PSI 20 forecast: probability ranges
PSI 20 level: 9,468 on 16 Oct 2026. The 80% probability range for the end of 2027 is 8,493 to 11,916, with a median of 10,206. Over the next 3 months the range is 8,914 to 10,237 and over 12 months 8,283 to 11,468. These are ranges, not targets: they show how far the level could plausibly move, built from past moves at today's volatility of 11.9% a year.
Where the level could go: 50%, 80% and 90% probability bands
Forecast ranges by horizon
| Horizon | 10% | 25% | Median | 75% | 90% | Chance higher |
|---|---|---|---|---|---|---|
| 1 week | 9,237 | 9,367 | 9,480 | 9,574 | 9,669 | 53% |
| 1 month | 9,077 | 9,301 | 9,516 | 9,731 | 9,914 | 57% |
| End of 2026 | 8,914 | 9,279 | 9,591 | 9,873 | 10,175 | 61% |
| 3 months | 8,914 | 9,264 | 9,612 | 9,906 | 10,237 | 61% |
| 6 months | 8,704 | 9,348 | 9,757 | 10,176 | 10,568 | 67% |
| 12 months | 8,283 | 9,242 | 10,054 | 10,737 | 11,468 | 70% |
| End of 2027 | 8,493 | 9,216 | 10,206 | 11,072 | 11,916 | 72% |
Read it as: over 3 months, there is a 10% chance the level ends below the "10%" column, a 10% chance it ends above the "90%" column, and an 80% chance it ends in between. "Chance higher" is the share of comparable past periods, scaled to today's volatility, that ended above the starting level. The centre follows a long-run equity drift of about 6% a year, a stated assumption. It deliberately does not use this index's own past performance, which would reward past winners.
How likely is each level?
| Level | Move from today | Chance above it in 3 months | in 12 months |
|---|---|---|---|
| 7,570 | −20.0% | 98% | 97% |
| 8,520 | −10.0% | 96% | 87% |
| 9,000 | −5.0% | 88% | 80% |
| 9,230 | −2.5% | 77% | 75% |
| 9,710 | +2.5% | 42% | 64% |
| 9,940 | +5.0% | 23% | 55% |
| 10,420 | +10.0% | 6% | 35% |
| 11,360 | +20.0% | <1% | 12% |
The chance that the level is above a level on the target date (not that it touches the level on the way). Built from the same volatility-adjusted history as the ranges above.
More on PSI 20
Track record: how often did the outcome land inside the ranges?
| Horizon | Tests | Inside 50% range | Inside 80% range | Inside 90% range |
|---|---|---|---|---|
| 1 week | 293 | 55% | 84% | 91% |
| 1 month | 292 | 56% | 85% | 91% |
| 3 months | 290 | 57% | 87% | 93% |
| 6 months | 287 | 60% | 83% | 92% |
| 12 months | 281 | 64% | 86% | 92% |
Walk-forward back-test since 2 Dec 2002: every 21 trading days the model was run using only the data available that day, and we counted whether the level later finished inside the 50%, 80% and 90% ranges. A well-calibrated forecast scores about 50%, 80% and 90%. Tests for long horizons overlap, so they are not independent. This is the history of this one asset; the pooled results for all stock indices are on the forecasts page. The raw 80% range for stock indices was a little too narrow in the back-test (it held 82% to 85% of the time depending on the horizon), so the ranges are widened by 1.05× to 1.10×. That factor was fitted once on all stock indices, so the table is partly in-sample.
Live record: every forecast is stored once a week and checked when its horizon passes. The first snapshots were stored on 5 Oct 2026; the 1-week ones resolve within days and the 12-month ones in a year. Results will appear here as they come in, good or bad.
How the ranges are built
- Take the daily price history and measure today's volatility with an exponentially weighted average of squared returns (the RiskMetrics method, decay 0.94).
- For every past day, record how far the price moved over the horizon in units of that day's volatility. This keeps the real shape of past outcomes: fat tails and skew included.
- Take the 5th, 10th, 25th, 50th, 75th, 90th and 95th percentiles of the last ten years of those outcomes and scale them back by today's volatility. That gives the bands.
- Run the same recipe walk-forward through history to count how often the ranges held, and widen them where they were too narrow.
What it does not do. It does not know about earnings, central-bank meetings, elections or news, and it assumes the next period is drawn from the same distribution as the last ten years, so a regime change can break it. Treat the ranges as the minimum width of what could happen, not the maximum. Full method and results: forecasts. Related reading: drawdown, standard deviation and other risk measures, volatility and the VIX.
Frequently asked questions
What is the PSI 20 forecast for 2027?
We do not give a single target. Based on the PSI 20 history and today's volatility, the model puts the 80% probability range for the end of 2027 at 8,493 to 11,916 (median 10,206), from 9,468 on 16 Oct 2026. The 50% range is 9,216 to 11,072. These are statistical ranges, not predictions.
Will PSI 20 go up or down?
Nobody knows. Looking at past stock indices moves scaled to current volatility, the chance of finishing higher than today over 3 months is about 61% and over 12 months about 70%. That is a base rate, not a call: the range of outcomes is far wider than the tilt.
How accurate are these forecasts?
In a walk-forward back-test since 2 Dec 2002, the 80% range for PSI 20 over 3 months held 87% of the time and the 50% range 57% (targets are 80% and 50%); over 12 months 86% and 64%. The full table is on this page.
How is this different from an analyst price target?
A price target is one number from an analyst's opinion. These are ranges built from the actual distribution of past moves, adjusted for how volatile the market is today, with a published record of how often the outcome landed inside them. They say how much could happen, not what will.
How often is the forecast updated?
Every night after the close. The page address stays the same; the numbers and dates move with the market.
Educational information, not financial advice. Ranges describe how far prices have typically moved; they are not predictions or price targets, and past results do not guarantee future ones.