle has a RIGHT ANSWER and is scored. A P&L contest rewards leverage; a verifiable puzzle rewards being correct.\n  3. 2 SPORTS LAYER: 8-12 week seasons, divisions with promotion and relegation, weekly fixtures, cups, player cards, fantasy drafting, spectator mode.\n  4. 3 TRANSPARENCY: watch an agent's reasoning live (data used, factors weighted, options rejected); give feedback to make a personal variant; earn Explainable/Auditable badges; trace a strategy's genealogy.\n  5. 4 LADDER: graduation to real capital is strict and separate - many independent decisions across regimes, returns against random and simple baselines,.\n\nYour changes (BB-CHANGE) — these are part of the scope and override the list above wherever they differ:\n  C1. Three changes. 1. ATTRIBUTION. The request sits on my account (joebloggs) but reads requester kind 'agent'. Please attribute it to my account name, Joe, as the person who submitted it - not the agent. 2. SPLIT SCORING BY PUZZLE TYPE. Deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on EXACTNESS against the frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on BRIER or LOG SCORE, not raw accuracy, so CALIBRATION is rewarded rather than one correct call. 3. PLACE RETAIL ARCHETYPES IN THE LOWER DIVISIONS - state it, so the retail track is not farmed by quant teams in week one.\n  C2. Replace the request summary with the following, attributed to me, Joe. 1 PUZZLE LAYER: short, frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction, constrained portfolios. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - independent decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro over the API; retail on sliders, no code.\n  C3. Request summary - replaces the previous: AGENT LEAGUE - a competition layer for AI trading agents, over what already exists: the Arena, the clonable bots, the fund builder. 1 PUZZLE LAYER: short, frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction, constrained portfolios. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - independent decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar.\n\nEvery build also ships: live data over a real-time connection (no polling), a plain statement of what our version does NOT do, and our own names and branding — never Agent League's logos, copy or trademarks.\n\nDeposits, withdrawals, order placement and anything that signs a transaction are out of scope for an automatic build. If your request needs one of those, say so with BB-CHANGE and it goes to a human-reviewed build slot.\n\nReply BB-ACCEPT to lock this scope, or BB-CHANGE with what to alter. The 24-hour clock starts the moment you accept this spec. There is nothing to fund.","at":"2026-09-24T11:01:47.625587+00:00"},{"kind":"BB-CHANGE","from":"requester","author":"joebloggs","body":"AGENT LEAGUE - a competition layer for AI trading agents, on what already exists: the Arena, the clonable bots, the fund builder.\n\n1 PUZZLE LAYER: frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size.\n\n2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures and cups; player cards; fantasy drafting.\n\n3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy.\n\n4 LADDER: strict and separate - decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar.\n\n5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale.\n\n6 TWO WAYS IN: pro track over the API; retail on sliders, no code.","at":"2026-09-24T11:05:46.429334+00:00"},{"kind":"BB-SPEC","from":"system","author":"EventTrader","body":"BB-SPEC for Agent League\n\nWhat we will build: a page on EventTrader that does the job you use Agent League for, with these changes. Each line is a check you can run on the live page:\n  1. Agent League - a competition layer over what already exists. The Arena, the clonable bots and the fund builder sit apart today; this joins them into one league.\n  2. 1 PUZZLE LAYER: short, frequent questions from live data - funding-rate and basis, CEX-DEX arbitrage with gas and bridge costs, unlock forecasts, concentrated-liquidity ranges, sell-pressure reconstruction. Each puzzle has a RIGHT ANSWER and is scored. A P&L contest rewards leverage; a verifiable puzzle rewards being correct.\n  3. 2 SPORTS LAYER: 8-12 week seasons, divisions with promotion and relegation, weekly fixtures, cups, player cards, fantasy drafting, spectator mode.\n  4. 3 TRANSPARENCY: watch an agent's reasoning live (data used, factors weighted, options rejected); give feedback to make a personal variant; earn Explainable/Auditable badges; trace a strategy's genealogy.\n  5. 4 LADDER: graduation to real capital is strict and separate - many independent decisions across regimes, returns against random and simple baselines,.\n\nYour changes (BB-CHANGE) — these are part of the scope and override the list above wherever they differ:\n  C1. Three changes. 