Skip to content

Written down before it played out.

Each risk rating is chained to the one before it the moment it is made. Changing or backdating an old entry breaks every hash after it. A look back at what was rated, not a forecast and not advice.
Illustration, generated with AI

Both sides of the log

Rated high or severe, hour passed
70
Stood 70% or more lower one hour later
5
Had not fallen 70% one hour later
65

From the latest 500 ratings, a short window that is not yet a reliable measure.

Latest high ratings, whatever happened next

  • $FIONA rated high at 2026-11-22 02:08 UTC

    Had not fallen 70% one hour later

    8b777e7cfcc4a937a2404e25a3197331f7363a8ec074537682ef313f79112999

  • $RIZZMAS67 rated high at 2026-11-22 02:08 UTC

    Had not fallen 70% one hour later

    2cc1193e50dea391b9ea2c3222b529304a35434df1acdf3c0fb747ce46674d69

  • $SOLRAIDERS rated high at 2026-11-22 02:08 UTC

    Had not fallen 70% one hour later

    2978724342ebb3c875b72082b093f8ea28db756774ddda4411b1d67fd27f2e92

  • $Pumpmas rated high at 2026-11-22 02:06 UTC

    Stood 70% or more lower one hour later

    f90f661a408c867b9a955822f0b5ebc0ee645e76d668a4e653965f8daa610cbb

  • $pumped rated high at 2026-11-22 02:06 UTC

    Had not fallen 70% one hour later

    78754724d2426347b8038e2cef0422496dc941a0b73da999219a0a8dc01ce385

  • $ину rated high at 2026-11-22 02:06 UTC

    Had not fallen 70% one hour later

    15d1a4cf714960ca3fdccc7bb6568084f3489cdd504a0be6ed3d6afd2c2cfcaa

  • $V15 rated high at 2026-11-22 02:00 UTC

    Had not fallen 70% one hour later

    8e7c24cfa98b66353ff390abeafbfc7a3729c853a95db40fba050a9074eaca23

  • $DCTY rated high at 2026-11-22 01:56 UTC

    Had not fallen 70% one hour later

    81c4f59b1481a57a5069dd46a323d9e101cb3886b7faf4bf086b8ce529aa0c0c

  • $PUMPMAS rated high at 2026-11-22 01:56 UTC

    Had not fallen 70% one hour later

    d4e26dc533bbd9967f5b520405c292b3a5047935bcfc08640457225e85bf103d

  • $RIZZMASRIZZMAS rated high at 2026-11-22 01:55 UTC

    Had not fallen 70% one hour later

    ecf708cbd8fc12e0cc69e8509d651d295ef97f3b2d49d67922c9ad2f0997c28f

  • $PushCoin rated high at 2026-11-22 01:55 UTC

    Had not fallen 70% one hour later

    e2b5da34c5ab4667c1da22b61bd156d28995c8e9137031f2fd4eb32acddb5883

  • $PI rated high at 2026-11-22 01:50 UTC

    Had not fallen 70% one hour later

    86b14d4031e2c9f46d7b32754799661888245099f769335ecd32beb93937bdcf

  • $RIZZMAS31 rated high at 2026-11-22 01:50 UTC

    Had not fallen 70% one hour later

    4a307efdb772cdf0ad37ccc6e0ba98d6bedfc68ac57c9ac5c8199c68e346b8fc

  • $? rated high at 2026-11-22 01:48 UTC

    Had not fallen 70% one hour later

    6bf3dfd607a185148004cf895e22b6e59ae8ca1458ee6f9a34701e8bc7fd49af

  • $RIZZMAS rated high at 2026-11-22 01:48 UTC

    Had not fallen 70% one hour later

    80575e9e7729d4a573c84739c6c39cbaec1c0eeb25148e74e52bf0b9b172994e

  • $CashJak rated high at 2026-11-22 01:45 UTC

    Had not fallen 70% one hour later

    ed3314fa0032b8712021c2ea0982c29e0be690bf1eb6a6557d544e8dd47fd402

  • $ELF rated high at 2026-11-22 01:41 UTC

    Had not fallen 70% one hour later

    e2ad3e1e7f39f451c8d108cd4f19700d898b426908a552d1d1799c7bbc706e17

  • $stick rated high at 2026-11-22 01:40 UTC

    Had not fallen 70% one hour later

