Backtest a signal rule
Define criteria, pick a horizon, and see what historically happened after every match — sample size, hit rate, drawdowns and the full distribution, using only data available at each moment.
Backtests are a Pro feature
Pro runs point-in-time backtests over the full history with distributions, drawdowns and leakage checks. Sign in to get started — browsing stays free.
Event studies stay open on every plan: they answer fixed questions like “what happens after thesis velocity triples?” with the same point-in-time method. Open event studies
Event studies
Pre-computed questions over the full point-in-time history, refreshed hourly. Free on every plan.
Event: Thesis velocity > 3× baseline
What happens after thesis velocity triples?
- Samples
- 43
- Median 1h
- 0.0%
- Median 24h
- +0.2%
- Median max DD 24h
- -17.4%
- Positive after 24h
- 51%
Event: 5+ high-quality authors in 24h
What happens after 5 high-quality authors converge?
- Samples
- 0
- Median 1h
- —
- Median 24h
- —
- Median max DD 24h
- —
- Positive after 24h
- —
Event: Social rising, price flat
What happens when social activity rises but price does not?
- Samples
- 95
- Median 1h
- -0.7%
- Median 24h
- -2.3%
- Median max DD 24h
- -12.6%
- Positive after 24h
- 42%
How backtests avoid look-ahead
Strict time-series testing: no look-ahead, dead tokens included, sample size always shown.
- 01
Point-in-time snapshots
Each candidate moment is a stored score + feature snapshot computed only from theses, prices and author records available at that time.
- 02
Entry and exit
Entry is the first price at or after the event (within 15 min); the outcome is the last price at or before event + horizon.
- 03
Drawdowns, not just returns
The deepest drop and the peak inside the horizon are measured for every event, and shown as prominently as returns.
- 04
Survivorship
Tokens that later died stay in the sample; you choose how a death inside the horizon is valued.
- 05
Sample size first
N < 10 is insufficient, < 30 low, < 100 medium confidence. Medians, trimmed means and full distributions — never one average.
FomoQuant provides analytics and historical/statistical context, not financial advice or guaranteed predictions.