Strategy Studio
Compose a perps swarm from real edges, give it an identity, and backtest it on live data.
Maps your words onto the real edge palette below — then tune the sliders. Grounded in evidence, not hype.
Long spot, short the same size in perps. Earn the funding, no price bet.
Steadiest edge, small drawdowns. Needs both legs and a solvent venue.
Tilt the way funding pays: short when longs pay, long when shorts pay.
Good income when calm; thins out when crowded, flips in a selloff.
Buy strength, sell weakness, and ride the move.
Only pays in real trends. Expect flat spells and sharp givebacks.
Hold the strongest coins in the watchlist, drop the laggards.
A real factor, but thin alts, slippage, and crashes on reversals.
Two coins usually move together; when they split, bet they reconverge.
Clean on paper, but crypto links snap without warning. Use a hard stop.
Enter as price breaks a recent high or low, betting it keeps running.
Most breakouts fail: many small losses, a few big wins. Stop required.
Buy sharp dips, sell sharp spikes inside a range.
Wins often, but the rare loss is brutal. Range-only, never against a trend.
Bet against stretched moves and funding extremes.
Thin evidence; fade too early and you're liquidated with the crowd.
Trade only when the 2-day vs 7-day trend is clear: |trendZ| ≥ 1.0. The same regime signal the live swarm runs on.
Same edge, three risk dials. Bands come from backtests: ranges, never promises.
Set the edge mix and universe, then Run against 120 days of real 1h candles. You get the full risk metrics and a buy-and-hold benchmark.
Bought on your curve at launch and locked until graduation — you can't dump on buyers. Real skin in the game.
Launch mints your swarm token + bonding curve on-chain (Base Sepolia) and starts its live engine on your DNA. Buyers speculate in $AGT; your on-chain track record (AGT Score) decides who wins.
- •Diversify across edges that don't move together, not just across coins. Carry, trend, reversion and stat-arb are driven by different things, and that is where real diversification comes from.
- •Volatility-target the whole book: trim exposure when the market gets choppy, add it back when it calms. In the research this lifted Sharpe from about 1.44 to 1.74 and shrank drawdowns (Moreira-Muir, JF 2017).
- •Bet a fraction of what the math says is optimal, never the full amount. About half keeps most of the growth at roughly half the swings, while the full bet invites 50% drawdowns. A quarter to a half is the sweet spot.
- •When you blend several edges, giving each an equal share of the risk usually beats clever optimization once you are out of sample. The fancy version looks better on paper, but estimation error eats the gain (DeMiguel-Garlappi-Uppal, RFS 2009).
- •Assume the comforting low correlations from your backtest all jump toward 1 in a crisis. Carry, reversion and fade tend to bleed together in the same liquidation cascade, so stress-test on the crash periods, not the calm average.
- •Size each venue as if the exchange could go to zero, because FTX did. Counterparty risk is the real danger for a delta-neutral book.
- •Treat every backtest as overfit until it is proven live. Judge a swarm by its on-chain track record (the AGT Score) and a deflated Sharpe, not by this preview.