Custom prediction AI models.
Backed by our own data intelligence.
DeepInvest builds bespoke prediction models for your universe and grades every call in public against what the market did — so you're predicting tomorrow's markets, not explaining yesterday's.
Trained on billions of historical data points across countries, commodities, industries and financial markets.
Every call, graded against reality.
Not a polished backtest — the model's actual time-locked predictions on live equities, each graded against what really happened. We show the record, not a highlight reel.
Models trained on the event the grader scores: exact session dates, judged against the bar knowable when the prediction was made, sqrt(h) times the daily sigma over the last 60 sessions. Labels and grades share one implementation.
Windows run from the day's close to the close of the h-th session after it. Models were trained on labels whose bar was read at the window end, so a move inside the window widened its own bar.
Predictions made seven days a week on calendar-day windows, by models trained under defects since corrected: in-sample calibration, price history partly invisible or mis-ordered at training time, and no 'down' events in training.
AUC:Ranking skill (0.5 = coin flip, 1 = perfect), measured within each day's model and averaged across days weighted by realised moves, so a day-to-day shift in score scale cannot masquerade as skill or its absence. Base rate: Fraction of outcomes that were significant moves. Avg p_jump: Average confidence assigned. Every era is graded under the same current definition of a significant move (grader g2.sqrt-h.60s.z2); what separates the eras is the definition their models were trained on. Each era's record stays on the page, shown separately so none flatters or drags another.
| Company | Horizon | Move predicted | Actual move | Outcome |
|---|---|---|---|---|
| SoftBank Group Corp. | 14d | 12% | -22.59% | quiet |
| Murata Manufacturing Co., Ltd. | 14d | 15% | -21.80% | quiet |
| Caterpillar Inc. | 14d | 13% | -21.60% | significant |
| Bayer AG | 14d | 13% | +20.00% | significant |
| Alphabet Inc. Class A | 14d | 18% | -18.95% | significant |
| Heidelberg Materials AG | 14d | 17% | +18.84% | significant |
| Caterpillar Inc. | 5d | 13% | -18.77% | significant |
| Shopify Inc. (Class A) | 14d | 13% | -17.16% | quiet |
| Hitachi, Ltd. | 14d | 13% | -17.14% | significant |
| Advanced Micro Devices Inc. | 5d | 10% | -16.84% | significant |
| Bayer AG | 5d | 9% | +16.41% | significant |
| Fresenius SE & Co. KGaA | 14d | 12% | +15.84% | significant |
| Meta Platforms Inc. | 14d | 13% | -15.65% | quiet |
| Heidelberg Materials AG | 5d | 15% | +15.38% | significant |
| Sanofi | 14d | 17% | +15.31% | significant |
| Tokyo Electron Limited | 14d | 13% | -15.12% | quiet |
| The Home Depot Inc. | 14d | 16% | +14.25% | significant |
| Advanced Micro Devices Inc. | 14d | 13% | -14.00% | quiet |
| Hitachi, Ltd. | 5d | 14% | -13.74% | significant |
| Glencore plc | 14d | 12% | -13.64% | quiet |
The world produces signals. We read them.
Turns billions of signals into our proprietary data intelligence — the foundation every custom model is trained on.
Prediction history, accuracy and risk.
Custom prediction models, graded in public.
Bespoke AI models trained on your universe and backed by our own data intelligence — with every call time-locked and graded against the outcome.