Learning

Live output right now

• Probability: calibrated (T-0001 promoted 2026-10-17). Expectancy gate: three-outcome formula (A-0003). Discord signal posts: off (owner switch).
• Graded calls so far: 315. Gate changes on record: T-0001 promote 2026-10-17; A-0002 rollback 2026-10-29.

Live news shadow trial (T-0016)

unvalidated (accruing) · graded 24 / 365 days (2026-10-17 to 2026-11-09) · 27 records · 5 rebalances

E-0009's news attention + headline tone, weekly long-short at perp costs. Each day's picks are recorded before the open they take effect at, then graded. Unvalidated: one decisive look after 365 graded days; nothing here changes live calls.

Live on-chain shadow trial (T-0017)

unvalidated (accruing) · graded 17 / 365 days (2026-10-24 to 2026-11-09) · 20 records · 4 rebalances

E-0011's transaction-count growth (Coin Metrics), weekly long-short at perp costs on 23 coins. Recorded before the open it takes effect at, graded the same way as T-0016. Unvalidated: one decisive look after 365 graded days; nothing here changes live calls.

Live unlock shadow trial (T-0019)

unvalidated (accruing) · graded 14 / 365 days (2026-10-27 to 2026-11-09) · 17 records · 17 rebalances

Short the Hyperliquid perp 7 days before large team/investor token unlocks (E-0014), from the schedule as published that day. Recorded before the open each short takes effect at; most days hold no short. Unvalidated: one decisive look after 365 graded days; nothing here changes live calls.

Live unlock shadow trial, 30-day history (T-0020)

unvalidated (accruing) · graded 14 / 365 days (2026-10-27 to 2026-11-09) · 17 records · 17 rebalances

T-0019's rule with trader-realistic eligibility: a perp needs 30 days of liquid Hyperliquid history instead of 220, so more unlocks are tradeable. Decided on its own, never pooled with T-0019. Unvalidated: one decisive look after 365 graded days; nothing here changes live calls.

Live unlock shadow trial, thin perps (T-0021)

unvalidated (accruing) · graded 14 / 365 days (2026-10-27 to 2026-11-09) · 17 records · 17 rebalances

T-0020's rule with a $250K daily-volume floor instead of $1M, so more unlocks are tradeable, graded at 15 bps per side because thin perps cost more to trade (an assumption, not measured fills). Decided on its own, never pooled. Unvalidated: one decisive look after 365 graded days; nothing here changes live calls.

Calibration (shadow)

bb-1.0.n286 · built Nov 13, 12:38 AM ET · N=286 · global target-hit 16% (90% 13% to 20%) · Brier prior 0.325 vs calibrated 0.129

Hierarchical Beta-binomial: each cell is pulled toward the global rate until it has enough of its own outcomes. Cells under the minimum N are suppressed, not shown as a probability.

Pre-registered trials (37)

