Methodology
Volum Invest AS publishes systematic, quantitative investment signals. This page describes, in plain language, how those signals are produced. It is provided to satisfy the methodology-disclosure expectation under MAR Article 20 and Commission Delegated Regulation (EU) 2016/958.
What the model does
Each instrument in our universe is scored by a multi-factor composite model that combines several families of evidence:
- Momentum: the strength and direction of the price trend over short and longer horizons.
- Volume: whether trading participation is confirming or fading.
- Volatility: the level and trend of realised volatility.
These factors are combined into a single composite score, and instruments are ranked cross-sectionally (relative to one another) by that score. Buy, sell, and trim signals follow from where an instrument's score sits relative to the model's entry and exit thresholds. The process is fully systematic: no discretionary judgment is applied to individual signals.
The detailed factor definitions, transformations, weights, and threshold values are proprietary and are not disclosed.
What the signals mean, and over what horizon
- BUY: the model's entry conditions are live for that instrument.
- SELL: the model has exited the position; its exit conditions have fired.
- TRIM: a partial reduction, not a full exit. It is raised in two distinct situations: when price has reached the upper band, and when price has closed below its trade average. The first is a strength signal, the second a weakness signal; both call for reducing rather than closing the position.
Every BUY carries a tier. Since 30 October 2026 each BUY is published with one of three words, High conviction, Strong or Standard, from the strength of the model's score on the day the signal fired, read against two fixed levels of the model that are not disclosed. The model portfolio's slot rules read the same tiers. In the five-year backtest on which the tiers were set, entries on stronger tiers won more often and earned more per day held; the closed trades by tier are shown on Signal History. A tier describes the reading; it is not a probability or a forecast, and it changes nothing about when a position is opened or sold.
The model does not use a fixed holding period. A position opens when the entry conditions hold and stays open until an exit condition fires. There are seven, and the first to trigger closes the position: the model's holding score falling below the exit threshold; a tightened exit floor that applies after 30 days to positions opened in specific market conditions; a shift into a high- or crisis-volatility regime; a sharp one-day fall on exceptionally heavy volume; a close a set amount below the entry price (a loss limit, since 5 November 2026); a defined break in trend on a position already in profit; and a combined market-wide volatility trigger. Since 19 October 2026 the holding score is a separate reading from the same model rather than the published composite score, so a position can be sold while its published score still looks strong. How positions are opened is unchanged.
Market-wide conditions can suspend new entries. Separately from the per-instrument score, two conditions of the market as a whole block the model from opening any new position while they hold: a sustained downtrend in the broad market index, and an elevated reading of the market volatility index. Neither affects positions already open, which continue to run their normal exit rules, and neither changes any instrument's score. Their effect is that BUY signals can be absent across the whole universe for reasons that have nothing to do with the individual instruments. The thresholds are proprietary.
The practical consequence is that holding periods vary widely between instruments and market regimes: some positions close within days, others run for months. Because there is no single representative figure, we publish the actual distribution rather than an average: every closed position on Signal History shows its own entry date, exit date, days held and exit reason. That page is the authoritative record of realised holding periods and covers all closed positions, not a selection.
The Trade signal
Since 27 October 2026 the service also publishes the Trade signal: a second, short-term recommendation produced from the same model scores as the main signal, under its own entry and exit rules.
- Buy at the next open: on the last completed session the instrument's score had been strong for two sessions in a row, its price had pulled back towards the lower part of a range that is still rising, and the session was a quiet down day. Of the market-wide conditions described above, the elevated volatility reading suspends Trade signal entries as well; since 8 November 2026 a sustained downtrend in the broad market index no longer does, and one of the Trade signal's requirements on an instrument's longer-term price history was relaxed at the same time.
- Sell at the next open: the price has reached the upper part of its range, or another defined exit condition has fired. There is no fixed holding period; most Trade signals last about two weeks.
- Tiers: every Trade signal carries a tier, High conviction, Strong or Standard, from the strength of the model's score on the two sessions that decided it. In the backtest, stronger tiers have won more often and earned more per trade. A tier describes the reading; it is not a probability or a forecast.
- Plans: subscribers on the Pro plan receive the strongest Trade signal of each session, and subscribers on the Max plan receive every one. The choice is made by the plan alone, so every subscriber on a plan receives the same Trade signals.
