Forecast accuracy
How accurately Ironline forecasts movement.
Ironline forecasts how much a stock is likely to move, and shows how often those forecasts were right. Everything here uses price data Ironline owns. The methods inside the tools are not published.
We only publish forecasts that held up in historical testing. Everything else is shown as plain information, never as a prediction.
We double-check every published number by rebuilding it from scratch with separate code.
The Day 90% range held 88–91% in every year from 2010 to 2025, including 2020 and 2022.
In testing, 2010–2025.
This is the public version: forecast accuracy and how we test, only. Ironline does not publish win rates, returns, or portfolio performance claims. Ironline publishes forecast accuracy — how often the Expected Move ranking identified the higher-moving stock in historical testing. Forecast accuracy is not the same as a win rate or a trading return. It was prepared Oct 11, 2026 from the final verified results, and an independent review recalculated every figure.
Results
Summary
- Expected Move works. Take two stocks on the same date. The one Ironline forecast to move more did move more 61% of the time for the next day, 74% for the next week and a half, and about 80% for the next three months and the next year. A coin flip would be 50%.
- A higher ranking accuracy means the Expected Move forecast correctly identified which of two stocks would move more. It does not show whether buying that stock would have been profitable.
- The price ranges are honest where they're shown. A range labelled 90% held about 89 to 90 times in 100, in both halves of the history.
- Earnings reports matter. When a report fell inside the window, moves were often bigger than the range, so the dashboard warns members when that happens.
Method
How we test
516 companies were tested, including ones that later stopped trading. The 516 companies were selected based on data availability and coverage. 1.14 million Expected Move forecasts were checked, and more than 20,000 comparisons were recorded and counted across the research.
- Owned data only. Daily prices are rebuilt from minute-by-minute trade data Ironline owns. We also use SEC filings, Federal Reserve economic data, and short-borrow fees from Ironline's market-data provider. No vendor's ready-made results were used.
- Rules written first. Every test's pass rule was written down and fingerprinted before any result existed. A tool is used only if it passed the rule as written.
- Two halves. 2010 to 2025 is split into 2010–2017 and 2018–2025, and a result has to hold in both.
- A later check year. 2026 so far is shown separately as a check. It was looked at once before, so it isn't a blind test.
- No peeking at the future. Every forecast uses only information available on its date. Deliberately scrambled future data was used to prove the tests can't see ahead.
- Independent review. A separate reviewer rebuilt the data and recalculated every published figure with their own code. All of them matched.
- Every comparison counted. This includes the ideas that didn't work, so the results aren't a highlight reel.
Live accuracy data is tracked from launch and will be reviewed before it is shown to members.
These results are exploratory. The same years, 2010 to 2025, were used both to build the tools and to test them, which can make results look better than they will be on new data. Past results do not guarantee future outcomes.
Expected Move
Expected Move: ranking accuracy
Of two stocks on the same date, how often the one forecast to move more did move more. Measured on daily closing prices.
| Window | 2010–2025 | 2010–2017 | 2018–2025 | 2026 check |
|---|---|---|---|---|
| Next day | 61.3% | 61.0% | 61.5% | 63.0% |
| Next week and a half (7 trading days) | 73.5% | 73.0% | 73.9% | 74.5% |
| A long trade's typical time (16 trading days) | 76.7% | 76.5% | 77.0% | 78.9% |
| A short trade's typical time (8 trading days) | 74.0% | 73.6% | 74.3% | 75.5% |
| Next three months (63 trading days) | 80.6% | 80.3% | 80.8% | 82.9% |
| Next year (252 trading days) | 80.4% | 80.3% | 80.5% | none closed yet |
Price ranges
Price ranges: how often the price finished inside
| Window | Range | 2010–2025 | 2010–2017 | 2018–2025 | 2026 check |
|---|---|---|---|---|---|
| Next day | 68% | 68.0% | 68.1% | 67.8% | 65.8% |
| Next day | 90% | 89.9% | 90.0% | 89.8% | 89.4% |
| Next week and a half | 68% | 67.2% | 67.1% | 67.3% | 64.5% |
| Next week and a half | 90% | 89.1% | 89.1% | 89.1% | 89.4% |
| Long trade's typical time | 68% | 66.8% | 66.9% | 66.8% | 65.2% |
| Long trade's typical time | 90% | 88.8% | 88.8% | 88.7% | 89.0% |
| Short trade's typical time | 90% | 89.0% | 88.9% | 89.2% | 90.5% |
Three months and one year are shown as a movement rank only: how a stock's expected movement compares with other stocks. Over those spans the whole market moves stocks together, so no range passed the test, and none is shown.
Earnings
When an earnings report falls inside the window
| Window | Range | Report inside the window | No report |
|---|---|---|---|
| Next day | 68% | 30.8% | 68.6% |
| Next day | 90% | 50.4% | 90.6% |
| Next week and a half | 68% | 52.3% | 69.1% |
| Next week and a half | 90% | 76.5% | 90.7% |
| Long trade's typical time | 90% | 83.6% | 90.5% |
| Short trade's typical time | 90% | 77.9% | 90.7% |
Limits
What these results do not show
- They don't show that any Ironline tool predicts whether a stock will go up or down.
- They are not performance claims. Ironline does not publish win rates, returns, or portfolio performance claims. Ironline publishes forecast accuracy — how often the Expected Move ranking identified the higher-moving stock in historical testing. Forecast accuracy is not the same as a win rate or a trading return. Nothing here shows or guarantees profit.
- They are exploratory and historical. Live results are tracked from launch.