• T-0001pass
  Calibrated Beta-binomial mapping vs the hand-set prior
  registered Oct 6, 10:58 PM ET · metric brier · min N 60 · manifest 251835100d · calibrated beat prior in both blocks: Brier 0.25 vs 0.363927, 0.153853 vs 0.336897
  amended by A-0001 (Oct 12, 8:51 PM ET, 4c362a5152): Exclude calls whose target was at or below zero from the live trials
  amended by A-0002 (Oct 20, 11:14 PM ET, bfbb6d7668): Expectancy gate counts all three outcomes (target, stop, expiry) at their measured value
  amended by A-0003 (Nov 9, 8:37 PM ET, 91361b6c81): Three-outcome expectancy gate restored; live statistics judge complete cohorts only; calls open and are judged only at executable prices; the gate fails closed
  amended by A-0004 (Nov 9, 8:45 PM ET, e376acd0e1): Sizing that sizes: heat and capital from the open calls, three-outcome Kelly, and a main book of at most 20 positions on $25,000
• T-0002fail
  Baseline: technical engine v0.2 calls vs doing nothing, BTC and SPY (history)
  registered Oct 6, 11:02 PM ET · metric mean_net_r · min N 100 · manifest c4895b7029 · oos1: n=2660 mean -0.0347R ci90 [-0.0966, 0.0238]; oos2: n=1956 mean -0.0692R ci90 [-0.1415, 0.0034]
• T-0003live: insufficient_n
  Crowded-funding filter on crypto calls (Hyperliquid funding z-score)
  registered Oct 6, 11:11 PM ET · metric lift_mean_net_r · min N 60 · manifest 9885f4438e · live: half1: n=12 kept=12 lift=0.0; half2: n=13 kept=13 lift=0.0
  amended by A-0001 (Oct 12, 8:51 PM ET, 4c362a5152): Exclude calls whose target was at or below zero from the live trials
  amended by A-0003 (Nov 9, 8:37 PM ET, 91361b6c81): Three-outcome expectancy gate restored; live statistics judge complete cohorts only; calls open and are judged only at executable prices; the gate fails closed
• T-0004holdout: fail · live: fail
  Challenger: drop SHORT calls (long-only)
  registered Oct 6, 11:11 PM ET · metric mean_net_r · min N 100 · manifest b67048ad68 · holdout: frozenHoldout: n=477 mean -0.1016R ci90 [-0.2406, 0.0405] vs champion -0.0468R · live: half1: n=143 kept=68 lift=0.912709; half2: n=143 kept=136 lift=0.031706
  amended by A-0001 (Oct 12, 8:51 PM ET, 4c362a5152): Exclude calls whose target was at or below zero from the live trials
  amended by A-0003 (Nov 9, 8:37 PM ET, 91361b6c81): Three-outcome expectancy gate restored; live statistics judge complete cohorts only; calls open and are judged only at executable prices; the gate fails closed
• E-0001exploration: explored
  Exploration: 15 literature-backed swing rules x 2 asset classes on development data
  registered Oct 6, 11:27 PM ET · metric alpha_t · min N 0 · manifest 9472b04fe7 · exploration: CRYPTO: none (t=0.6312); EQUITY: none (t=1.8058)
• T-0005confirmation: fail
  Confirmation on untouched data: 4 stock trend/dip rules and 2 crypto short-term momentum rules
  registered Oct 6, 11:32 PM ET · metric alpha_t · min N 0 · manifest cb1d775c4e · confirmation: EQUITY:tsmom_90_lo: fail fresh alpha -0.0069 t -0.4329, prior alpha 0.01 t 0.2593; EQUITY:trend_sma200_lo: fail fresh alpha 0.0047 t 0.3042, prior alpha 0.0127 t 0.3468; EQUITY:dip_uptrend_lo: fail fresh alpha -0.0084 t -0.9583, prior alpha -0.0216 t -0.5866; EQUITY:engine_v02_lo: fail fresh alpha -0.003 t -0.2059, prior alpha -0.0097 t -0.1752; CRYPTO:xs_mom_5_lo: fail fresh alpha -0.0159 t -0.2774, prior alpha 0.0011 t 0.0189; CRYPTO:xs_mom_5_ls: fail fresh alpha -0.1199 t -2.145, prior alpha 0.009 t 0.1595
