We Test the Folklore Against the Record
Crypto runs on folklore: extreme fear is a buy, high funding means a flush, altseason follows dominance breaks. We have the recorded history to check, so we do, in public. Each note states the question, the data, the method and the result. When the answer is "no edge", we publish that too; honest negatives are rarer and more useful than hopeful positives.
01
Research notes
Bitget withdrawals, 24 Sep 2026: the switches, not the rumourRecorded
657 of 4,930 asset and network pairs disabled at 20:40 UTC, three more than at 04:50. BTC, USDT, USDC, SOL open. ETH on Ethereum mainnet the one unannounced change. Prices, depth and open interest in line with peers.
Read the note →001 · Do extreme readings predict the next week?rec · recomputed daily
Five gauges, eight years of records, one narrow question. Mostly no, with two honest exceptions. Live tables.
Read the note →02
The rules every note follows
- The question is stated before the data is looked at, and the note answers that question, not a prettier one.
- Only recorded, dated series are used: the same daily archive the rest of this site runs on, never reconstructed history.
- Base rates always appear next to conditional results, and buckets thinner than 30 days are suppressed.
- Negative results are published with the same prominence as positive ones.
- Every note links the dataset and the exact API calls needed to reproduce it.