A backtest tests the idea. Paper trading tests the idea plus you plus the clock. They answer different questions, and doing them in the wrong order wastes months.
Last verified 13 September 2026 · written by the DexLadder team
What a backtest proves
That a fixed rule set would have produced a given equity curve over a given history, assuming your cost model is honest. That is a real and useful claim — it is the only cheap way to see a strategy through a bear market, a squeeze and a chop phase in an afternoon.
The four ways a backtest lies
Look-ahead bias. Using the close of a bar to make a decision inside that bar. The commonest and most flattering error there is.
Survivorship. Testing on the assets that still exist today, which quietly removes every project that went to zero.
Optimisation. Tuning parameters until the curve is beautiful. The curve is then a description of the past, not a hypothesis about the future.
Cost fiction. Ignoring fees, funding, spread and slippage. Add a realistic cost stack and many high-frequency systems invert.
What paper trading proves
That you can execute the rules in real time, at real speed, against live spread and depth — including the parts a backtest cannot represent: hesitating on an entry, widening a stop, doubling down after a loss. It also surfaces mechanical problems a historical test never sees, like the fact that your signal appears at a time when the pair is thin.
What neither one proves
Neither answers whether you will follow the plan when the money is real. Simulation removes the sensation that causes most rule-breaking, so treat flawless paper results as necessary and nowhere near sufficient.
The order that works
Write the rules down, completely, before touching data.
Backtest on a sample, with honest costs. Hold out a period the model never sees.
If it survives, paper trade it live for a minimum number of trades, not a minimum number of days.
Compare realised fills against the backtest's assumed fills. The gap between them is your execution cost, and it is the number most people never measure.
Go live at a size where losses are irrelevant, and scale only when behaviour matches the paper record.
DexLadder runs both halves on one device: Rewind for bar-by-bar historical replay, and the live paper desk for execution against the current market, using the same fee and impact model in each so the two are directly comparable.
How many trades before a paper record means anything?
Enough that one outlier cannot dominate — commonly fifty to a hundred, across more than one market condition. Days elapsed is the wrong unit.
Can I skip backtesting and just paper trade?
You can, but paper trading collects data at the speed of real time. A backtest sees years in minutes, which is the only practical way to test a strategy against conditions that are not happening right now.
Why do my live results differ from my backtest?
Almost always execution: assumed fills at mid price, ignored fees and funding, and slippage that scales with your size. Measure the gap rather than blaming the strategy.