• T-0006confirmation: fail
  Crypto crowding: long the least-crowded coins, short the most-crowded (Hyperliquid funding)
  registered Oct 6, 11:46 PM ET · metric alpha_t · min N 0 · manifest ed276e20a6 · confirmation: funding_xs_ls: fail alpha -0.135 t -3.2818 (halves -0.0876, -0.1814); funding_xs_lo: fail alpha -0.0988 t -3.7498 (halves -0.0836, -0.1)
• T-0007confirmation: pass · audit: pass_not_promoted
  Post-earnings drift from SEC EDGAR (standardized unexpected earnings)
  registered Oct 6, 11:50 PM ET · metric alpha_t · min N 0 · manifest 5dd38f631f · confirmation: pead_ls: fail alpha 0.007 t 1.672 (halves 0.01, 0.0039); pead_lo: PASS alpha 0.0089 t 2.5265 (halves 0.0113, 0.0063) · audit: Registered portfolio-alpha test passed (pead_lo t=2.53), but a market-adjusted event study of the same 697 long trades shows mean 20-day excess 0.0029 (naive t 1.08, 90% weekly-block CI [-0.0013, 0.0071]); 10-Q-dated events show none. Portfolio alpha was inflated by market return during sparse invested days (exposure ~6%, beta underestimated). Not promoted. Live use also needs EPS at the 8-K date (not in XBRL until the 10-Q).
• T-0008confirmation: fail
  Funding momentum: long the most-crowded coins (confirmation on 25 untouched coins)
  registered Oct 7, 12:30 AM ET · metric excess_t · min N 0 · manifest e643aff92a · confirmation: fundmom_ls: fail excess -0.0939 t -1.299 (halves 0.0233, -0.2112); alpha -0.0938 t -1.3033; fundmom_lo: fail excess -0.0432 t -1.075 (halves 0.0199, -0.1065); alpha -0.0645 t -1.4228
• T-0009confirmation: fail
  Drift after an EPS surprise against analyst consensus (Finnhub estimates, S&P 100)
  registered Oct 8, 9:42 PM ET · metric excess_t · min N 150 · manifest 5ffe35fe80 · confirmation: N=176 (153 beats, 23 misses); cs_ls: fail excess 0.001 t 0.0874 (halves 0.0002, 0.0018); cs_lo: fail excess 0.001 t 0.1138 (halves 0.008, -0.006)
• T-0010history: fail
  Crypto stop/target bracket sized for a 7-day horizon (engine v0.2 calls, history)
  registered Oct 16, 10:51 PM ET · metric mean_net_r · min N 100 · manifest 94acef5c11 · history: base_1.5_3.0: oos1 -0.1036R [-0.2057], oos2 -0.1232R [-0.2537], confirm -0.0824R [-0.1347]; b_1.5_1.5: oos1 -0.1023R [-0.1851], oos2 -0.1261R [-0.2353], confirm -0.0942R [-0.1394]; c_2.0_2.0: oos1 -0.0703R [-0.1504], oos2 -0.1211R [-0.2264], confirm -0.0667R [-0.1086]; d_1.0_1.0: oos1 -0.1805R [-0.2594], oos2 -0.1464R [-0.2457], confirm -0.1565R [-0.1983]; e_2.5_2.5: oos1 -0.0453R [-0.1209], oos2 -0.0923R [-0.1825], confirm -0.0496R [-0.0871]; f_2.0_3.0: oos1 -0.0735R [-0.1577], oos2 -0.117R [-0.2333], confirm -0.0551R [-0.0989]
• E-0002exploration+confirmation: no_finalist
  Crypto: slow, low-turnover signals from new free data (exploration + pre-committed confirmation)
  registered Oct 16, 11:35 PM ET · metric n/a · min N 0 · manifest 1e4bf4c1fe · exploration+confirmation: stable_flow_14: alpha_t est 0.0194 t 0.0982; stable_flow_30: alpha_t est 0.0317 t 0.1508; lowvol_xs_lo: excess_t est -0.045 t -1.093; lowvol_xs_ls: excess_t est -0.117 t -1.488; btc_catchup_lo: excess_t est -0.0544 t -2.1473
• E-0003exploration+confirmation: no_finalist
