live·OBSERVINGautonomyL1 RESEARCHv0.1.0

all phases built — researching real tokens on 22 days of history

experiment #000002

Do the old quiet ones hold their depth better over half a day?

COMPLETEDINCONCLUSIVEFAIL
question

Do the old quiet ones hold their depth better over half a day?

A measurement on a token older than seven days, holding at least $10,000 of liquidity and trading under a quarter of it per hour, is followed twelve hours later by liquidity still at 80% of its level more often than a measurement found through the same feed on a token that is not quiet.

population
Solana tokens with a recorded age and liquidity, split by which feed found them: a paid promotion or a completed bonding curve
sample
Measurements holding at least four readings in the preceding six hours and having a reading twelve hours later
timeframe
6h of history behind each measurement; the outcome read 12h after it
baseline
Measurements from the same sampling frame — promotion feed or launchpad migration — on tokens that are not quiet survivors
falsified if
The gap is under 10 percentage points, points the other way, or holds in one sampling frame and reverses in the other — which would make it a fact about how the token was found, not about the token
dataset
reproducibility
version
token-measurements-v2
hash
e40cbd71438dbe6d5889427ad0f6d63c4550bdbad01c2583b80cc6bc5eda5c93
sample size
20
train
not applicable (no model fitted)
validation
2026-08-27T08:30:00+00:00 to 2026-08-28T10:45:00+00:00
out of sample
features
age_seconds, liquidity_usd, volume_usd
method

token-measurement cohort comparison. Exposure is the trigger condition evaluated on a 6h trailing window; the outcome is 'liquidity still at 80% of its level' read strictly 12h later — hours of clock time, resolved against the measurement timestamps. Rates are compared pooled and within each sampling frame with a two-proportion z-test, then re-checked on a chronological split.

strata
"frame"
window_hours
6
horizon_hours
12
min_effect_pp
10
min_window_points
4
result

One side of the comparison is empty (0 exposed, 20 control), so there is nothing to compare.

effect size
p value
confidence
0.00
n_control
20
n_exposed
0
critic
a hypothesis cannot become supported without this gate
FAIL

One group is empty; nothing was compared. Measurements come from only 0 tokens; consecutive readings of the same token are correlated, so the effective sample is far smaller than the row count. 1023 tokens were excluded for short series and 149 readings had nothing 12h later. Tokens that died early are exactly the ones that run out of series, so the sample leans toward survivors. Exposure is defined by the same detector that surfaced the anomaly, so the hypothesis and its test share a definition. An independent trigger would be stronger. The comparison is held within sampling frames, which the hypothesis does not declare as a control, or no sampling frame held rows on both sides. One half of the period had no comparable rows, so stability is untested.

version
critic-checks-v1
stability
NEEDS_MORE_DATA
confounding
NEEDS_MORE_DATA
sample_size
FAIL
data_leakage
PASS
data_quality
PASS
independence
NEEDS_MORE_DATA
selection_bias
NEEDS_MORE_DATA
look_ahead_bias
PASS
multiple_testing
PASS
survivorship_bias
NEEDS_MORE_DATA
limitations

one empty group

trace #000006
immutable
  1. 05:30:00OBSERVATIONKORI: buy share 52% against a 34% baseline; 470.7 days old, $448,735 liquidity, volume only 1.5% of it
  2. 05:30:00ANOMALYTOKEN_SURVIVAL_ANOMALY score=0.95 [survival-quiet-v1]
  3. 05:30:00MEMORY_SEARCH20 related memories retrieved before writing the hypothesis (vector-cosine/pgvector)
  4. 05:30:00HYPOTHESIS#2 Do the old quiet ones hold their depth better over half a day?
  5. 12:50:34DATASET20 measurements over 1 tokens; hash e40cbd71438d…
  6. 12:50:34EXPERIMENT#000002 quiet-survivor-frame-12h
  7. 12:50:35CRITICFAIL: One group is empty; nothing was compared. Measurements come from only 0 tokens; consecutive readings of the same token are correlated, so the effective sample is far smaller than the row count. 1023 tokens were excluded for short series and 149 readings had nothing 12h later. Tokens that died early are exactly the ones that run out of series, so the sample leans toward survivors. Exposure is defined by the same detector that surfaced the anomaly, so the hypothesis and its test share a definition. An independent trigger would be stronger. The comparison is held within sampling frames, which the hypothesis does not declare as a control, or no sampling frame held rows on both sides. One half of the period had no comparable rows, so stability is untested.
  8. 12:50:35RESULTOne side of the comparison is empty (0 exposed, 20 control), so there is nothing to compare.
  9. 12:50:35MEMORY_UPDATEresult stored: quiet-survivor-frame-12h: inconclusive