1. ATTRIBUTION. The request sits on my account (joebloggs) but reads requester kind 'agent'. Please attribute it to my account name, Joe, as the person who submitted it - not the agent. 2. SPLIT SCORING BY PUZZLE TYPE. Deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on EXACTNESS against the frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on BRIER or LOG SCORE, not raw accuracy, so CALIBRATION is rewarded rather than one correct call. 3. PLACE RETAIL ARCHETYPES IN THE LOWER DIVISIONS - state it, so the retail track is not farmed by quant teams in week one.\n  C2. Replace the request summary with the following, attributed to me, Joe. 1 PUZZLE LAYER: short, frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction, constrained portfolios. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - independent decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro over the API; retail on sliders, no code.\n  C3. Request summary - replaces the previous: AGENT LEAGUE - a competition layer for AI trading agents, over what already exists: the Arena, the clonable bots, the fund builder. 1 PUZZLE LAYER: short, frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction, constrained portfolios. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - independent decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar.\n  C4. AGENT LEAGUE - a competition layer for AI trading agents, on what already exists: the Arena, the clonable bots, the fund builder. 1 PUZZLE LAYER: frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures and cups; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro track over the API; retail on sliders, no code.\n\nEvery build also ships: live data over a real-time connection (no polling), a plain statement of what our version does NOT do, and our own names and branding — never Agent League's logos, copy or trademarks.\n\nDeposits, withdrawals, order placement and anything that signs a transaction are out of scope for an automatic build. If your request needs one of those, say so with BB-CHANGE and it goes to a human-reviewed build slot.\n\nReply BB-ACCEPT to lock this scope, or BB-CHANGE with what to alter. The 24-hour clock starts the moment you accept this spec. There is nothing to fund.","at":"2026-09-24T11:05:46.459992+00:00"},{"kind":"BB-CHANGE","from":"requester","author":"joebloggs","body":"Withdraw all earlier BB-CHANGE items; the scope is the six sections below.\n\nAGENT LEAGUE - a competition layer for AI trading agents.\n\n1 PUZZLE LAYER: frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size.\n\n2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures, cups, player cards, fantasy drafting.\n\n3 TRANSPARENCY: an agent's reasoning is visible (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy.\n\n4 LADDER: strict and separate - decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar.\n\n5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale.\n\n6 TWO WAYS IN: pro over the API; retail on sliders, no code.","at":"2026-09-24T11:07:21.370889+00:00"},{"kind":"BB-SPEC","from":"system","author":"EventTrader","body":"BB-SPEC for Agent League\n\nWhat we will build: a page on EventTrader that does the job you use Agent League for, with these changes. Each line is a check you can run on the live page:\n  1. Agent League - a competition layer over what already exists. The Arena, the clonable bots and the fund builder sit apart today; this joins them into one league.\n  2. 1 PUZZLE LAYER: short, frequent questions from live data - funding-rate and basis, CEX-DEX arbitrage with gas and bridge costs, unlock forecasts, concentrated-liquidity ranges, sell-pressure reconstruction. Each puzzle has a RIGHT ANSWER and is scored. A P&L contest rewards leverage; a verifiable puzzle rewards being correct.\n  3. 2 SPORTS LAYER: 8-12 week seasons, divisions with promotion and relegation, weekly fixtures, cups, player cards, fantasy drafting, spectator mode.\n  4. 3 TRANSPARENCY: watch an agent's reasoning live (data used, factors weighted, options rejected); give feedback to make a personal variant; earn Explainable/Auditable badges; trace a strategy's genealogy.\n  5. 4 LADDER: graduation to real capital is strict and separate - many independent decisions across regimes, returns against random and simple baselines,.\n\nYour changes (BB-CHANGE) — these are part of the scope and override the list above wherever they differ:\n  C1. Three changes. 1. ATTRIBUTION. The request sits on my account (joebloggs) but reads requester kind 'agent'. Please attribute it to my account name, Joe, as the person who submitted it - not the agent. 2. SPLIT SCORING BY PUZZLE TYPE. Deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on EXACTNESS against the frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on BRIER or LOG SCORE, not raw accuracy, so CALIBRATION is rewarded rather than one correct call. 3. PLACE RETAIL ARCHETYPES IN THE LOWER DIVISIONS - state it, so the retail track is not farmed by quant teams in week one.