Published
Binance publishes two long/short account ratios for its perpetual pairs: a "global" ratio across all accounts (a proxy for retail), and a "top trader" ratio limited to its largest position holders (a rough proxy for more sophisticated money). Our live dashboard has been quietly pulling both every 30 minutes since early September. I took the 2026-09-26 23:00 UTC snapshot across the 24 pairs we track and lined the two ratios up side by side, expecting the usual mixed picture. Instead: 23 of the 24 pairs have both groups net long, and in most of those, the top traders are more long-biased than the retail crowd, not less.
The full snapshot, 24 pairs
Sorted by retail (global) long/short ratio, highest first. A ratio of 2.00 means twice as many long accounts as short accounts on that pair.
| Coin | Retail ratio | Top-trader ratio | 7d price |
|---|---|---|---|
| ETH | 2.73 | 1.68 | +1.95% |
| LTC | 2.55 | 3.96 | +24.49% |
| DOGE | 2.50 | 4.83 | +8.53% |
| ADA | 2.44 | 2.21 | +10.27% |
| AVAX | 2.35 | 2.89 | +10.15% |
| XRP | 2.33 | 2.75 | +6.73% |
| SUI | 2.30 | 2.37 | +32.50% |
| BNB | 2.04 | 1.43 | +1.35% |
| PEPE | 2.02 | 2.23 | +3.09% |
| DOT | 1.98 | 1.99 | +10.63% |
| APT | 1.84 | 2.78 | +16.06% |
| LINK | 1.80 | 1.67 | +12.67% |
| BCH | 1.78 | 2.34 | +32.53% |
| SHIB | 1.73 | 1.77 | +6.37% |
| POL | 1.61 | 1.79 | +13.52% |
| NEAR | 1.55 | 3.12 | +42.80% |
| UNI | 1.50 | 2.94 | +13.45% |
| SOL | 1.49 | 2.29 | +9.20% |
| XLM | 1.45 | 2.32 | +9.35% |
| HYPE | 1.40 | 1.73 | +0.62% |
| ATOM | 1.32 | 3.16 | +7.53% |
| ARB | 1.30 | 1.69 | +9.43% |
| BTC | 1.29 | 1.90 | +3.74% |
| TRX | 1.06 | 0.68 | -1.36% |
Averaged across all 24 pairs, the retail ratio sits at 1.85 (about 65% long / 35% short) and the top-trader ratio at 2.36 (about 70% long / 30% short) β top traders are, on average, more one-sided than the crowd they supposedly trade against.
23 of 24, and it's not close
Every pair above TRX has both columns above 1.00. That's 23 of 24 tracked pairs where retail accounts and top-trader accounts are pointed the same direction, at the same time, on the same day. Nine of the 24 have a retail ratio at 2:1 or worse. Four β DOGE, LTC, ATOM, NEAR β have a top-trader ratio at 3:1 or worse, meaning fewer than 1 in 4 of the largest tracked accounts on those pairs are short.
Smart money is leading the crowd, not fading it
The assumption I'd have made going in: top traders fade retail excess β when the crowd pulls too long, the "smart" side leans the other way. That's not what the data shows. In 19 of the 24 pairs, the top-trader ratio is higher than the retail ratio β DOGE's top traders are at 4.83:1 against a retail crowd already at 2.50:1; LTC's top traders sit at 3.96:1 against a 2.55:1 retail crowd. Both groups also sit on top of a market that's mostly been rallying β every pair except TRX shows a positive 7-day return, several (SUI +32.5%, NEAR +42.8%, BCH +32.5%) sharply so. Read plainly: the recent rally, the retail long bias, and the top-trader long bias are the same trade, stacked three times over, on almost every pair we track.
The one pair that splits: TRX
TRX is the single exception, and it's a clean one. Retail is barely long at 1.06 (52% long / 48% short β close enough to call it a coin flip). Top traders, meanwhile, sit at 0.68 β net short, 40% long / 60% short. It's also the only one of the 24 pairs still negative over 7 days, at -1.36%, while the other 23 average a double-digit gain. Correlation isn't causation from a single pair, but the pattern is at least internally consistent: the one coin where the largest accounts are betting against the crowd is also the one coin that hasn't joined the rally.
Why the lopsidedness matters more than the direction
None of this says the rally is over or that being long is wrong β 23 pairs being long together doesn't make them wrong together. What it changes is what a single stop-out is worth. When close to 70% of tracked open interest on a pair sits on one side, a move against that side doesn't just close those positions β the forced selling from the first tier of liquidations pushes price further, which triggers the next, non-linearly. That's the mechanic behind moves like the one covered in ETH's $255M short-liquidation squeeze β same cascade, opposite side. Our own liquidation heatmap is built to show exactly where those clusters sit by price level. And if you're holding several of the 23 long-crowded pairs at once expecting diversification, the math in "your five diversified longs are really one bet" applies directly β five positions on pairs that are all long-crowded aren't five uncorrelated bets, they're one leveraged bet on the crowd staying right, sized five times over.
What I'm checking before sizing anything right now
Leverage and volatility were already my first two checks via the time to liquidation calculator. This data adds a third: how crowded is this specific pair, on both the retail and top-trader side, right now β not just whether I think the direction is right. A position on TRX and the identical position on DOGE aren't the same trade at the same leverage this week; one sits against a near-even crowd, the other sits inside a trade nearly 5 out of 6 large accounts have already made. I'm checking the live ratio on our dashboard before opening anything above 10x, the same way I now check funding rate per pair rather than assuming the median.