How accurate is Coneview?
Every forecasting service claims accuracy. Almost none publishes the times it was wrong. This page does, and it updates automatically — we cannot quietly drop a bad result.
Across the 6 instruments and 18 horizons below, the model beats a naive baseline on 2 of 18. On the rest it has no edge — and we show those rows rather than hiding them. That is the honest state of short-horizon market forecasting, ours included.
What we're measuring
- •Out-of-sample. The model trains only on data before a given date, then forecasts the period after it. It never sees the answer it is graded on — a walk-forward backtest, repeated across history.
- •Directional accuracy. How often the model called up-or-down correctly. On its own this number is close to meaningless.
- •The baseline.Markets drift upward, so "always say higher" scores surprisingly well. Every model must be measured against that, and the difference is the only figure that matters.
- •Edge. Accuracy minus baseline. Positive means the model added something. Zero or negative means it did not, and we label it no edge.
Live results
| Instrument | Horizon | Accuracy | Baseline | Edge | Verdict |
|---|---|---|---|---|---|
| Bitcoin | 1-day | 49.8% | 50.7% | -0.9% | no edge |
| 7-day | 51.1% | 52.4% | -1.3% | no edge | |
| 30-day | 53.4% | 53.8% | -0.4% | no edge | |
| Gold Spot | 1-day | 52.7% | 53.8% | -1.1% | no edge |
| 7-day | 57.9% | 57.9% | +0.0% | no edge | |
| 30-day | 54.4% | 64.5% | -10.1% | no edge | |
| Apple | 1-day | 52.7% | 53.7% | -1.0% | no edge |
| 7-day | 57.7% | 58.0% | -0.3% | no edge | |
| 30-day | 60.8% | 63.5% | -2.7% | no edge | |
| S&P 500 | 1-day | 53.8% | 54.8% | -0.9% | no edge |
| 7-day | 61.1% | 60.6% | +0.5% | has edge | |
| 30-day | 67.1% | 67.3% | -0.2% | no edge | |
| EUR/USD | 1-day | 48.7% | 51.2% | -2.5% | no edge |
| 7-day | 50.0% | 50.7% | -0.7% | no edge | |
| 30-day | 51.5% | 52.1% | -0.6% | no edge | |
| Crude Oil Futures | 1-day | 52.6% | 52.8% | -0.2% | no edge |
| 7-day | 51.7% | 53.4% | -1.6% | no edge | |
| 30-day | 52.8% | 51.5% | +1.3% | has edge |
Refreshed automatically. Each instrument's full scorecard, including probability calibration, lives on its own forecast page.
Why so many rows say "no edge"
Because markets are close to efficient, and short-horizon direction is mostly noise. A model that claimed an edge everywhere would be overfitting its own history and lying to you about it. When the direction is a coin toss, Coneview says so — and the range becomes the entire forecast, which is still genuinely useful information about how much room the market plausibly has.
This is the whole premise of the product. We would rather tell you we don't know than sell you a confident number we invented. Read more on what makes a probability trustworthy.
Past performance does not guarantee future results. Backtested figures are out-of-sample but remain historical. Coneview provides statistical market information, not financial advice.