This is the third and final episode in the mean reversion series with Cesar Alvarez from Alvarez Quant Trading. The first two episodes covered strategy construction, exits, position sizing, and risk control. This one is a Q&A format: listener questions on everything from one-day hold setups in XAU to shorting mechanics, sector overweighting, and how to think about tail risk.
The content is dense and practical. Cesar doesn’t pad answers. If he hasn’t tested something, he says so. What follows is a structured summary of the most useful material from the session.
Watch the full episode below, then read on for the complete breakdown.
One-day holds and entry signals for XAU
For a one-day hold strategy on gold (XAU), Cesar recommends looking for heavily stretched readings. Two specific conditions he uses: a 2-period RSI below 1, or a Connors RSI below 5. Both set up short-term bounces reasonably well.
The trade-off is frequency. The tighter the threshold, the stronger the edge but the fewer the trades. If RSI below 1 doesn’t generate enough setups, loosening to 1.5 or 2 produces more trades but a smaller average edge. You’re always making that choice.
Sector concentration: biotech and oil
Mean reversion setups cluster in sectors. When biotech or oil sells off hard, a strategy can end up fully weighted in one corner of the market. Cesar has tested both full exclusion and percentage caps (no more than 20% in any one sector).
The honest result: sector caps reduce returns. The concentration, as uncomfortable as it feels, does increase performance. It also increases portfolio volatility.
Cesar’s personal approach is to allow the concentration without a hard cap. He notes that roughly 90% of the time it works out fine, and 10% of the time it produces an outsized drawdown. His rationale: if the concentration would cause a trader to stop following the system, add the cap. Rules that keep you trading are more valuable than rules that optimise raw returns.
Small caps versus large caps
The question of whether mean reversion works better in large caps than small caps came up. Cesar’s answer: small caps tend to have larger edges, not smaller ones.
The reason is structural. Large quant funds can’t trade small caps with thin volume. Individual traders dominate those names. That creates more exploitable overreaction when stocks get stretched.
The cost is higher volatility. Larger moves in both directions come with the territory. Cesar finds this trade-off acceptable and sees it in both his mean reversion and breakout strategies.
Handling the rubber band that snaps
One listener raised a specific problem: stocks that stay stretched for 15 to 20 days, well past the typical 5-day mean reversion hold, wiping out profits from multiple prior trades. Cesar calls this the rubber band snapping rather than bouncing.
His fix is an n-bar exit. If the position hasn’t resolved after a set number of days (say 7), he exits regardless of RSI or moving average condition. This limits exposure to the situations where the expected reversion simply doesn’t happen. The trade-off is occasionally exiting before a delayed bounce. Cesar treats that as acceptable.
Entering and exiting at the close
Several listeners asked about close-to-close execution, since backtesting often shows an edge in the overnight period that gets lost by exiting at the open.
Cesar’s take: the close versus open edge is not consistent year to year. Looking at a full decade, closing exits might appear stronger on average, but in any individual year it flips. He personally exits at the open, not the close, for practical reasons. Executing at the close requires running scans 15 minutes before the close and placing all orders in the final 5 minutes. That’s operationally fragile. A data outage, a slow connection, or a last-minute price move that disqualifies a signal can leave the backtest and live trading diverging in ways that are hard to diagnose. Exiting at the open is simpler and more predictable.
Volume as a filter: doesn’t work
Cesar has tried, extensively, to make volume work as a filter across mean reversion, breakout, and trend following strategies. He cannot get it to add value. It doesn’t improve net profit or reduce drawdowns in any meaningful way across his testing. He has not found a way to use it. If a listener has a specific method that works, he’d like to hear it.
Shorting mean reversion: what changes
Shorting works, but Cesar no longer does it. Two reasons. First, a short can gap up 100% overnight. He has been in several such trades. The asymmetry is brutal: a long position can fall to zero in theory, but real overnight gaps don’t do that. A short can double against you in a single print.
Second, short availability is invisible in backtesting. You can’t know whether a stock was borrowable at the time of a historical signal. Backtest results and live trading diverge in ways that are hard to attribute.
For those who do want to short, the rules need to be exaggerated, not just inverted. A long entry at RSI below 10 doesn’t become a short entry at RSI above 90. The short equivalent is RSI above 99, held there for 3 or more days, stretched 10% above the 5-period moving average. You’re looking for a more extreme setup because the risk on the wrong side is asymmetric.
Key metrics and minimum thresholds
Cesar’s primary metrics: CAGR, top 5 drawdowns, top 5 draw lengths, and yearly returns by year. He does not use Sharpe ratio or Sortino.
His minimum thresholds for a strategy worth trading:
- CAGR: at least 15 to 18% per year over the last 10 years, for large caps or ETFs. For smaller cap stocks with thin volume, he expects 25 to 30%.
- Drawdown: preferably in the mid-20s. Below 15% is ideal; above 35% is a reason to reconsider.
- Yearly distribution: no single year should dominate. A strategy that made all its money in 2008 and flat-lined since needs scrutiny.
Tail risk and position sizing
Tail risk in mean reversion is almost always overnight. Cesar has never seen a mean reversion position drop 50% over multiple days. It happens in a single gap down.
His approach: size positions so that if any single stock opens down 50%, the account can continue trading. If that scenario would stop you, position sizes are too large. For anyone trading Russell 3000 or smaller cap names, that gap will happen. Sizing to survive it is the only protection available.
Portfolio allocation to mean reversion
Cesar keeps approximately 30 to 40% of his total trading capital in mean reversion strategies, spread across five strategies. The rest is in breakout, trend, ETF, and volatility strategies. Different strategies behave differently across market regimes. Low-volatility markets shrink mean reversion edges; breakouts tend to do better in those conditions.
Adjusting parameters over time
Cesar checks parameters roughly every one to two years, but not to update them. The check is a robustness test: does the current parameter sit in the middle of a range that all works reasonably well, or does it happen to be the worst performer in the neighbourhood? If the nearby parameters are all doing poorly, the concern is structural, not one of bad luck with a specific setting.
He does not run annual re-optimisation. The only time people want to adjust parameters is when the strategy is underperforming, which is exactly the wrong moment to do it.
Related episodes
- Building mean reversion strategies: Cesar Alvarez Part 2
- Low-effort trading strategies: Cesar Alvarez
- How to build mean reversion trading strategies: Stefan Friedrichowski
Want to learn more about building better trading systems? Subscribe to the Better System Trader podcast for weekly interviews with the world’s top systematic traders and quantitative researchers.