    7c598551d1965e3eb8c274f1f688cdd3c4a1b8b4180ef833e6c2b3cd1003e337

  • $RIZZMAS67 rated high at 2026-11-22 01:35 UTC

    Had not fallen 70% one hour later

    4d1d50955bc3864c5bdaa54905df19f3e7c0a8f41d3f2a5ab66bf3271627bbb6

  • $SOLRAIDERS rated high at 2026-11-22 01:35 UTC

    Had not fallen 70% one hour later

    99ebf01f85d8fada5b2c1e497f42be2977e331e1898d02c4360d949e0dd6918e

Closing hash per day

  • 2026-11-22still running559 ratings19f496fad186f07fc0cb7ec173ea7485785c859211bc7c25f2544300657b3ee0
  • 2026-11-214,089 ratingsfe1abdf048c7d7b82102c1f6b515dd47187ae83780bbd00dd948fd3af12f3975Sealed on Solana
  • 2026-11-204,101 ratings5674694b38487943f00a902fa85f0461ff95e5bf4e76e935f0be54f780b7630eSealed on Solana
  • 2026-11-194,076 ratings4774c023fd40b584828e7449782bbf5bbdaf73975e3c71d41c59ec875db3c197Sealed on Solana
  • 2026-11-184,186 ratingsa47865f5b78e7f8886262465edac8577079313855fcd3da9299318ab5265a04cSealed on Solana
  • 2026-11-174,281 ratings37b6b20a2e0a4774aaf6e23425a0eeee8064f9ec52aa06e5cff1a3259199ec0dSealed on Solana
  • 2026-11-164,346 ratingse0376276266defc465262c40b186ad1851b10107662a630f4c76d6cf46e60f0aSealed on Solana
  • 2026-11-154,149 ratings012a58999701d358d1be8589582919fcd6370292b537a5cfd678c107e1eed9ceSealed on Solana
  • 2026-11-144,409 ratings4940f8c30168916b05bdc3cfd13ecd739ab7eef0436b0a16988c928b070ea0c6Sealed on Solana
  • 2026-11-134,381 ratings1317e819bf6cf25db912f93bd43b36991944ea2297319989296aea71bb13764bSigned by FvzurR…
  • 2026-11-124,183 ratings7d41edfc481501e5735e2b2ae8ae8b18842fb2b854beb4d18b31fad7c8e07a7fSigned by FvzurR…
  • 2026-11-114,215 ratings6be7c4a294027464c433edb6a7ef699f96d422fcb190fd2fdf898e75877de007Signed by FvzurR…
  • 2026-11-103,980 ratings1b35b29baa8e0eb78c309db3e7ea9f6e3dde8f7a7efee1f3be9964f1e323829fSigned by FvzurR…
  • 2026-11-093,832 ratingsd97f6f595b37d4a60ca0abf331d6e2040c19ccfafcb82ccf5877c534f5d9bd19Signed by FvzurR…

Check it yourself

The raw ratings of a day are released 30 days after it ended, one JSON entry per line. Each hash is the SHA-256 of prev|i|t|mint|risk|rule|mc0 with empty text for a missing value, and each entry carries the hash of the one before. The last hash of the day has to match the closing hash published above.

verify.mjs does it for you: Node, no packages. Run node verify.mjs proof-DAY.jsonl and compare the last hash it prints.

Each finished day is also sealed: its closing hash is signed with an Ed25519 key and written to Solana as a memo by the same wallet, so the time it was fixed does not depend on this site. Add --seal seal-DAY.json from /proof/seal/DAY.json to check the signature too.

80,715

ratings in the log. Information only, not financial advice.

Every class, side by side

High is a warning, so most High coins should fall. The test is whether they fall more often than Low ones. Here is every class, with how many coins it rated, whatever the order turns out to be.

ClassRated, hour passedFell 70%+
Low8,1254.4%
Moderate4,52911.5%
Elevated41,9479.3%
High15,55315.7%
Very high43139.9%
All rated coins70,58510.4%

Coins rated 5 Nov 2026 to 22 Nov 2026, counted when the hour had passed. A class shows a rate from 50 coins on. Outcomes before 5 Nov were measured differently and are left out. Measured by CON, not audited by a third party. Not a forecast, not advice.

