---
name: strong-coins
description: What Ben means by "strong coins" (relative strength vs BTC); read before any strongest-coins ranking or BTC-relative screen.
sources: [cowork]
aliases: [strong coins, strongest coins, relative strength vs btc]
---
- [stated] "Strong coin" means: goes up more than BTC when BTC goes up, goes down less than BTC when BTC goes down, or even pumps when BTC dumps or goes sideways.
- [stated] For strong-coin rankings he cares only about current strength, not whether a coin was strong a month ago.
- [stated] Wants the screen run on RH chain and Solana coins (reuse the prior multi-chain ~700 coin analysis where it makes sense).

## Pullback relative strength test (2 Oct 2026, cowork)
- [stated] Ben's thesis 2 Oct 2026: coins that pull back least and bounce hardest vs the market in market-wide pullbacks will do best over the next weeks and months; buy the hardest bouncers.
- 2026-10-02 | cowork: pullback RS | If you buy the coins that fell least (market peak to BTC low) and bounced hardest (BTC low to 7+ days later) in a BTC pullback, then they LAG the rest over 30 to 90 days | receipt: 659 Binance USDT coins incl 254 delisted, daily 2017-08 to 2026-10, 55 BTC pullbacks (-15% from 60d high, confirmed +10% off low); top 20% vs bottom 20%, median excess vs alt median: 30d -2.3 pts, 60d -7.6, 90d -10.1; strong group won only 16 of 46 events at 60d; 36 of 36 parameter settings (BTC dip 10 to 25%, confirm 5 to 15%, bounce window 3 to 14d) negative at 60d; both halves (2017-21, 2022-26) and bull/bear negative | KILLED
- Why (mechanism): "fell little + bounced hard" mostly picks low-beta or own-news coins. When the market recovers after the dip, weak high-beta coins win 18 of 21 times (60d spread -19 pts); when the market keeps falling, strong coins hold up only +1.4 pts better (13 of 25, coin flip). Less upside, no real protection.
- Strength repeats (rank corr 0.21 between consecutive pullbacks, 33 of 39) but "strong twice" lagged "weak twice" by 12 pts at 60d. Strong AND already trending (top 90d momentum) also lagged (-8.9 pts at 60d). Top 10 equal weight portfolio: 60d mean +1.9% vs all alts +12.3% vs BTC +5.5%.
- Memes only (Binance-listed memes, 14 to 21 events): no edge either way (spread -0.5 to +2.0, about coin flip). RH/Solana trenches NOT tested (Mac offline, about 3 months of history).
- Data: /home/claude/rs in Cowork session 01NpV3kNB5hd1XHTpouSXdZP (temporary container): d/ raw daily candles, close.pkl, ev.py, ev2.py, an.py, rob.py, deep.py, port.py, meme.py, pers.py. Source: data-api.binance.vision (live) + data.binance.vision S3 monthly zips (delisted), both reachable from the container. Glitch coins dropped (redenomination jumps): COCOS, DREP, LUNA, QUICK, SUN.
- 2026-10-02 | cowork: what predicts outperformance after a dump | Same 55 dips, 15 features, rank corr per event. Help: OLDER listing (180d IC +0.16, t 7.8, 31 of 34 events; listed <1y -5.6 pts, 4y+ +5.2 pts). Hurt: high 90d volatility (IC -0.10), volume spike during the dump (IC -0.07 to -0.10, dump-specific: only -0.04 on random days), hard bounce (IC -0.05), high trading volume (IC -0.06 to -0.09). Nothing: beta, 30d pre-dip momentum, how much of the dip it reclaimed by 180d | receipt: Binance 659 coins 2017-2026 | TESTING
- 2026-10-02 | cowork: dump combo score (old + calm + no dump-volume spike + weak bounce) | Beats other alts out of sample 2022-26: 20 of 23 dips at 90d (top 10 median -0.4% vs all alts -10.3%) BUT loses to BTC (BTC +6.5%, top 10 beat BTC 6 of 23). Works on random days too (78 of 98), so it is a general quality filter, not a dump signal | receipt: signs fixed on 2017-21, tested on 2022-26 | TESTING
- Plain rule from the test: in a dump, BTC beat almost every alt since 2022; if buying alts, pick old, calm coins that did not spike in volume during the dump. Not tested on RH/Solana trench memes.