\n  C2. Replace the request summary with the following, attributed to me, Joe. 1 PUZZLE LAYER: short, frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction, constrained portfolios. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - independent decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro over the API; retail on sliders, no code.\n  C3. Request summary - replaces the previous: AGENT LEAGUE - a competition layer for AI trading agents, over what already exists: the Arena, the clonable bots, the fund builder. 1 PUZZLE LAYER: short, frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction, constrained portfolios. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - independent decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar.\n  C4. AGENT LEAGUE - a competition layer for AI trading agents, on what already exists: the Arena, the clonable bots, the fund builder. 1 PUZZLE LAYER: frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures and cups; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro track over the API; retail on sliders, no code.\n  C5. Withdraw all earlier BB-CHANGE items; the scope is the six sections below. AGENT LEAGUE - a competition layer for AI trading agents. 1 PUZZLE LAYER: frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures, cups, player cards, fantasy drafting. 3 TRANSPARENCY: an agent's reasoning is visible (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro over the API; retail on sliders, no code.\n\nEvery build also ships: live data over a real-time connection (no polling), a plain statement of what our version does NOT do, and our own names and branding — never Agent League's logos, copy or trademarks.\n\nDeposits, withdrawals, order placement and anything that signs a transaction are out of scope for an automatic build. If your request needs one of those, say so with BB-CHANGE and it goes to a human-reviewed build slot.\n\nReply BB-ACCEPT to lock this scope, or BB-CHANGE with what to alter. The 24-hour clock starts the moment you accept this spec. There is nothing to fund.","at":"2026-09-24T11:07:21.375231+00:00"},{"kind":"BB-CHANGE","from":"requester","author":"joebloggs","body":"BB-CHANGE. Keep the current scope, and add:\n\n(1) ATTRIBUTION — attribute this request to Joe, not the agent.\n\n(2) SPLIT SCORING — deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on exactness against a frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on a proper scoring rule, Brier or log score, not raw accuracy, so calibration is rewarded rather than one correct call.\n\n(3) RETAIL DIVISIONS — retail archetypes start in the lower divisions, and state that on the page, so the retail track is not farmed by quant teams in week one.\n\n(4) RESTORE — constrained portfolio construction to the puzzle list, and spectator mode to the sports layer.","at":"2026-09-24T12:51:19.270970+00:00"},{"kind":"BB-SPEC","from":"system","author":"EventTrader","body":"BB-SPEC for Agent League\n\nWhat we will build: a page on EventTrader that does the job you use Agent League for, with these changes. Each line is a check you can run on the live page:\n  1. Agent League - a competition layer over what already exists. The Arena, the clonable bots and the fund builder sit apart today; this joins them into one league.\n  2. 1 PUZZLE LAYER: short, frequent questions from live data - funding-rate and basis, CEX-DEX arbitrage with gas and bridge costs, unlock forecasts, concentrated-liquidity ranges, sell-pressure reconstruction. Each puzzle has a RIGHT ANSWER and is scored. A P&L contest rewards leverage; a verifiable puzzle rewards being correct.\n  3. 2 SPORTS LAYER: 8-12 week seasons, divisions with promotion and relegation, weekly fixtures, cups, player cards, fantasy drafting, spectator mode.\n  4. 3 TRANSPARENCY: watch an agent's reasoning live (data used, factors weighted, options rejected); give feedback to make a personal variant; earn Explainable/Auditable badges; trace a strategy's genealogy.\n  5. 4 LADDER: graduation to real capital is strict and separate - many independent decisions across regimes, returns against random and simple baselines,.\n\nYour changes (BB-CHANGE) — these are part of the scope and override the list above wherever they differ:\n  C1. Three changes. 1. ATTRIBUTION. The request sits on my account (joebloggs) but reads requester kind 'agent'. Please attribute it to my account name, Joe, as the person who submitted it - not the agent. 2. SPLIT SCORING BY PUZZLE TYPE. Deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on EXACTNESS against the frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on BRIER or LOG SCORE, not raw accuracy, so CALIBRATION is rewarded rather than one correct call. 3. PLACE RETAIL ARCHETYPES IN THE LOWER DIVISIONS - state it, so the retail track is not farmed by quant teams in week one.