Signal Scorecard

For each signal: how often it fired on coins held back from tuning, and how often the coin then stood 70% or more lower one hour later. The range is a 95% interval. Method

SignalFiredRightBase rate
Dev sell32715%5%
Bundle detection7,7652%2%
Fresh wallet detection2,43623%9%
Holder concentration2,09311%9%
Developer history1,2718%9%
Right
The coin stood 70% or more below its market cap at the signal, one hour later.
Base rate
How often that happened to all test coins, signal or not. A signal earns its place only if the lower end of its range is above it.
Test period
21,489 coins rated after 14 Nov 2026, held back while the thresholds were tuned on the coins before.
Minimum sample
A signal shows once it fired on 100 test coins.

Past outcomes of coins we analysed, measured by CON, not audited by a third party. Not a forecast for any coin, not advice.

Forecast record

Every forecast is written to a chained register at the moment it is made, with the features known then, and resolved later against what happened. Members see a probability only for a class whose model beats the base rate on a later holdout period, with its range. Everywhere else it says not proven.

Outcome forecasts

Measured at the market cap and liquidity checks after the rating.

  • 7 days · rule token_survival@1

    Token Survival

    The coin still has a market 7 days after the rating: a price above zero at the 7-day check.

    Not proven: too few cases

    No model trained yet: the first models need 30 days of recorded outcomes.

    No holdout yet. The first models need 30 days of recorded outcomes.

  • 24 hours · rule liquidity_outflow@1

    Liquidity Outflow

    Within 24 hours the pool falls below half of its liquidity at the rating, or below 500 USD.

    Not proven

    Does not beat the base rate with its range: skill 0.0002, range -0.0079 to 0.0102.

    Holdout cases
    1,402 (881 happened)
    Brier score
    0.242 vs 0.242 base rate
    Skill over base rate
    0.000 (-0.008 to 0.010)
    Base rate
    54%
    Hits at 50%
    855
    False alarms
    484
    Missed
    26
    Register, resolved
    none yet
    Calibration of Liquidity Outflow on the holdout41% predicted, 41% observed, 63 cases; 53% predicted, 63% observed, 1280 cases; 70% predicted, 78% observed, 59 cases0%100% predicted100%0
    Predicted against observed on the holdout. On the dashed line a forecast of 30% happens 30% of the time.
  • 7 days · rule graduation@1

    Graduation Probability

    A coin on the bonding curve at the rating moves to a pool within 7 days (seen at the 24-hour or 7-day check).

    Not proven: too few cases

    No model trained yet: the first models need 30 days of recorded outcomes.

    No holdout yet. The first models need 30 days of recorded outcomes.

  • 6 hours · rule momentum_continuation@1

    Momentum Continuation

    Market cap is higher at 1 hour than at the rating, and higher again at 6 hours than at 1 hour.

    Beats the base rate on the holdout

    Beats the base rate on the holdout: skill 0.0208, range 0.0165 to 0.0251.

    Holdout cases
    18,000 (822 happened)
    Brier score
    0.043 vs 0.044 base rate
    Skill over base rate
    0.021 (0.017 to 0.025)
    Base rate
    4%
    Hits at 50%
    0
    False alarms
    0
    Missed
    822
    Register, resolved
    1,051, Brier 0.045
    Calibration of Momentum Continuation on the holdout4% predicted, 4% observed, 17585 cases; 14% predicted, 11% observed, 413 cases; 21% predicted, 0% observed, 2 cases0%100% predicted100%0
    Predicted against observed on the holdout. On the dashed line a forecast of 30% happens 30% of the time.
  • 1 hour · rule early_momentum@1

    Early Momentum

    Market cap reaches at least twice its value at the rating within the first hour (checks at 5, 15 and 60 minutes).

    Beats the base rate on the holdout

    Beats the base rate on the holdout: skill 0.0098, range 0.0074 to 0.0124.

    Holdout cases
    18,000 (1,465 happened)
    Brier score
    0.074 vs 0.075 base rate
    Skill over base rate
    0.010 (0.007 to 0.012)
    Base rate
    8%
    Hits at 50%
    0
    False alarms
    0
    Missed
    1465
    Register, resolved
    1,942, Brier 0.090
    Calibration of Early Momentum on the holdout8% predicted, 7% observed, 14729 cases; 13% predicted, 13% observed, 3251 cases; 22% predicted, 29% observed, 17 cases; 34% predicted, 0% observed, 3 cases0%100% predicted100%0
    Predicted against observed on the holdout. On the dashed line a forecast of 30% happens 30% of the time.

Market structure forecasts

Measured from recorded trading after the rating: holders, sells, large trades, roles, funding groups, narratives and market volume. Observations, not actions.

