The backtest engine walks your strategy through past market data and reports what it would have done with it: every entry, every exit, and every branch it never took. It was rebuilt from scratch for the reboot, and it runs on your machine like everything else.
- Engine
- Rebuilt from scratch
- Runs on
- Your machine
- Modes
- Single and multi
- Input
- The strategy file
- Output
- Order-by-order log
- Cost
- None
| Strategy | Pair | Period | Status | Progress |
|---|---|---|---|---|
| Odayakawa | BTC/USDT | May 27, 2026 → Aug 27, 2026 | Done | — |
| dca-bearish-ema200 | ETH/USDT | Jan 1, 2026 → Aug 27, 2026 | Running | |
| Supertrend futures | SOL/USDT | May 27, 2026 → Aug 27, 2026 | Running | |
| Bollinger squeeze | BNB/USDT | Jun 1, 2026 → Aug 27, 2026 | Running | |
| MA crossover 12h | XRP/USDT | Feb 14, 2026 → Aug 27, 2026 | Running | |
| Ichimoku swing | ADA/USDT | May 27, 2026 → Aug 27, 2026 | Queued | |
| RSI reversal | LINK/USDT | May 27, 2026 → Aug 27, 2026 | Queued | |
| Grid AVAX | AVAX/USDT | — | ReadyNeeds config | — |
| Smart DCA | DOGE/USDT | Jan 1, 2026 → Aug 27, 2026 | Ready | — |
What a run gives you back
A backtest is not a score. It is a record of decisions, which is the only thing you can actually learn from.
Run a queue, not a run
One strategy against three pairs, or three variants of the same idea, or the same strategy across four periods. The runner takes a list and works through it, running some in parallel and queueing the rest, so you read the results side by side instead of one at a time.
- ready-to-runConfigured, waiting for you to start it.
- queuedAccepted, waiting for a slot.
- runningWalking through the data now.
- doneFinished. The report is on disk.
- errorStopped, and it says why.
The five states the runner can be in. There is no sixth.
What a backtest is not
This section exists because the alternative is letting a nice-looking curve do your thinking for you.
- It is not a forecastA backtest tells you how a set of rules behaved in a market that already happened. It says nothing about the next one. Past results, backtested or live, do not predict future performance.
- It is not the live marketA replay fills orders against historical candles. Real fills happen against a real book, with real spread, real latency and real slippage. The gap between the two is smallest on liquid pairs and large orders make it worse.
- It is not proof, and it can be fittedTune the parameters until the curve looks good and you have described the past, not found a rule. A strategy that only works on the window you tuned it on is the normal outcome, not the unlucky one.
- It is not a substitute for trading smallThe honest order is: replay it, then run it on money you can lose, then decide. Skipping the middle step is where the expensive lessons live.
The step between the replay and the money
Paper trading runs the strategy against the live feed, in real time, with no funds behind it. It is the only way to see how a strategy behaves against a market that has not happened yet, and it is still in development.
Where it sits on the roadmapBefore and after the replay
- KryllOS SystemThe desktop itself: windows, apps, and a store to install them from.
- Drag-and-Drop EditorCompose a strategy from blocks, wire them up, read the result out loud.
- MarketplaceInstall what the community published, or publish and get paid.Soon
- Kryll AIDescribe a strategy, get blocks. Then have them audited.Soon
- KRL TokenThe currency of the store, and where a share of every sale goes.
Replay it first.
The backtest engine ships with KryllOS and runs on your own hardware. There is no per-run cost, because there is no server to pay for.
Trading crypto-assets carries a risk of losing your capital.