  Crypto: attention, lottery (MAX), 52-week-high, and a walk-forward combination of 8 signals (exploration + pre-committed confirmation)
  registered Oct 16, 11:49 PM ET · metric n/a · min N 0 · manifest 2c64485339 · exploration+confirmation: attention_lo: excess_t est 0.0321 t 0.7434; attention_ls: excess_t est 0.0455 t 0.6897; maxlow_lo: excess_t est -0.0602 t -1.5729; maxlow_ls: excess_t est -0.111 t -1.5338; high52_lo: excess_t est 0.0314 t 0.9252; high52_ls: excess_t est 0.0199 t 0.3208; combo_ridge_lo: excess_t est -0.0292 t -0.9651; combo_ridge_ls: excess_t est -0.0749 t -1.4861
• E-0004exploration: explored
  Exploration: 52-week high, Donchian breakout, ADX trend, volume-confirmed breakout and vol-managed momentum x 2 asset classes on development data
  registered Oct 16, 11:53 PM ET · metric alpha_t · min N 0 · manifest 544f376323 · exploration: CRYPTO: none (t=0.79); EQUITY: adx_trend_lo (t=2.1004)
• T-0011history: fail
  Volatility-regime filter on replayed engine v0.2 calls: VIX term structure (stocks) and DVOL percentile (crypto)
  registered Oct 16, 11:53 PM ET · metric lift_mean_net_r · min N 200 · manifest 099f094279 · history: half1: n=2319 kept=1920 lift=0.012495; half2: n=2320 kept=2053 lift=0.013501
• T-0012confirmation: fail
  Confirmation on untouched data: ADX(14) > 25 trend rule, long-only, stocks
  registered Oct 17, 12:00 AM ET · metric alpha_t · min N 0 · manifest d09401ea74 · confirmation: EQUITY:adx_trend_lo: fail fresh alpha -0.005 t -0.5242, prior alpha -0.0174 t -0.4195
• T-0013history: fail
  DVOL gate on crypto LONG calls, confirmation on 24 untouched coins
  registered Oct 17, 12:00 AM ET · metric lift_mean_net_r · min N 200 · manifest 8ca3053c9a · history: half1: n=615 kept=426 lift=0.026064; half2: n=615 kept=397 lift=0.075324
• E-0005exploration+confirmation: no_finalist
  Crypto breadth, point-in-time universe: 52-week-high and attention (and their blend) on every Coinbase USD coin, weekly and monthly (exploration + pre-committed confirmation)
  registered Oct 17, 12:17 AM ET · metric n/a · min N 0 · manifest 329155a043 · exploration+confirmation: 427 coins with history; high52_w: excess -0.0341 t -0.9737; high52_m: excess -0.0523 t -1.5373; attention_w: excess -0.0723 t -1.6205; attention_m: excess -0.059 t -2.0321; blend_w: excess -0.0841 t -2.1313; blend_m: excess -0.1032 t -2.8881
• T-0014confirmation: fail
  Crypto hype reversal: long the coins furthest from their 52-week high with fading volume (confirmation on data E-0005 never scored)
  registered Oct 17, 12:36 AM ET · metric excess_t · min N · manifest afacde6e96 · confirmation: contra_blend_m: fail earlierCoins excess 0.0058 t 0.2118, broadOutOfTime excess 0.0044 t 0.1385; contra_blend_w: fail earlierCoins excess -0.0351 t -1.4568, broadOutOfTime excess 0.0285 t 0.7799
• E-0006exploration+confirmation: no_finalist
  Crypto cross-sectional momentum on the broad point-in-time universe, monthly, with and without a BTC trend gate (exploration + pre-committed confirmation)
  registered Oct 17, 12:42 AM ET · metric n/a · min N 0 · manifest b1f8e8f527 · exploration+confirmation: mom28_m: excess -0.1142 t -2.9988; mom90_m: excess -0.0825 t -2.113; mom28_m_btc: excess -0.083 t -2.6522
• T-0015exploration+confirmation: no_finalist
