Advanced Strategies and Review · Crypto Trading Guide
A convincing historical chart can be a poor test. Keep future information out, count failed experiments and simulate the conditions you could actually trade.
Write the rule before testing
Specify the asset universe, data frequency, signal timing, order logic, sizing and exit rules. If a signal uses a candle’s close, do not assume execution at that same close without a defensible mechanism. Future data, revised labels or a token list containing only survivors can make the past look easier than it was.
Separate development from evaluation
Use chronological development and validation periods, then reserve an untouched final test where appropriate. Record every parameter combination tried. Selecting the best of hundreds of variations can fit noise. Repeatedly inspecting and adjusting to a supposed test set turns it into another development set. A later forward simulation provides additional evidence but no guarantee.
Model what could fill
Include fees, bid and ask differences, funding, borrowing and realistic delays. A candle touching a limit does not prove your order filled. If a bar contains both stop and target prices, the sequence may be unknown. Use finer data or a conservative stated assumption. Retain delisted assets and periods of outages when they belong to the strategy’s universe.
Report more than the headline return
Imagine a fictional test produces $1,200 gross profit across 200 trades. If omitted costs average $7 per trade, the result becomes a $200 loss. Examine drawdowns, tail losses, turnover, capacity and sensitivity to modest parameter changes. Compare with a simple benchmark and explain how cash flows were handled. A strategy that works only at one exact setting deserves scepticism. Preserve the data version and test log so another reviewer can reproduce the result.
Try it yourself
You select the best strategy after viewing the final test period, then report that period as unseen evidence. What is wrong?
Show the worked answer
The period influenced selection and is no longer untouched. A fresh evaluation is needed, and the number of attempted configurations should be disclosed.
Further reading
Educational material. Examples and rates are hypothetical, not trade recommendations. Product terms, eligibility and risks vary by venue and jurisdiction.
