Accuracy is often the first number people look at, and the easiest to misread. A report can show a high hit rate with poor returns, or modest accuracy with better risk-adjusted results.
CryptoCrispy publishes signal accuracy and trade performance reports so users can inspect methodology, sample size, and coin-level differences before changing a strategy.
Directional accuracy is not the same as profit
Directional accuracy measures how often a prediction matched the later price direction. It does not measure how large the winning trades were, how large the losers were, or what fees did to the result.
ROI, when it is reported on a trade-by-trade basis, is closer to economic outcome. Even then, average ROI can hide a few large winners. Read both numbers together.
Start with sample size and window
Ten trades can look perfect and still mean nothing. A useful report states:
- How many predictions were included
- The date range and whether the window is rolling
- Which coins and timeframes were in scope
- Whether fees, quality filters, or exclusions applied
Compare a weekly snapshot with a longer window before you decide the model improved or broke. One quiet week is not a regime change.
Coin-level results beat a single average
A blended accuracy figure can be lifted by one liquid coin and dragged down by another. If you only trade a subset of the universe, read the rows that match your watchlist.
How to use CryptoCrispy reports in practice
- Open the weekly AI Assistant check and note the window, coin set, and prediction count.
- Compare that week with the coin-level insight summary so one liquid asset does not hide the rest.
- Read the methodology page so you know what “accuracy” means in that document.
- Only then decide whether to change alerts, size, or bot settings.
If the weekly window looks unusually strong or weak, open a longer snapshot such as the monthly assistant window before you treat it as a new regime.
Takeaways
- Accuracy answers “how often”, not “how much”.
- Ignore headline percentages until you know the sample.
- Use multi-period reports to avoid overfitting to one week.
Continue with how to use backtesting and performance reports or review the current CryptoCrispy reports library.
CryptoCrispy provides market analytics and trading tools. This article is general information, not personal financial advice. Past report figures are not a reliable guide to future results.