A bidirectional, non-directional strategy that places concentric grids around exchange-aggregated liquidation zones. Up to five concurrent grids across ~50 scanned perpetuals, scored every 30 minutes on seven dimensions, rotated continuously toward the highest-conviction setups.
Unlike spot markets, perpetual futures expose a continuously updating ledger of leveraged positions whose liquidation prices are publicly inferable from open-interest data. When price approaches a dense cluster, forced flow compresses price toward that cluster, producing a brief window of mean-reverting behaviour. Most retail systems are directional momentum strategies trying to predict the next leg. Farms is structural — it monetises the oscillation around well-defined liquidation anchors without taking a view on direction.
Up to five concurrent assets, each running concentric grids spanning long max-pain to short max-pain. Long-side and short-side fills both monetise the same structural anchor — direction-agnostic by construction.
Every 30-minute refresh, the rotation engine evaluates each active grid against the current farmability cross-section. Grids that fall out of the top five for two hours rotate out; replacements that pass the entry gate rotate in. Capital follows opportunity.
Position-level hard stops 1.5% beyond the lowest grid level with ATR trailing on the seed. Asset-level breakout detection pauses redeployment for 4 hours after a confirmed breakout. Portfolio-level circuit breakers at 12% peak drawdown and 5% daily loss, both with auto-resume.
The current production parameters are grounded in a 32-day out-of-sample observation window covering 10 April through 12 May 2026. The scanner produced 77,734 zone snapshots across 50 symbols at consistent 30-minute cadence. What that data says, plainly:
Across all firing-ready entries during the window, the median forward return was positive at every horizon — +0.13% at 6h, +0.46% at 24h, +0.66% at 48h, +1.13% at 7d. Sixty-two percent of entries produced positive 7-day returns.
Of the five concurrent grid slots, expected occupied capacity over the window was approximately 1.0 of 5 (20%). Half of all snapshots showed zero firing-ready candidates. The strategy waits patiently for favourable conditions rather than deploying capital continuously.
A subset of assets (BTC, TRX, BNB, LTC, BCH, PAXG) exhibits clean mean-reversion with low stop-hit rates. A second subset (NEAR, XRP, WLD, HYPE, ENA) shows high forward returns on average but high realised volatility around those returns. A universe prune is scheduled in the next strategy revision.
Capital is idle roughly 80% of the time in the observed window — the entry gate rejects setups whose price has drifted too far from the anchor. If you expect every dollar to be at work continuously, this strategy is the wrong shape.
Mean-reversion edge is conditional on price oscillating around the anchor. Assets in persistent one-direction trends do not retest the anchor, and grids deployed on them either rotate out at minor losses or get stopped on breakouts. Trending regimes are the structural drag.
Past performance does not predict future returns. The 32-day observation window covered a specific market state and does not guarantee replication. The strategy depends on liquidation-zone topology remaining a meaningful anchor — market structure changes can degrade the edge.
Farms trades leveraged perpetual futures on Hyperliquid. The portfolio drawdown halt at 12% reduces tail risk but does not eliminate it. The full deposit is at risk of loss.
Deposit USDC into the Hyperliquid vault. The strategy trades for you. Withdraw any time — fees only on profit.
Same liquidation-zone thesis, different execution. Zones DCAs into dislocations across graduated levels and holds for mean-reversion. Live since March.
18-coin TradingView + Python hybrid on Hyperliquid. Mean-reversion at the exhaustion of liquidation cascades, with per-coin tuning.
$7/mo signals service for the buy-the-dip retail investor. No leverage, execute on your own broker.