  Crypto at perpetual-futures costs: funding momentum, 5-day reversal, 28-day trend, long-short on the broad Hyperliquid-listed universe (exploration + one-shot confirmation on the untouched last year)
  registered Oct 17, 12:55 AM ET · metric · min N · manifest 55ef202cbc · exploration+confirmation: fundmom_ls: excess -0.0445 t -0.509; xsrev5_ls: excess -0.1575 t -1.5714; tsmom28_ls: excess -0.2133 t -2.2321
• E-0007exploration+confirmation: no_finalist
  Crypto regime switch: momentum in trending markets, reversal otherwise, monthly on the broad point-in-time universe (exploration + one-shot confirmation on the untouched last year)
  registered Oct 17, 12:55 AM ET · metric n/a · min N 0 · manifest 9d11f9ff9a · exploration+confirmation: regime_breadth_m: excess -0.0052 t -0.1693; regime_btc_m: excess -0.0593 t -1.8336
• E-0008exploration+confirmation: no_finalist
  Crypto meta-labeling: a walk-forward model that predicts each engine call's net R and keeps only the calls it expects to win (exploration + one-shot confirmation on the untouched last year)
  registered Oct 17, 3:46 PM ET · metric n/a · min N 0 · manifest 31dfaa4203 · exploration+confirmation: explore: all {'n': 13581, 'mean_net_r': -0.0828, 'ci90': [-0.1354, -0.0293], 'hit_rate': 0.0998}, taken {'n': 3447, 'mean_net_r': -0.1188, 'ci90': [-0.2324, -0.0072], 'hit_rate': 0.1129}, share 0.2538
• E-0009exploration+confirmation: fail
  Crypto news and events: news attention, the live headline lexicon, and keyword event trades on the free Alpaca news archive, at perp costs (exploration + one-shot confirmation on the last year)
  registered Oct 17, 4:17 PM ET · metric n/a · min N 0 · manifest f1c1b0b376 · exploration+confirmation: attn_ls: excess 0.0125 t 0.3297; tone_ls: excess 0.0458 t 1.6293; attn_tone_ls: excess 0.0759 t 2.5931; ev_pos_long: excess -0.0132 t -1.0945; ev_neg_short: excess 0.0077 t 0.7443
• T-0016running
  Live shadow trial: crypto news attention + headline tone, weekly long-short at perp costs (E-0009 attn_tone_ls, forward only)
  registered Oct 17, 4:47 PM ET · metric excess_t · min N 365 · manifest 8fce49ef27
• E-0010exploration+confirmation: void
  Crypto on-chain network activity and valuation (Coin Metrics) and options-implied fear (Deribit DVOL), at perp costs (exploration + one-shot confirmation on the last year)
  registered Oct 24, 3:50 PM ET · metric n/a · min N 0 · manifest f2672dee0a · exploration+confirmation: run void before any result: the Coin Metrics community API returns no rows for BNB (catalog lists it; pro-only) and only 519 rows for DOT (ends 2022-06). Re-registered as E-0011 on the 23 coins it serves.
• E-0011exploration+confirmation: fail
  Crypto on-chain network activity and valuation (Coin Metrics) and options-implied fear (Deribit DVOL), at perp costs (re-run of void E-0010 on the 23 coins the free API serves; exploration + one-shot confirmation on the last year)
  registered Oct 24, 3:51 PM ET · metric n/a · min N 0 · manifest 38e50ff905 · exploration+confirmation: netgrowth_ls: excess 0.1238 t 1.7663; txgrowth_ls: excess 0.1618 t 2.1938; mvrv_value_ls: excess -0.0298 t -0.3169; growth_value_ls: excess -0.0003 t -0.0029; dvol_fear_long: excess -0.0999 t -5.8528
• T-0017running
  Live shadow trial: crypto transaction-count growth, weekly long-short at perp costs (E-0011 txgrowth_ls, forward only)
