How We Calculate Flip Probability
When indicators diverge from the current regime, the dashboard shows a historical probability that the regime will change within the next 20 trading days. This is not a model prediction — it is a direct look-up from 30 years of data.
How It Works
We scan every day in our dataset (1996–2026) and find all days where the same regime was active and the same set of indicators were diverging. We then count what percentage of those days were followed by a regime flip within 20 trading days. That percentage is the flip probability shown on the dashboard.
It only appears when at least 10 historical matches exist. Rare combinations with fewer instances are excluded — showing a probability based on 2 or 3 occurrences would be meaningless.
What Counts as Diverging
An indicator is diverging only when it actively signals the opposite direction to the current regime. Neutral readings (zero) are excluded — they carry no information either way.
VIX and AUD/JPY are contrarian indicators. A rising VIX signals risk-off (not risk-on), and a falling AUD/JPY signals risk-off. Divergence is always assessed against these effective signals, not the raw values.
Time Windows
The dashboard shows probabilities across four windows: 5, 10, 20, and 30 trading days. The 20-day window is the headline — it offers the best balance between catching genuine flips early and avoiding false alarms.
5 days
High precision, low recall. Catches only imminent flips.
10 days
Intermediate. Starts to separate signal from noise.
20 days — headline
Best precision/recall balance across 30 years.
30 days
Highest recall, but also catches more false alarms.
VIX + AUD/JPY diverging in risk-on regime — flip probability by window (n=295)
What the Numbers Mean
A flip probability of 36% means that in 36% of historical cases with the same diverging indicators, the regime changed within 20 trading days. The base rate — across all days regardless of divergence — is roughly 33–35%. A reading above the base rate means the current indicator combination has historically added genuine predictive signal.
Example: VIX and AUD/JPY diverging in a risk-on regime — 295 historical matches, 36% resulted in a flip within 20 days. That is slightly above the base rate, meaning these two indicators together provide a modest but real lift over random chance.
Neutral Regime Periods
When the Alphameter score is in the neutral band, we inherit the last directional regime as context. The question being asked is always "what will the next directional regime be?" not "will it stay neutral?" — so a neutral period following a risk-on episode is treated as if it were still risk-on for the purposes of this calculation.
For educational and informational purposes only. Not financial or investment advice. Historical probabilities are not guarantees of future results.