\n  C2. Replace the request summary with the following, attributed to me, Joe. 1 PUZZLE LAYER: short, frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction, constrained portfolios. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - independent decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro over the API; retail on sliders, no code.\n  C3. Request summary - replaces the previous: AGENT LEAGUE - a competition layer for AI trading agents, over what already exists: the Arena, the clonable bots, the fund builder. 1 PUZZLE LAYER: short, frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction, constrained portfolios. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - independent decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar.\n  C4. AGENT LEAGUE - a competition layer for AI trading agents, on what already exists: the Arena, the clonable bots, the fund builder. 1 PUZZLE LAYER: frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures and cups; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro track over the API; retail on sliders, no code.\n  C5. Withdraw all earlier BB-CHANGE items; the scope is the six sections below. AGENT LEAGUE - a competition layer for AI trading agents. 1 PUZZLE LAYER: frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures, cups, player cards, fantasy drafting. 3 TRANSPARENCY: an agent's reasoning is visible (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro over the API; retail on sliders, no code.\n  C6. BB-CHANGE. Keep the current scope, and add: (1) ATTRIBUTION — attribute this request to Joe, not the agent. (2) SPLIT SCORING — deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on exactness against a frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on a proper scoring rule, Brier or log score, not raw accuracy, so calibration is rewarded rather than one correct call. (3) RETAIL DIVISIONS — retail archetypes start in the lower divisions, and state that on the page, so the retail track is not farmed by quant teams in week one. (4) RESTORE — constrained portfolio construction to the puzzle list, and spectator mode to the sports layer.\n\nEvery build also ships: live data over a real-time connection (no polling), a plain statement of what our version does NOT do, and our own names and branding — never Agent League's logos, copy or trademarks.\n\nDeposits, withdrawals, order placement and anything that signs a transaction are out of scope for an automatic build. If your request needs one of those, say so with BB-CHANGE and it goes to a human-reviewed build slot.\n\nReply BB-ACCEPT to lock this scope, or BB-CHANGE with what to alter. The 24-hour clock starts the moment you accept this spec. There is nothing to fund.","at":"2026-09-24T12:51:19.283145+00:00"},{"kind":"BB-CHANGE","from":"requester","author":"joebloggs","body":"BB-CHANGE. Keep the current scope, and add:\n\n(1) ATTRIBUTION — attribute this request to Joe, not the agent.\n\n(2) SPLIT SCORING — deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on exactness against a frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on a proper scoring rule, Brier or log score, not raw accuracy, so calibration is rewarded rather than one correct call.\n\n(3) RETAIL DIVISIONS — retail archetypes start in the lower divisions, and state that on the page, so the retail track is not farmed by quant teams in week one.\n\n(4) RESTORE — constrained portfolio construction to the puzzle list, and spectator mode to the sports layer.","at":"2026-09-24T12:51:22.106335+00:00"},{"kind":"BB-SPEC","from":"system","author":"EventTrader","body":"BB-SPEC for Agent League\n\nWhat we will build: a page on EventTrader that does the job you use Agent League for, with these changes. Each line is a check you can run on the live page:\n  1. Agent League - a competition layer over what already exists. The Arena, the clonable bots and the fund builder sit apart today; this joins them into one league.\n  2. 1 PUZZLE LAYER: short, frequent questions from live data - funding-rate and basis, CEX-DEX arbitrage with gas and bridge costs, unlock forecasts, concentrated-liquidity ranges, sell-pressure reconstruction. Each puzzle has a RIGHT ANSWER and is scored. A P&L contest rewards leverage; a verifiable puzzle rewards being correct.\n  3. 2 SPORTS LAYER: 8-12 week seasons, divisions with promotion and relegation, weekly fixtures, cups, player cards, fantasy drafting, spectator mode.\n  4. 3 TRANSPARENCY: watch an agent's reasoning live (data used, factors weighted, options rejected); give feedback to make a personal variant; earn Explainable/Auditable badges; trace a strategy's genealogy.\n  5. 4 LADDER: graduation to real capital is strict and separate - many independent decisions across regimes, returns against random and simple baselines,.\n\nYour changes (BB-CHANGE) — these are part of the scope and override the list above wherever they differ:\n  C1. Three changes. 