Why the Menu Is Short

Most platforms compete on how much they let you pick — hundreds of coins, every timeframe from one minute up, leverage to the moon. It reads as freedom. But a backtest is only as honest as the market underneath it, and on most of those choices the simulation is quietly guessing.

We took the opposite bet. The Lab only puts a combination on the menu if it can grade it the hard, honest way — real costs, real fills, no flattering assumptions. So what's missing from the menu isn't missing by accident. Every choice we left off is one we'd have had to fudge to include. A short menu isn't a smaller product; it's a promise about the ones that made the cut.

BTC, ETH, SOL, XRP. Two reasons, both about honesty.

Deep enough to simulate truthfully. The engine subtracts real costs — fees, slippage, and funding — from every trade. Those costs are only knowable on a liquid market. On a thin altcoin, your own order moves the price, fills are unpredictable, and a backtest that pretends otherwise is fiction dressed as data. These four are among the most liquid perpetuals in existence, so the slippage and fills the engine models are the ones you would actually meet.

Long enough, and permanent enough. A fair grade needs history across every kind of market — bull, crash, and the long flat stretches where nothing works (that's why the window is a fixed two years). The four carry years of continuous history through all of it, and they're the least likely to delist out from under a strategy. And they aren't four copies of one chart: a blue-chip anchor, a second major, a fast high-beta mover, and an idiosyncratic one — a strategy tested across them has met genuinely different temperaments.

The honest trade is depth over breadth: we'd rather breed real champions on four markets we can simulate truthfully than hand you a thousand strategies on coins where the number is a guess.

There's a speed band where a backtest can keep its promise. These three sit inside it — and the edges are cut off for concrete reasons.

Too fast, and the simulation breaks. Below a few minutes, an outcome is decided by microstructure — the exact spread, the queue you're in, the millisecond of your fill — none of which a candle-and-tick replay can honestly reproduce, and all of which real costs quietly eat. A one-minute "edge" is usually just a simulation being kinder than a real exchange. There's a simpler problem too: these are signals a human receives and acts on, over push or Telegram. A signal you'd have to fill within seconds is one you can't actually use.

Too slow, and there's nothing to prove. Stretch to daily bars and a strategy fires so rarely that, even over a long window, it books only a handful of trades — far too few for the gates and the Monte-Carlo shuffle to certify it as anything but luck. Validation needs a real sample.

So the band. 15m, 30m, and 60m are fast enough to gather a trustworthy number of trades, and slow enough that the costs, the fills, and you can all keep up. Three rungs — active, middle, patient — to match your temperament, without fragmenting into noise.

This one deserves the longest answer, because the obvious objection is fair: if every strategy ran at 1×, couldn't you just add your own 3×, 5×, even 100× at the exchange? You could. It simply wouldn't give you what the 3× and 5× cards give you — and here's the exact reason.

Leverage is not a multiplier you bolt onto a finished strategy. It changes which trades survive. Take a 1× trade that fell 30% underwater before recovering to close +10% — at 1×, a winner you calmly held. But a 5× position is liquidated once the coin moves only about a fifth against it, so that same 30% dip doesn't cost you 30% — it ends the trade at the bottom. You're wiped out and never see the recovery. Same entries, same exits, opposite outcome. "The 1× strategy at 5×" isn't the 5× strategy at all; it's a strategy that gets stopped out of existence on the very drawdowns the 1× version rode straight through.

The engine knows this, because it checks — second by second, inside every candle — whether a position at that leverage would have been liquidated before the trade ever reached its target. It also charges the real cost of the size: fees and funding are paid on the full leveraged position, so a 5× trade carries several times the cost drag a 1× backtest ever subtracted. Multiply a 1× return by five and you've ignored both the liquidations and the costs — the number is a fantasy.

That's why the Lab breeds a separate champion for each leverage. The 5× card isn't the 1× strategy turned up loud — it's a strategy that had to survive 5× liquidation and pay 5× costs to earn its grade, usually with its own, more careful rules. The 3× and 5× cards are the strategies that are genuinely best at that leverage, with the danger already priced in. That is the thing you cannot reproduce by turning a dial on a 1× card.

"Then why stop at 5× — why not 10×, 100×?" The same honesty test. Push leverage high enough on these timeframes and ordinary intrabar noise liquidates almost everything before any edge can play out — no strategy survives the engine's liquidation check, so there would be nothing honest to sell you. A 100× card would only "work" in a simulation that looked away at the wrong moment, and we won't build that card. The ceiling isn't us deciding how much risk you're allowed to take at your own exchange — it's the point past which we can no longer prove the strategy is real.