Model v1.0 — September validation report
The methodology is still v1.0. This page is the performance review, not a new model. Weights did not change. Buy-and-hold, a random directional call, and simple momentum are now scored on each Signal page from that symbol’s own closes. Those figures are in-sample. They are not a universe result, so quote pages still withhold the expected-return figure. The S&P 500, a sector ETF, and analyst consensus are not on the replay.
- Updated
- September 2026
- Source
- Stored model calls and quote inputs
- Method
- TECHi Forward factor model
- Version
- v1.0
- Confidence
- Simple baselines are per symbol. Market baselines are not published.
- Freshness
- experimental
- Status
- Experimental
Baselines
- Buy and holdScored on each Signal page from that symbol’s stored closes. The comparison is in-sample and overlapping. It is not a universe result, so it does not unlock ranks or expected returns.
- Random directional callScored on each Signal page as the expected hit rate of a uniform random call, one time in three. It is not a seeded simulation and not a universe result.
- Sector ETFNot published. The price replay does not include a sector ETF series on the same dates.
- S&P 500Not published. The price replay does not include an S&P 500 series on the same dates.
- Analyst consensusNot published. The replay does not store a historical analyst rating on each sample date.
- Simple momentumScored on each Signal page as the sign of the trailing 20-session return known on the sample date. It is not a universe result.
What has to be true before a strong label
- At least 100 matured predictions.
- Validation confidence of at least 0.6.
- At least 30 names before a sector calibration is shown.
- Data completeness of at least 0.7, or the state is insufficient data.