I trade three coins, not fifteen
I trade three coins, not fifteen
Coinbase lists fifteen perpetual futures markets. Trader-7 trades three.
The instinct on seeing a longer menu is to expand. BTC, ETH, SOL, and then XRP, DOGE, LINK, AVAX, ADA, SUI, the lot. More coins means more setups, means more trades, means more chances for the system to find an edge. That is the obvious story.
It is also wrong, and the reason it's wrong is sitting in my P&L.
Why those three in the first place
I picked BTC, ETH, and SOL because of the Alpha Arena S1 results. Six frontier models traded ten coins each. The clearest signals showed up on those three. Deep liquidity, narrative density an LLM can actually reason about, intraday volatility wide enough for a 2.5:1 risk-to-reward target to fill. The other seven coins were noisier, and model performance was noisier with them.
That was the call. It's still the call. The question is whether I should add a fourth.
What SOL taught me about the cost of "more"
Since Sprint 120 (2026-03-21), SOL on Trader-7 has done 14 trades, 20% win rate, minus $126.08 P&L. BTC over the same window did fine. ETH did fine. SOL did not.
Sprint 125 cut SOL position size by 50%. The intent was bleed-control while I figured out what was specifically wrong. It didn't bleed-control. It just bled at half the rate.
Sprint 137 PR-C, shipped 2026-04-30, hard-excluded SOL. The system is back to two coins until I either understand what's broken about how the LLM picks SOL setups, or confirm the problem is structural and the exclusion stays permanent.
The lesson is not that SOL is bad. The lesson is that I added SOL with the same care I'd add any well-justified asset, and it still cost me real money for a month while I was busy looking somewhere else. Adding a fourth coin is not a free option.
What nobody tells you about altcoin perps
Most alts are not a different bet. They are a leveraged bet on BTC with extra symbol-specific noise.
If BTC dumps, AVAX dumps. ETH dumps, LINK dumps. The correlation matrix on a fifteen-minute timeframe is not subtle. When a "diversification" trade goes wrong, every position goes wrong together, and perp leverage means you eat the move four to five times over.
True decorrelation on this list is rare. XRP gets close. Its driver is a regulatory and payments narrative running on its own clock: ETF speculation, court rulings, banking partnerships. These don't track BTC's macro flow the way an L1 token's narrative does. That's why XRP is the one coin on the menu I'd seriously paper-test if I were adding anything.
The rest mostly aren't worth the slot.
The honest read on the rest of the menu
| Coin | Verdict | Why |
|---|---|---|
| XRP | Strongest candidate | Independent driver, deep liquidity, distinct catalysts |
| LINK | Second-tier | Real RWA narrative, but heavily ETH-correlated |
| AVAX | Skip | High BTC/ETH beta, no edge over what ETH already gives me |
| ADA / DOT | Skip | Slow intraday volatility, weak narrative flow for an LLM to read |
| BCH / LTC | Skip | Declining mindshare, dead-coin risk |
| DOGE | Skip | Sentiment-driven, the LLM can't predict the next Elon tweet |
| SHIB | Skip | Long-term downtrend drift plus retail noise |
| SUI | Maybe later | Volatile, independent Asia flow, but too young for the LLM to anchor on history |
| XLM / HBAR | Skip | XLM is XRP-lite, HBAR perp liquidity is thin |
If you read that and think it's just taste, fair, but the criteria are explicit. Independent drivers. Deep liquidity. Narrative an LLM can read. ATR profile that fits a 2:1 risk-to-reward floor. Most of the list fails at least one of those.
The fix isn't coverage
Trader-7's real problem right now is not "we're missing trades." For one of the three coins it already trades, the system was picking bad entries for a month. Adding a fourth coin while I'm still untangling that is curiosity, not strategy.
Order of operations: prove the BTC/ETH-only book is profitable for two to three weeks. Understand whether SOL exclusion stays or comes back. Then, if I genuinely want a fourth coin, paper-test XRP for thirty days before any live capital touches it.
Three is not a starting line. Three might be the right number.
If this was useful and you fancy chipping in for a coffee: buymeacoffee.com/jamiewatters.