Methodology / FAQ
How does the SPY regime model work?
The model does not predict the future. It evaluates the current SPY market environment and checks whether the setup resembles historical phases that were followed by strong recoveries.
Core Logic
A regime score, not a crystal ball.
BottiAi calculates a daily regime score after the US market close. The score describes how strongly the current SPY setup resembles earlier market-bottom and recovery regimes.
A high score does not mean: Buy now. It means: The current environment resembles historical market phases where attractive entries later emerged.
The model was trained only on data before 2020-01-01. Everything from 2020 onward was not visible during training and is used as an out-of-sample test on unseen market data.
Weak, early, or failed signals remain visible. That transparency matters: this is not presented as a perfect trading system, but as a focused timing tool for research around longer-term ETF or stock buying decisions.
Plain English
- 01The model runs once per day after the US market close.
- 02It analyzes the current SPY market environment.
- 03It compares today's setup with prior market-bottom and recovery regimes.
- 04It outputs a score, not a buy recommendation.
- 05The track record remains publicly reviewable.
- 06False alarms and weak signals are also shown.
- 07No financial advice and no guarantee.
The basic idea: do not forecast where SPY will trade tomorrow. Instead, identify when today's regime resembles historically attractive entry phases.
01
Daily run
After the US close, the new SPY dataset is processed. The public service then shows the current status or, without Live Access, a delayed view.
02
Historical similarity
The score measures similarity, not certainty: structure, stress, and recovery profile are compared with prior market regimes.
03
Transparent record
Strong, weak, and open signals stay visible so users can inspect how the logic behaved on unseen data since 2020.
What is publicly reviewable
- Delayed SPY chart with historical signal markers.
- Out-of-sample signal count since 2020-01-01.
- 20D and 60D forward returns for historical signals.
- Open and completed cases clearly separated in the track record.
- Sample CSV to inspect the data format.
What is intentionally not published
Internal weights, threshold logic, and non-public indicator variants remain inside the product. The process, signal data, and track record remain public enough for review without turning the live logic into a static trading rule.