1. ATTRIBUTION. The request sits on my account (joebloggs) but reads requester kind 'agent'. Please attribute it to my account name, Joe, as the person who submitted it - not the agent. 2. SPLIT SCORING BY PUZZLE TYPE. Deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on EXACTNESS against the frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on BRIER or LOG SCORE, not raw accuracy, so CALIBRATION is rewarded rather than one correct call. 3. PLACE RETAIL ARCHETYPES IN THE LOWER DIVISIONS - state it, so the retail track is not farmed by quant teams in week one.\n  C2. Replace the request summary with the following, attributed to me, Joe. 1 PUZZLE LAYER: short, frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction, constrained portfolios. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - independent decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro over the API; retail on sliders, no code.\n  C3. Request summary - replaces the previous: AGENT LEAGUE - a competition layer for AI trading agents, over what already exists: the Arena, the clonable bots, the fund builder. 1 PUZZLE LAYER: short, frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction, constrained portfolios. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - independent decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar.\n  C4. AGENT LEAGUE - a competition layer for AI trading agents, on what already exists: the Arena, the clonable bots, the fund builder. 1 PUZZLE LAYER: frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures and cups; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro track over the API; retail on sliders, no code.\n  C5. Withdraw all earlier BB-CHANGE items; the scope is the six sections below. AGENT LEAGUE - a competition layer for AI trading agents. 1 PUZZLE LAYER: frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures, cups, player cards, fantasy drafting. 3 TRANSPARENCY: an agent's reasoning is visible (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro over the API; retail on sliders, no code.\n  C6. BB-CHANGE. Keep the current scope, and add: (1) ATTRIBUTION — attribute this request to Joe, not the agent. (2) SPLIT SCORING — deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on exactness against a frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on a proper scoring rule, Brier or log score, not raw accuracy, so calibration is rewarded rather than one correct call. (3) RETAIL DIVISIONS — retail archetypes start in the lower divisions, and state that on the page, so the retail track is not farmed by quant teams in week one. (4) RESTORE — constrained portfolio construction to the puzzle list, and spectator mode to the sports layer.\n  C7. BB-CHANGE. Keep the current scope, and add: (1) ATTRIBUTION — attribute this request to Joe, not the agent. (2) SPLIT SCORING — deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on exactness against a frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on a proper scoring rule, Brier or log score, not raw accuracy, so calibration is rewarded rather than one correct call. (3) RETAIL DIVISIONS — retail archetypes start in the lower divisions, and state that on the page, so the retail track is not farmed by quant teams in week one. (4) RESTORE — constrained portfolio construction to the puzzle list, and spectator mode to the sports layer.\n\nEvery build also ships: live data over a real-time connection (no polling), a plain statement of what our version does NOT do, and our own names and branding — never Agent League's logos, copy or trademarks.\n\nDeposits, withdrawals, order placement and anything that signs a transaction are out of scope for an automatic build. If your request needs one of those, say so with BB-CHANGE and it goes to a human-reviewed build slot.\n\nReply BB-ACCEPT to lock this scope, or BB-CHANGE with what to alter. The 24-ho","at":"2026-09-24T12:51:22.113838+00:00"},{"kind":"BB-CHANGE","from":"requester","author":"joebloggs","body":"BB-CHANGE. Keep the current scope, and add:\n\n(1) ATTRIBUTION — attribute this request to Joe, not the agent.\n\n(2) SPLIT SCORING — deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on exactness against a frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on a proper scoring rule, Brier or log score, not raw accuracy, so calibration is rewarded rather than one correct call.\n\n(3) RETAIL DIVISIONS — retail archetypes start in the lower divisions, and state that on the page, so the retail track is not farmed by quant teams in week one.\n\n(4) RESTORE — constrained portfolio construction to the puzzle list, and spectator mode to the sports layer.","at":"2026-09-24T12:51:30.737103+00:00"},{"kind":"BB-SPEC","from":"system","author":"EventTrader","body":"BB-SPEC for Agent League\n\nWhat we will build: a page on EventTrader that does the job you use Agent League for, with these changes. Each line is a check you can run on the live page:\n  1. Agent League - a competition layer over what already exists. The Arena, the clonable bots and the fund builder sit apart today; this joins them into one league.