The Trade signal has its own backtested history and its own live record, which began on 27 October 2026. Its own figures are shown on its own page and are never combined with the main signal's trade statistics. Since 8 November 2026 the model portfolio also takes Trade signals, only into a slot that is free and only in a name it does not already hold, so its return figure combines the two signals.
Sensitivity of the results
The published figures depend on the model's threshold and weight settings, and they are not equally stable under every condition. Three sensitivities matter most to anyone reading the performance:
- Entry threshold. The score required to open a position directly controls how many positions the model takes and their average quality. Lowering it admits more positions at a lower average win rate; raising it does the reverse. Changes here move results materially in both directions.
- Market regime. Results differ substantially between rising and falling markets, and between calm and volatile ones. Performance measured over a period dominated by one regime should not be read as representative of the others.
- Universe composition. The score is cross-sectional: each instrument is ranked against the others scored that day, so results depend on which instruments are in the universe. Adding, removing or reweighting the universe changes the ranking, and therefore the signals.
Model results are also subject to regime change, data limitations and errors in third-party market data. The signals are recomputed continuously, and past settings are not a guarantee that current settings will perform similarly. See the Full Disclaimer.
Universe
The model covers a broad universe of roughly 1,500 liquid instruments across the United States, Canada, the Nordics and the rest of Europe. Of these, about 80 are ETFs spanning broad-market exposure, sector SPDRs, thematic and leveraged products, and international funds (including Asia-focused ETFs); the rest are individual stocks. Instruments are included based on data availability and liquidity; names without sufficient price history are excluded. The exact selection rules are proprietary.
Update frequency
Signals are recomputed every 15 minutes throughout the trading day, using the latest available market data.
Data sources
Price, volume, and volatility inputs are sourced from public market data via the Yahoo Finance (yfinance) interface, using dividend-adjusted open/high/low/close (OHLC) data. Backtests use the same data.
Performance
Unless a figure is explicitly labelled otherwise, historical performance shown on the site is backtested / hypothetical: it is produced by applying the current model rules to past data and does not represent signals that were published live at the time. Backtested performance has inherent limitations and is not a reliable indicator of future results. See the Full Disclaimer.
Macro Regime (US growth and inflation)
The Macro Regime page is a separate model from the signal engine, and it is not a trading signal. It describes the direction of the United States economy: whether growth is speeding up or slowing down against a year earlier, and whether inflation is rising or falling against a year earlier. Each quarter, and each month, is placed in one of four regimes by those two answers: R1 (growth up, inflation down), R2 (growth up, inflation up), R3 (growth down, inflation up) and R4 (growth down, inflation down). The R stands for regime; the number is the cell.
- Growth. For the quarters, real GDP. The current quarter is estimated from the monthly data already published (employment, industrial production, retail sales, jobless claims). Beyond the current quarter the model does not forecast GDP; it assumes a normal pace of about 2% a year, adjusted for the effect of interest-rate changes, which reach growth with a lag of two to six quarters. For the months, a growth measure built from real consumer spending, industrial production, real income after transfers, real retail sales and hours worked.
- Inflation. The consumer price index, all items, projected month by month from gasoline (oil and pump prices), rents (market rents, which reach the index with about a year's lag) and the trend in everything else.
- Direction. Whether growth or inflation is up or down in a period is decided against the same period a year earlier, which is already published. A call is marked too close to call when the expected move is very small.
- Accuracy. The hit rates shown are from a backtest of month-end runs from January 2017, each using only the data available at that date, scored against the first published release. They are backtested and hypothetical, not a live record, and past accuracy is not a reliable indicator of future accuracy. The direction of next month's inflation and of the monthly growth measure are the calls the backtest supports best; the regime for quarters further out is much less certain.
- Data and cadence. Public data from the Federal Reserve's FRED database (which carries the BEA, BLS, Federal Reserve and University of Michigan series) and Zillow's market rents. The model is rebuilt on weekdays after the US data releases, and the page refreshes from its output within the hour.
The page carries the current quarter and the current month for every subscriber; the path further ahead is included in the Pro and Max plans. The detailed model settings are proprietary and are not disclosed.
Changes to the methodology
The methodology evolves as the model is refined. This page describes the current methodology and will be updated to reflect material changes.
Last updated: 2026-10-30