  • 24 hours · rule dev_sell_risk@1

    Dev Sell Risk

    The creator wallet held tokens at the rating and its recorded net holding is at most half of that 24 hours later.

    Beats the base rate on the holdout

    Beats the base rate on the holdout: skill 0.2894, range 0.2634 to 0.319.

    Holdout cases
    2,346 (1,360 happened)
    Brier score
    0.173 vs 0.244 base rate
    Skill over base rate
    0.289 (0.263 to 0.319)
    Base rate
    58%
    Hits at 50%
    1239
    False alarms
    479
    Missed
    121
    Register, resolved
    none yet
    Calibration of Dev Sell Risk on the holdout2% predicted, 2% observed, 328 cases; 15% predicted, 10% observed, 31 cases; 25% predicted, 23% observed, 43 cases; 36% predicted, 35% observed, 69 cases; 46% predicted, 50% observed, 157 cases; 55% predicted, 59% observed, 337 cases; 65% predicted, 70% observed, 615 cases; 75% predicted, 77% observed, 522 cases; 84% predicted, 87% observed, 217 cases; 93% predicted, 81% observed, 27 cases0%100% predicted100%0
    Predicted against observed on the holdout. On the dashed line a forecast of 30% happens 30% of the time.
  • 6 hours · rule unusual_selling@1

    Unusual Selling

    Within 6 hours after the rating, recorded sells reach at least twice the recorded buys in SOL and at least 5 SOL.

    Beats the base rate on the holdout

    Beats the base rate on the holdout: skill 0.0621, range 0.0516 to 0.0721.

    Holdout cases
    6,368 (1,466 happened)
    Brier score
    0.166 vs 0.177 base rate
    Skill over base rate
    0.062 (0.052 to 0.072)
    Base rate
    22%
    Hits at 50%
    7
    False alarms
    7
    Missed
    1459
    Register, resolved
    1,051, Brier 0.172
    Calibration of Unusual Selling on the holdout5% predicted, 5% observed, 1083 cases; 16% predicted, 16% observed, 832 cases; 25% predicted, 26% observed, 3007 cases; 33% predicted, 34% observed, 1271 cases; 44% predicted, 45% observed, 161 cases; 52% predicted, 54% observed, 13 cases; 63% predicted, 0% observed, 1 cases0%100% predicted100%0
    Predicted against observed on the holdout. On the dashed line a forecast of 30% happens 30% of the time.
  • 24 hours · rule holder_growth@1

    Holder Growth

    Wallets with a positive recorded holding grow to at least 1.5 times their number at the rating within 24 hours (from 10 holders, recorded trades, not an on-chain count).

    Beats the base rate on the holdout

    Beats the base rate on the holdout: skill 0.0563, range 0.0327 to 0.0807.

    Holdout cases
    3,938 (868 happened)
    Brier score
    0.163 vs 0.173 base rate
    Skill over base rate
    0.056 (0.033 to 0.081)
    Base rate
    19%
    Hits at 50%
    118
    False alarms
    148
    Missed
    750
    Register, resolved
    none yet
    Calibration of Holder Growth on the holdout7% predicted, 9% observed, 506 cases; 15% predicted, 15% observed, 1781 cases; 24% predicted, 28% observed, 1029 cases; 35% predicted, 41% observed, 254 cases; 43% predicted, 36% observed, 102 cases; 55% predicted, 29% observed, 24 cases; 65% predicted, 27% observed, 109 cases; 74% predicted, 34% observed, 56 cases; 86% predicted, 82% observed, 57 cases; 93% predicted, 80% observed, 20 cases0%100% predicted100%0
    Predicted against observed on the holdout. On the dashed line a forecast of 30% happens 30% of the time.
  • 24 hours · rule holder_distribution_change@1

    Holder Distribution Changes

    The share of the ten largest recorded holders rises by at least 10 percentage points within 24 hours (from 10 holders).

    Beats the base rate on the holdout

    Beats the base rate on the holdout: skill 0.1065, range 0.0849 to 0.123.

