02 · Create a forecast

Turn one pre-open signal into one closing-level forecast.

Once the 09:15 New York futures observation is locked, M1-v1.0 maps that signal into a predicted Nasdaq Composite return and then into one exact forecast of the day's closing level.

Input

Start with information that existed before the cash open.

Signal\(r^{NQ}_{t,\mathrm{pre}}\)
Level anchor\(S_{t-1}\)

The first quantity is the Nasdaq-100 futures return measured on the fixed pre-open window. The second is the previous Nasdaq Composite close.

Frozen production model

M1-v1.0 converts the signal into a return forecast.

M1-v1.0 is a least-absolute-deviations median regression. It was selected using historical out-of-sample testing before the prospective forecast is evaluated.

\[\widehat r_t=\widehat\alpha+\widehat\beta r^{NQ}_{t,\mathrm{pre}}\]
The model is not re-selected after observing the target day's market close. The prospective experiment asks how a frozen rule performs on genuinely new dates.
Level transformation

Convert the predicted return into the thing the site actually publishes.

Graham Says forecasts the Nasdaq Composite closing level, not merely a return. The previous close anchors the prediction.

\[\widehat S_t=S_{t-1}\exp\!\left(\widehat\alpha+\widehat\beta r^{NQ}_{t,\mathrm{pre}}\right)\]
Production sequence

From observation to immutable forecast.

01Read signal

Use the locked pre-open futures return.

02Estimate return

Apply the frozen M1-v1.0 relationship.

03Anchor level

Start from the prior Nasdaq Composite close.

04Create level

Transform the return forecast into an index level.

05Lock

Write and publish the target-date forecast.

Model detail

Want the econometrics rather than the pipeline?

The dedicated M1 page contains the historical tournament result, LAD objective, backtest chart, prospective record and full mathematical explanation.

Historical winner · production modelM1 · NQ-LAD
Explore the model →
Previous step← 01 · Observe futuresNext step03 · Score after close →