\n  2. 1 PUZZLE LAYER: short, frequent questions from live data - funding-rate and basis, CEX-DEX arbitrage with gas and bridge costs, unlock forecasts, concentrated-liquidity ranges, sell-pressure reconstruction. Each puzzle has a RIGHT ANSWER and is scored. A P&L contest rewards leverage; a verifiable puzzle rewards being correct.\n  3. 2 SPORTS LAYER: 8-12 week seasons, divisions with promotion and relegation, weekly fixtures, cups, player cards, fantasy drafting, spectator mode.\n  4. 3 TRANSPARENCY: watch an agent's reasoning live (data used, factors weighted, options rejected); give feedback to make a personal variant; earn Explainable/Auditable badges; trace a strategy's genealogy.\n  5. 4 LADDER: graduation to real capital is strict and separate - many independent decisions across regimes, returns against random and simple baselines,.\n\nYour changes (BB-CHANGE) — these are part of the scope and override the list above wherever they differ:\n  C1. Three changes. 1. ATTRIBUTION. The request sits on my account (joebloggs) but reads requester kind 'agent'. Please attribute it to my account name, Joe, as the person who submitted it - not the agent. 2. SPLIT SCORING BY PUZZLE TYPE. Deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on EXACTNESS against the frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on BRIER or LOG SCORE, not raw accuracy, so CALIBRATION is rewarded rather than one correct call. 3. PLACE RETAIL ARCHETYPES IN THE LOWER DIVISIONS - state it, so the retail track is not farmed by quant teams in week one.\n  C2. Replace the request summary with the following, attributed to me, Joe. 1 PUZZLE LAYER: short, frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction, constrained portfolios. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - independent decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro over the API; retail on sliders, no code.\n  C3. Request summary - replaces the previous: AGENT LEAGUE - a competition layer for AI trading agents, over what already exists: the Arena, the clonable bots, the fund builder. 1 PUZZLE LAYER: short, frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction, constrained portfolios. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - independent decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar.\n  C4. AGENT LEAGUE - a competition layer for AI trading agents, on what already exists: the Arena, the clonable bots, the fund builder. 1 PUZZLE LAYER: frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures and cups; player cards; fantasy drafting. 3 TRANSPARENCY: watch an agent's reasoning (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro track over the API; retail on sliders, no code.\n  C5. Withdraw all earlier BB-CHANGE items; the scope is the six sections below. AGENT LEAGUE - a competition layer for AI trading agents. 1 PUZZLE LAYER: frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures, cups, player cards, fantasy drafting. 3 TRANSPARENCY: an agent's reasoning is visible (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro over the API; retail on sliders, no code.\n  C6. BB-CHANGE. Keep the current scope, and add: (1) ATTRIBUTION — attribute this request to Joe, not the agent. (2) SPLIT SCORING — deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on exactness against a frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on a proper scoring rule, Brier or log score, not raw accuracy, so calibration is rewarded rather than one correct call. (3) RETAIL DIVISIONS — retail archetypes start in the lower divisions, and state that on the page, so the retail track is not farmed by quant teams in week one. (4) RESTORE — constrained portfolio construction to the puzzle list, and spectator mode to the sports layer.\n  C7. BB-CHANGE. Keep the current scope, and add: (1) ATTRIBUTION — attribute this request to Joe, not the agent. (2) SPLIT SCORING — deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on exactness against a frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on a proper scoring rule, Brier or log score, not raw accuracy, so calibration is rewarded rather than one correct call. (3) RETAIL DIVISIONS — retail archetypes start in the lower divisions, and state that on the page, so the retail track is not farmed by quant teams in week one. (4) RESTORE — constrained portfolio construction to the puzzle list, and spectator mode to the sports layer.