  registered Oct 24, 4:02 PM ET · metric excess_t · min N 365 · manifest 35bbc5f15b
• E-0012exploration+confirmation: fail
  Crypto delta-neutral funding carry: long Coinbase spot + short Hyperliquid perp, collecting funding (exploration + one-shot confirmation on the last year)
  registered Oct 24, 4:03 PM ET · metric n/a · min N 0 · manifest 0e0517305d · exploration+confirmation: carry_top_m: net 0.1709/yr t 5.533; carry_btc_eth: net 0.1536/yr t 7.273; carry_hyst: net 0.0954/yr t 3.282
• E-0013confirmation: fail
  Out-of-period replication: do E-0011's on-chain growth signals (and a bull-market-only version) hold on 2018-07..2022-07, a period no rule of theirs has seen?
  registered Oct 24, 4:43 PM ET · metric n/a · min N 0 · manifest f1537b1399 · confirmation: txgrowth_ls: excess -0.0349 t -0.3524 (halves 0.071 / -0.141); netgrowth_ls: excess -0.0909 t -0.9468 (halves 0.0328 / -0.2147); txgrowth_bull_ls: excess 0.0045 t 0.0537 (halves 0.0998 / -0.0909)
• E-0014exploration+confirmation: no_finalist
  Crypto token unlocks: short the Hyperliquid perp around large scheduled cliff unlocks, at perp costs (exploration + one-shot confirmation on the last year)
  registered Oct 24, 6:53 PM ET · metric n/a · min N 0 · manifest c5400d3919 · exploration+confirmation: pre7_short: excess 0.0256 t 1.4586; pre30_short: excess 0.0407 t 1.254; post7_short: excess 0.0013 t 0.0733; pre7_short_insider: excess 0.0289 t 1.8542
• T-0018confirmation: fail
  One-shot confirmation on the untouched last year: short the Hyperliquid perp 7 days before large team/investor token unlocks (E-0014 pre7_short_insider)
  registered Oct 26, 8:35 PM ET · metric · min N · manifest 5c10a6ea12 · confirmation: pre7_short_insider 2025-10-01..2026-10-01: excess 0.0198 t 1.2924, 154 windows, mean 0.0054, won 0.149
• T-0019running
  Live shadow trial: short the Hyperliquid perp 7 days before large team/investor token unlocks (forward only)
  registered Oct 26, 8:35 PM ET · metric excess_t · min N 365 · manifest 3c228a3db8
• T-0020running
  Live shadow trial: T-0019's pre-unlock short with trader-realistic eligibility (30 days of perp history instead of 220; forward only)
  registered Oct 26, 10:10 PM ET · metric excess_t · min N 365 · manifest b14359529a
• T-0021running
  Live shadow trial: T-0020's pre-unlock short with a $250K Hyperliquid volume floor instead of $1M, graded at 15 bps per side (forward only)
  registered Oct 26, 10:26 PM ET · metric excess_t · min N 365 · manifest 734ac19fb7
• E-0015exploration+confirmation: no_finalist
  Crypto protocol fundamentals: weekly long-short on DefiLlama fee and revenue growth on Hyperliquid perps, at perp costs (exploration + one-shot confirmation on the last year)
  registered Oct 26, 10:52 PM ET · metric n/a · min N 0 · manifest a826b6e0d9 · exploration+confirmation: fee_mom30_ls: excess 0.0088 t 0.1221; fee_mom90_ls: excess 0.1276 t 1.5068; fee_vs_price_ls: excess -0.0194 t -0.2204; rev_mom30_ls: excess -0.0238 t -0.4309
• M-0001first-look: selection shown
  Measurement: do ORACLE's live calls beat a random same-day pick, and the market? (placebo and benchmark attribution of every matured call)
  registered Nov 9, 8:14 PM ET · metric n/a · min N 0 · manifest d8a6fef793 · first-look: v1 stock longs since 10-09: n 148 on 7 days, vs placebo 0.2221R t 2.2331, permutation p 0.0044