Open Interest as a Conviction Gauge: Reading Net OI by Cohort on Hyperliquid

Aggregate open interest is one of the most-watched metrics in perp trading and one of the least-understood. Every analytics dashboard shows it, every commentator references it, and almost everyone is reading it wrong. "OI is rising" tells you that more positions are being opened. It doesn't tell you who's opening them, what direction they're leaning, or whether the marginal new position is informed conviction or chasing noise.

On Hyperliquid, you can decompose OI by behavioral cohort. Instead of one aggregate number, you get a breakdown: how much OI is Money Printer wallets, how much is Smart Money, how much is Exit Liquidity, and crucially — what direction each cohort is leaning. The aggregate hides the composition. The composition is the actual signal.

This article unpacks how to read net OI by cohort on Hyperliquid, what the four common cohort patterns mean, and how to combine cohort OI with price action to filter trades that are likely to succeed from trades that are likely to fail.

What aggregate OI actually measures

Open interest is the total notional value of all open positions on a market. Every long is matched by a short, so OI is the size of one side. Rising OI means new positions are being opened; falling OI means positions are being closed.

The usual interpretation: rising OI with rising price = bullish conviction (longs adding). Rising OI with falling price = bearish conviction (shorts adding). Falling OI with rising price = short covering (shorts capitulating). Falling OI with falling price = long capitulation.

This framework works on a CEX where you can't see who's opening positions. On Hyperliquid, you can — and the "conviction" interpretation falls apart fast once you look at composition.

Decomposing OI by cohort

The HyperTracker API exposes OI breakdowns by cohort: how much of the total OI on BTC, ETH, SOL, or any HIP-3 market is held by each of the 16 cohorts, broken down by long vs. short. A typical call:

oi = api.get("/oi/breakdown/BTC")
# Returns:
# {
#   "asset": "BTC",
#   "total_oi": 142_300_000,
#   "by_cohort": {
#     "money_printer": {"long": 28_400_000, "short": 12_100_000, "net": +16_300_000},
#     "smart_money":   {"long": 19_800_000, "short": 14_500_000, "net": +5_300_000},
#     "consistent_grinder": {"long": 8_200_000, "short": 7_900_000, "net": +300_000},
#     ...
#     "exit_liquidity": {"long": 4_500_000, "short": 11_200_000, "net": -6_700_000},
#     ...
#   },
#   "timestamp": "2026-06-06T14:30:00Z"
# }

The "net" field per cohort tells you how each behavioral class is leaning. The aggregate of all nets is by definition zero (every long is matched by a short), but the distribution of nets across cohorts is the signal.

Four cohort OI patterns

After looking at hundreds of OI snapshots across BTC, ETH, SOL, and a handful of HIP-3 markets, four recurring patterns explain most of the conviction signals worth acting on.

Pattern 1: Informed convergence

High-PnL cohorts (Money Printer + Smart Money) heavily net long; low-PnL cohorts (Exit Liquidity + Semi-Rekt) heavily net short. This is the convergence pattern: the wallets that have historically been right are positioned one direction, and the wallets that have historically been wrong are positioned the other.

The signal: strong continuation in the high-PnL direction is more likely than reversal. This pattern often shows up at the start of trends, before the move becomes obvious in price. By the time price has moved enough for retail to notice, the high-PnL cohorts are already positioned.

How to act: trade aligned with the high-PnL cohort consensus. Use full or expanded position size given the signal quality.

Pattern 2: Crowded losers

Low-PnL cohorts (Exit Liquidity + Semi-Rekt + Full Rekt) heavily net long; high-PnL cohorts neutral or net short. The unprofitable wallets are stacking the same side; the profitable wallets are not joining them.

The signal: mean reversion is more likely than continuation. Crowded loser positioning is fuel for a squeeze. The longer it sustains, the more violent the unwind tends to be.

How to act: contrarian short candidates. Tighter stops because squeezes happen fast; the trade either works in days or doesn't work at all.

Pattern 3: Cohort disagreement

Money Printer net long, Smart Money net short (or vice versa). The two high-PnL cohorts are explicitly disagreeing. This usually reflects time-horizon differences — Money Printer wallets often hold longer than Smart Money — but it can also reflect genuine analytical disagreement.

The signal: reduced confidence in either direction. The signal is ambiguous, which is itself information.

How to act: smaller positions if you take the trade at all. The market isn't giving you a clean read; act accordingly.

Pattern 4: Universal alignment

All cohorts on the same side. Money Printer through Giga-Rekt, all net long (or all net short). This is rare and usually indicates an extreme news catalyst or coordinated market move.

The signal: terminal phase of a trend. When even the unprofitable cohorts have piled into the right side, the marginal buyer is exhausted. The next move is usually a reversal.

How to act: fade the universal alignment. This is the most contrarian of all four patterns, and the highest-conviction reversal signal cohort data gives you.

Combining cohort OI with price action

The cohort patterns above are strongest when combined with price-action context. Same cohort distribution can have very different implications depending on what price has been doing:

Informed convergence + price consolidating: the strongest setup. High-PnL cohorts are positioning before the move. Trade aligned, expect a breakout.

Informed convergence + price trending in cohort direction: still good, but you may be late. Smaller positions, tighter management.

Crowded losers + price grinding against them: classic squeeze setup. Position contrarian to the losers, wait for the unwind.

Crowded losers + price trending in their direction: the rare case where losers are right. Wait — don't fade losers when they're being rewarded by price. Re-evaluate after price stalls.

Cohort disagreement + range-bound price: sit out. Both signals are ambiguous; no edge.

Universal alignment + parabolic price: the high-leverage fade setup. Tight stop, look for the first sign of reversal.

Why this beats aggregate OI

Most trading commentary that references OI is implicitly making a conviction argument: "rising OI = new conviction is coming in." But conviction is a statement about the traders, not about the positions. Whether new OI represents conviction depends entirely on who's opening it.

Aggregate OI conflates conviction from Money Printers with chasing from Exit Liquidity wallets. The two have opposite forward signals — Money Printer accumulation usually leads price; Exit Liquidity accumulation usually fades. Treating them as the same number (aggregate OI) throws away the directional information that makes the metric useful.

Cohort decomposition recovers that information. You're no longer reading "OI is up 8%." You're reading "Money Printer net long position grew $4M while Exit Liquidity net long grew $7M — the marginal new conviction is mostly retail chasing, not informed accumulation." The same headline OI move means very different things in those two scenarios.

API access

The cohort OI breakdown is exposed via /oi/breakdown/{asset} on any paid tier. Pulse ($179/mo) covers REST polling at 30s intervals — adequate for swing trades. WebSocket subscription on Surge ($499/mo) for sub-second updates if you need real-time cohort flip detection.

Get cohort-level OI data for Hyperliquid →

The bigger framing

The conviction interpretation of OI is intuitive but lossy. It works on data where you can't see composition. On Hyperliquid, you can — and the composition contains 80% of the signal that "rising OI" tries to communicate but can't.

Most traders are still reading aggregate OI like it's 2021. The traders who've upgraded to cohort decomposition are extracting a structural information advantage that compounds every time the aggregate metric tells a misleading story to everyone else.

Open interest tells you that positions are being opened. Cohort OI tells you whether they're the positions you should be on the same side of.