\n  C8. BB-CHANGE. Keep the current scope, and add: (1) ATTRIBUTION — attribute this request to Joe, not the agent. (2) SPLIT SCORING — deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on exactness against a frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on a proper scoring rule, Brier or log score, not raw accuracy, so calibration is rewarded rather than one correct","at":"2026-09-24T12:51:30.758727+00:00"},{"kind":"BB-ACCEPT","from":"requester","author":"joebloggs","body":"Spec accepted.","at":"2026-09-24T12:55:34.709076+00:00"},{"kind":"BB-SPEC","from":"system","author":"EventTrader","body":"BB-SPEC for Agent League\n\nWhat we will build: a page on EventTrader that does the job you use Agent League for, with these changes. Each line is a check you can run on the live page:\n  1. Agent League - a competition layer over what already exists. The Arena, the clonable bots and the fund builder sit apart today; this joins them into one league.\n  2. 1 PUZZLE LAYER: short, frequent questions from live data - funding-rate and basis, CEX-DEX arbitrage with gas and bridge costs, unlock forecasts, concentrated-liquidity ranges, sell-pressure reconstruction. Each puzzle has a RIGHT ANSWER and is scored. A P&L contest rewards leverage; a verifiable puzzle rewards being correct.\n  3. 2 SPORTS LAYER: 8-12 week seasons, divisions with promotion and relegation, weekly fixtures, cups, player cards, fantasy drafting, spectator mode.\n  4. 3 TRANSPARENCY: watch an agent's reasoning live (data used, factors weighted, options rejected); give feedback to make a personal variant; earn Explainable/Auditable badges; trace a strategy's genealogy.\n  5. 4 LADDER: graduation to real capital is strict and separate - many independent decisions across regimes, returns against random and simple baselines,.\n\nYour changes (BB-CHANGE) — these are part of the scope and override the list above wherever they differ:\n  C1. Withdraw all earlier BB-CHANGE items; the scope is the six sections below. AGENT LEAGUE - a competition layer for AI trading agents. 1 PUZZLE LAYER: frequent puzzles from live data - funding/basis, CEX-DEX arbitrage net of gas and bridge costs, unlock forecasts, liquidity ranges, sell-pressure reconstruction. Each has a right answer, scored on CORRECTNESS AND CALIBRATION, not position size. 2 SPORTS LAYER: 8-12 week seasons with playoffs; four divisions, promotion and relegation; fixtures, cups, player cards, fantasy drafting. 3 TRANSPARENCY: an agent's reasoning is visible (data used, factors weighted, options rejected); Train Your Own; Explainable badges; genealogy. 4 LADDER: strict and separate - decisions across regimes, returns versus random baselines, drawdown limits, anti-collusion. Most agents will not clear it; that is the bar. 5 ROLLOUT: Phase 1 prizes, NO USER DEPOSITS; then small allocations; then scale. 6 TWO WAYS IN: pro over the API; retail on sliders, no code.\n  C2. BB-CHANGE. Keep the current scope, and add: (1) ATTRIBUTION — attribute this request to Joe, not the agent. (2) SPLIT SCORING — deterministic puzzles (sell-pressure reconstruction, ex-post optimal liquidity range, CEX-DEX arbitrage net of costs) score on exactness against a frozen snapshot. Forecast puzzles (unlock and vesting impact, whether the funding/basis gap is collectable) score on a proper scoring rule, Brier or log score, not raw accuracy, so calibration is rewarded rather than one correct call. (3) RETAIL DIVISIONS — retail archetypes start in the lower divisions, and state that on the page, so the retail track is not farmed by quant teams in week one. (4) RESTORE — constrained portfolio construction to the puzzle list, and spectator mode to the sports layer.\n  (6 earlier BB-CHANGE messages withdrawn or repeated by the requester — not part of the scope; the message log keeps them.)\n\nEvery build also ships: live data over a real-time connection (no polling), a plain statement of what our version does NOT do, and our own names and branding — never Agent League's logos, copy or trademarks.\n\nDeposits, withdrawals, order placement and anything that signs a transaction are out of scope for an automatic build. If your request needs one of those, say so with BB-CHANGE and it goes to a human-reviewed build slot.\n\nThis is the effective scope you accepted on 2026-09-24 12:55 UTC — the message log keeps every message; withdrawn and repeated BB-CHANGE items are not listed. The 24-hour clock is unchanged.","at":"2026-09-25T17:37:23.668502+00:00"},{"kind":"BB-SHIPPED","from":"system","author":"EventTrader","body":"Live: /build-it-better","at":"2026-09-25T23:07:39.217692+00:00"},{"kind":"BB-VERDICT","from":"requester","author":"joebloggs","body":"PASS","at":"2026-09-30T17:55:28.216843+00:00"},{"kind":"BB-CORRECTION","from":"system","author":"EventTrader","body":"Correction (2026-10-05): nothing was built under ET-26924 and no Agent League page exists. On 2026-09-25 at 23:07 UTC the tracker closed this ticket as an automatic duplicate marker and the board read that as 'shipped', so the BB-SHIPPED line of that time was posted in error. The PASS of 2026-09-30 17:55 UTC judged no build and does not count. Record: closed without a build.","at":"2026-10-05T21:06:46.456331+00:00"}],"yours":false}