Mean Reversion Trading

Mean reversion is the tendency of a price to return toward its historical average over time. After a sharp move up or down, the price “snaps back” rather than continuing in the same direction. This is measured using the z-score:

z = (current_price - mean) / standard_deviation

A z-score of -2 means the price is two standard deviations below its mean — statistically unusual and likely to revert. A z-score near 0 means the price is at its average.

Not all OSRS items behave the same way. Items that mean-revert reliably share a common trait: a supply anchor that creates a price floor or ceiling.

• Shop-anchored items (runes, feathers, vials): NPC shops sell unlimited quantities at a fixed price, creating a hard supply floor. Prices can spike above shop price due to convenience demand but always get pulled back.
• Cost-anchored items (crafted goods, potions): The cost of raw materials creates an input-cost floor. Prices rarely stay below crafting cost for long.

Statistically, mean-reverting items show a first-order autocorrelation (AC1) below -0.2, meaning a price increase today predicts a decrease tomorrow. This negative autocorrelation is the mathematical signature of mean reversion.

The practical strategy is straightforward:

1. Identify items that are confirmed mean-reverters (AC1 < -0.2).
2. Buy when the z-score drops to -2.0 or below (price is abnormally low).
3. Sell when the z-score returns toward 0 (price has reverted to its mean).

Research validated this approach across 200+ items: using a z-score threshold of 2.5 with a 7-day holding period on confirmed mean-reverting items yields a 71% win rate. This is a significant edge in a market where random entries produce roughly 50/50 outcomes.

Tighter thresholds (higher z) give fewer but higher-quality signals. The trade-off is fewer trading opportunities.

Demand-driven items — boss drops, rare equipment, and items tied to meta shifts — often trend rather than revert. When a new boss is released or a combat meta changes, prices can move persistently in one direction for weeks.

These items show positive autocorrelation (AC1 > +0.1), meaning a price increase today predicts another increase tomorrow. Applying mean-reversion strategy to trending items is a losing proposition — you'd be buying into a falling knife or selling short a breakout.

Always check an item's classification before trading. The Screener page shows AC1 values and item categories to help you avoid this mistake.

The most powerful approach combines multiple signal sources. When a Prophet ML forecast and a z-score signal agree on direction, the combined signal achieves a 64.7% win rate — significantly higher than either signal alone (roughly 55% each in isolation).

This ensemble effect works because the two signals capture different information. Prophet models trend and seasonality from historical patterns, while z-score captures statistical extremes. Agreement between independent signals reduces false positives.

On the Strategies page, signals that have multi-strategy agreement are highlighted. Prioritize these over single-strategy signals.