    Holdout cases
    3,779 (1,066 happened)
    Brier score
    0.182 vs 0.204 base rate
    Skill over base rate
    0.106 (0.085 to 0.123)
    Base rate
    32%
    Hits at 50%
    233
    False alarms
    227
    Missed
    833
    Register, resolved
    none yet
    Calibration of Holder Distribution Changes on the holdout4% predicted, 6% observed, 705 cases; 16% predicted, 18% observed, 397 cases; 26% predicted, 26% observed, 851 cases; 35% predicted, 34% observed, 741 cases; 43% predicted, 40% observed, 625 cases; 56% predicted, 51% observed, 443 cases; 64% predicted, 50% observed, 16 cases; 71% predicted, 0% observed, 1 cases0%100% predicted100%0
    Predicted against observed on the holdout. On the dashed line a forecast of 30% happens 30% of the time.
  • 6 hours · rule whale_accumulation@1

    Whale Accumulation

    Within 6 hours, single buys of at least 10 SOL exceed single sells of at least 10 SOL by 20 SOL or more.

    Beats the base rate on the holdout

    Beats the base rate on the holdout: skill 0.0256, range 0.0026 to 0.0477.

    Holdout cases
    6,368 (74 happened)
    Brier score
    0.011 vs 0.011 base rate
    Skill over base rate
    0.026 (0.003 to 0.048)
    Base rate
    2%
    Hits at 50%
    0
    False alarms
    0
    Missed
    74
    Register, resolved
    1,051, Brier 0.007
    Calibration of Whale Accumulation on the holdout1% predicted, 1% observed, 6264 cases; 13% predicted, 10% observed, 101 cases; 21% predicted, 50% observed, 2 cases; 39% predicted, 0% observed, 1 cases0%100% predicted100%0
    Predicted against observed on the holdout. On the dashed line a forecast of 30% happens 30% of the time.
  • 6 hours · rule smart_money_entry@1

    Smart-Money Entry

    Within 6 hours, at least one wallet with a recorded role (KOL, smart money list or smart money score from wallet intelligence) buys without a holding at the rating.

    Not proven: too few cases

    Only 0 cases on one side, 20 needed.

    Holdout cases
    5,739 (0 happened)
    Brier score
    0.000 vs 0.000 base rate
    Skill over base rate
    n/a
    Base rate
    0%
    Hits at 50%
    0
    False alarms
    0
    Missed
    0
    Register, resolved
    924, Brier 0.000
    Calibration of Smart-Money Entry on the holdout1% predicted, 0% observed, 5739 cases0%100% predicted100%0
    Predicted against observed on the holdout. On the dashed line a forecast of 30% happens 30% of the time.
  • 6 hours · rule wallet_cluster_activity@1

    Wallet Cluster Activity

    Within 6 hours, at least three wallets with the same known funding source trade the coin. A shared source does not prove one person.

    Not proven: too few cases

    Only 0 cases on one side, 20 needed.

    Holdout cases
    6,368 (0 happened)
    Brier score
    0.000 vs 0.000 base rate
    Skill over base rate
    n/a
    Base rate
    0%
    Hits at 50%
    0
    False alarms
    0
    Missed
    0
    Register, resolved
    1,051, Brier 0.000
    Calibration of Wallet Cluster Activity on the holdout1% predicted, 0% observed, 6368 cases0%100% predicted100%0
    Predicted against observed on the holdout. On the dashed line a forecast of 30% happens 30% of the time.
  • 24 hours · rule narrative_growth@1

    Narrative Growth

    The strongest word of the name or ticker appears in at least 5 launches in the 24 hours after the rating and in at least 1.5 times as many as in the 24 hours before.

    Not proven

    Does not beat the base rate with its range: skill 0.0016, range -0.0048 to 0.0077.

    Holdout cases
    5,444 (940 happened)
    Brier score
    0.145 vs 0.145 base rate
    Skill over base rate
    0.002 (-0.005 to 0.008)
    Base rate
    22%
    Hits at 50%
    0
    False alarms
    0
    Missed
    940
    Register, resolved
    none yet
    Calibration of Narrative Growth on the holdout8% predicted, 14% observed, 94 cases; 16% predicted, 13% observed, 1696 cases; 24% predicted, 19% observed, 3473 cases; 31% predicted, 16% observed, 178 cases; 44% predicted, 33% observed, 3 cases0%100% predicted100%0
    Predicted against observed on the holdout. On the dashed line a forecast of 30% happens 30% of the time.
  • 24 hours · rule regime_expansion@1

    Market Regime

    Recorded market volume in the 24 hours after the rating reaches at least 1.25 times the median daily volume of the seven stored days before (the expansion threshold of market_regime@1).

    Not proven: too few cases

    No model trained yet: the first models need 30 days of recorded outcomes.

    No holdout yet. The first models need 30 days of recorded outcomes.

33,569 entries in the forecast register. A probability measured on past coins, not a recommendation and not a price target. Information only.