Module 11 of 14 · Faster & relative value
Fast mean reversion with a trend filter
Module 6 showed that trend fades at horizons of a few days and mean reversion appears instead. A strategy can harvest it: estimate an equilibrium price — for example a short EWMA of daily closes, about a week long — and lean against deviations from it. Price above equilibrium means short, below means long, with the size of the forecast set by the distance in units of daily volatility.
On its own this is dangerous. Mean reversion keeps buying as a market falls — “catching a falling knife” — so it has steady small profits and occasional severe losses, the negative skew from Module 1. Two fixes make it far safer:
- Trend filter. Only take mean-reversion trades that agree with a slow trend signal. In an uptrend you buy dips below equilibrium but never short the rallies; when the trend flips, the position is closed. The trend overlay does the job of a stop-loss without arbitrary stop levels.
- Volatility regime scaling from Module 10, applied more aggressively: mean reversion loses most in volatility spikes.
The other problem is cost. Holding periods are a day or two, so turnover is enormous. The fix is in execution: because the position is a direct function of price, you can calculate in advance the exact prices at which you would want one contract more or one fewer, and rest limit orders there. You get paid the spread instead of paying it, and pay commission only. Cancel the ladder at the end of each session and use a market order if price gaps through several rungs overnight.
🎛 Limit-ladder lab
Ideal position
-6.59
contracts
Hold (rounded)
-7
short
Next sell / buy at
100.684 / 100.592
nearest rungs
| Rung | Order | Limit price | Position after fill |
|---|---|---|---|
| buy 3 | BUY 1 | 100.408 | -4 |
| buy 2 | BUY 1 | 100.500 | -5 |
| buy 1 | BUY 1 | 100.592 | -6 |
| sell 1 | SELL 1 | 100.684 | -8 |
| sell 2 | SELL 1 | 100.777 | -9 |
| sell 3 | SELL 1 | 100.870 | -10 |
Forecast = scalar × (equilibrium − price) ÷ daily price volatility, capped at ±20, and switched off when it disagrees with the slow trend filter. Because the position depends only on price, you can solve in advance for the exact prices at which you would buy or sell one more contract and rest limit orders there: you collect the spread instead of paying it. A gap through several levels is handled with a market order. Illustrative inputs; risk target 20%.
Powerful, and demanding
A trend-filtered mean-reversion strategy diversifies a trend-and-carry portfolio well because its returns come from a different horizon. Backtests of strategies like this can look spectacular, which is a reason for suspicion rather than celebration: limit-order fills are hard to simulate honestly, and hourly history is short.
It also asks a lot of the operator: fully automated execution, intraday monitoring, no position buffering, enough capital to hold several contracts in each market, and hard trade limits (Module 14), since a strategy that places dozens of orders a day can lose a lot of money very quickly when something breaks. Treat it as an advanced, separately run sleeve — not a starting point.
Things to try
- • Move the current price below the equilibrium with the filter off, then switch to 'Slow trend is DOWN': the long position disappears.
- • Raise volatility from 12% to 30%: the rungs spread out, because each contract now represents more risk and needs a bigger price move.
- • Raise capital to $3m: rungs get closer together. Eventually they are less than a tick apart — the capacity limit of fast strategies.
Original interactive lesson. Labs run on a seeded synthetic market in your browser, built to illustrate mechanics, not to predict any real market. Contract figures are illustrative; check exchange specifications. Educational analysis, not investment advice.