A score means nothing if the simulation behind it flatters the strategy. So before we talk about what a good grade is, it's worth knowing how carefully the grade is earned — and the shortcut we refused to take.
Most backtests grade a strategy on four numbers per candle: where it opened, its high, its low, and where it closed. Fast, and good enough for a rough sketch. But those four points hide the one thing that decides a trade's fate — the order things happened in.
Say a candle touched both your stop-loss and your take-profit. Which came first? The four points can't tell you — the high and the low are just two dots, with no clock attached. A lazy engine gets to assume, and it will quietly assume whichever one flatters the result. Do that ten thousand times and you've manufactured a beautiful strategy that never existed.
So the engine doesn't work from the four points. It reconstructs what actually happened inside each candle, in the real sequence it happened — and settles the trade on whichever event genuinely came first. When a stop and a target both sit inside one candle, there's no coin-flip and no convenient guess: it takes the one that truly triggered first.
That single choice is the difference between a grade you can lean on and a grade that's quietly lying to you. It's slower and it's fussier — and it's the whole reason a score here means what it says.
There's a subtler way a backtest can cheat itself: acting on information it wouldn't have had yet. When a strategy's signal fires on a candle, it's tempting to pretend you bought at that same candle's close — a price you only know because the candle already finished.
The engine won't. A signal on one candle fills at the next candle's open — the first price you could realistically have acted on. It's a small rule with a big effect: it closes the door on hindsight, so the strategy is only ever rewarded for what it could actually have done in real time.
A return that ignores costs is a fantasy. Every simulated trade here pays what a real one would — the exchange's cut, the slippage of getting filled at a slightly worse price than you hoped, and the periodic funding fee that perpetual futures charge for holding a position. All of it is subtracted inside the run, before any number reaches your screen.
And a position still hanging open when the test ends doesn't get to count a paper profit — an unclosed trade is an unfinished one, left out until it actually closes. The score you see is what's left after the bills, not before.
None of this makes a strategy good. It makes the grade honest — which is the part you can't check by eye and therefore have to trust. An impressive number from a sloppy simulation is worse than useless, because it earns your confidence without deserving it.
That's the bar we hold: the score is measured the hard, slow, unflattering way, so that when a card impresses you, it impressed the toughest examiner in the room first. What that grade can't promise is the future — for how we guard against a strategy that merely memorized its past, see The Four Gates.
An honest simulation makes a grade trustworthy — it doesn't make the outcome certain. Every result here is hypothetical performance on past data, net of costs, and nothing on CoinLAB is financial advice. A strategy graded fairly can still lose in live markets. Always do your own research.