Regime Models — Telonics Trading

Neither model predicts direction — only how far price is likely to travel, and how cleanly. And neither is a crystal ball. Both are measurably better than chance, and both are a long way from certain.

More information

what they forecast: How efficiently the next session will move: how far price ends up from where it opened, compared with the total distance it covered getting there. A typical day travels about twenty times further than it finishes from its open. The second model forecasts how volatile that session will be. Both cover the 9:30am–4pm ET session only, never the overnight.
when it is decided: At the previous session’s 4pm ET close, using only that session’s data, and published before 9:30am ET. A forecast that misses the open is thrown out rather than counted.
how it is judged: Against levels published alongside the forecast, worked out from the previous 252 sessions. The bar is set the evening before, never after the result is known.
how accurate they are: Both models beat chance; neither is close to certain, and we think the way they are scored matters more than the score itself. The efficiency model is a ranking tool — it separates sessions that will travel efficiently from ones that will chop, and that separation is consistent, but it is a tendency across many days rather than a call on any one of them. The volatility model produces the higher raw accuracy figure, and we deliberately do not lead with it: volatility is strongly persistent, so a forecast of “the same as today” already scores well, and any honest measurement has to credit the model only with what it adds on top of that.

Forecasts describe the expected character of a trading session. They are not trading advice, not a signal to buy or sell anything, and say nothing about direction. Past forecasting accuracy does not guarantee future accuracy. Trading futures involves substantial risk of loss and is not suitable for all investors.