QERN Forecast
We run a price-forecast model over PSX and US stocks and publish how it does, including when it does badly. Powered by an open-source AI model for candlestick data. Not investment advice.
The forecasts themselves are withheld
In forecasts scored after the fact across 214 windows, the model's median was closer to the outcome than assuming no change in only 26% of cases (50% would be no skill), so its forecasts are not shown.
A forecast is only shown once the evidence says it is better than assuming nothing changes. This page publishes that evidence either way. The model is being tested on more markets and configurations, and its forecasts are scored against what actually happens every night.
Live results
214 windows , 2026-10-09 to 2026-10-16; forecasting 10 sessions ahead.
The model did not beat simply assuming the price stays where it is.
The ranges were too narrow: they should hold 80% of outcomes and held 56%.
Live: forecasts scored after the fact. Recorded 2026-10-31 on 39 symbols, 8 sampled paths per forecast. Windows from adjacent days overlap, so the count overstates independent observations.
Forecast against outcome
3883 forecast steps scored against what actually happened, since 2026-10-09. Each is compared with the real session of that step, not the calendar date, so holidays do not shift anything.
By day
| Session | 1 ahead | 5 ahead | 10 ahead |
|---|---|---|---|
| 2026-10-30 | 49% beat · 44% in range n=39 | 38% beat · 46% in range n=39 | 23% beat · 54% in range n=39 |
| 2026-10-29 | 33% beat · 41% in range n=39 | 31% beat · 62% in range n=39 | 21% beat · 46% in range n=39 |
| 2026-10-28 | 31% beat · 67% in range n=39 | 26% beat · 64% in range n=39 | 28% beat · 67% in range n=39 |
| 2026-10-27 | 34% beat · 74% in range n=38 | 21% beat · 66% in range n=38 | 28% beat · 56% in range n=39 |
| 2026-10-26 | 44% beat · 56% in range n=39 | 26% beat · 59% in range n=39 | 21% beat · 54% in range n=39 |
| 2026-10-23 | 33% beat · 74% in range n=39 | 38% beat · 54% in range n=39 | 42% beat · 58% in range n=19 |
| 2026-10-22 | 41% beat · 26% in range n=39 | 28% beat · 59% in range n=39 | — |
| 2026-10-21 | 42% beat · 63% in range n=38 | 33% beat · 64% in range n=39 | — |
| 2026-10-20 | 38% beat · 67% in range n=39 | 31% beat · 67% in range n=39 | — |
| 2026-10-19 | 28% beat · 79% in range n=39 | 36% beat · 62% in range n=39 | — |
By week
| Week of | 1 ahead | 5 ahead | 10 ahead |
|---|---|---|---|
| 2026-10-26 | 38% beat · 56% in range n=194 | 28% beat · 59% in range n=194 | 24% beat · 55% in range n=195 |
| 2026-10-19 | 37% beat · 62% in range n=194 | 33% beat · 61% in range n=195 | 42% beat · 58% in range n=19 |
| 2026-10-12 | 37% beat · 63% in range n=175 | 26% beat · 37% in range n=19 | — |
Most recent scored forecasts
| Symbol | Made | Ahead | Range forecast | Happened | In range |
|---|---|---|---|---|---|
| AAPL | 2026-10-29 | 1 | $340.50 – $354.23 | $344.77 | yes |
| AAPL | 2026-10-23 | 5 | $334.00 – $367.89 | $344.77 | yes |
| AAPL | 2026-10-16 | 10 | $308.69 – $360.21 | $344.77 | yes |
| AMZN | 2026-10-29 | 1 | $240.40 – $247.39 | $240.35 | no |
| AMZN | 2026-10-23 | 5 | $253.68 – $270.25 | $240.35 | no |
| AMZN | 2026-10-16 | 10 | $256.78 – $292.54 | $240.35 | no |
| BAHL | 2026-10-29 | 1 | PKR 153.60 – PKR 159.19 | PKR 150.89 | no |
| BAHL | 2026-10-23 | 5 | PKR 149.15 – PKR 153.09 | PKR 150.89 | yes |
| BAHL | 2026-10-16 | 10 | PKR 147.01 – PKR 154.52 | PKR 150.89 | yes |
| BRK.B | 2026-10-29 | 1 | $538.72 – $542.57 | $542.82 | no |
| BRK.B | 2026-10-23 | 5 | $510.28 – $526.29 | $542.82 | no |
| BRK.B | 2026-10-16 | 10 | $486.02 – $519.92 | $542.82 | no |
| DGKC | 2026-10-29 | 1 | PKR 173.80 – PKR 183.14 | PKR 178.15 | yes |
| DGKC | 2026-10-23 | 5 | PKR 179.06 – PKR 190.39 | PKR 178.15 | no |
| DGKC | 2026-10-16 | 10 | PKR 171.04 – PKR 199.43 | PKR 178.15 | yes |
| DIA | 2026-10-29 | 1 | $520.16 – $529.78 | $517.00 | no |
| DIA | 2026-10-23 | 5 | $506.38 – $524.53 | $517.00 | yes |
| DIA | 2026-10-16 | 10 | $503.00 – $544.38 | $517.00 | yes |
| EFERT | 2026-10-29 | 1 | PKR 195.41 – PKR 199.94 | PKR 194.99 | no |
| EFERT | 2026-10-23 | 5 | PKR 184.96 – PKR 189.14 | PKR 194.99 | no |
| EFERT | 2026-10-16 | 10 | PKR 177.63 – PKR 188.64 | PKR 194.99 | no |
| ENGROH | 2026-10-29 | 1 | PKR 251.28 – PKR 259.33 | PKR 256.34 | yes |
| ENGROH | 2026-10-23 | 5 | PKR 242.43 – PKR 263.49 | PKR 256.34 | yes |
| ENGROH | 2026-10-16 | 10 | PKR 255.01 – PKR 297.23 | PKR 256.34 | yes |
| FFC | 2026-10-29 | 1 | PKR 550.68 – PKR 567.16 | PKR 559.49 | yes |
| FFC | 2026-10-23 | 5 | PKR 517.14 – PKR 557.37 | PKR 559.49 | no |
| FFC | 2026-10-16 | 10 | PKR 548.89 – PKR 661.19 | PKR 559.49 | yes |
| GLD | 2026-10-29 | 1 | $359.37 – $372.22 | $355.81 | no |
| GLD | 2026-10-23 | 5 | $366.79 – $379.41 | $355.81 | no |
| GLD | 2026-10-16 | 10 | $371.30 – $406.81 | $355.81 | no |
Tell us what you think
The model failed its first test, so we are not showing its forecasts. Would you want to see them anyway, clearly labelled as untested?
How to read this
- It is model output, not research and not a recommendation to buy or sell. The range is the model's own spread over sampled paths, which the track record above shows is not always well calibrated.
- Prices are PSX quotes (about 15 minutes delayed, from the community psxdata-api) and Alpaca's IEX feed. Neither is adjusted for splits or bonus issues, so a symbol is not forecast until it has about three months of history after one.
- Forecasts are made every night after the close. "As of" is the last session the model saw.
- The model itself is open source. The pipeline around it, the data handling and the testing are QERN's.