Building a mean reversion strategy is one thing. Building one that actually holds up across different market conditions is another. In this second episode of a three-part series, Cesar Alvarez from Alvarez Quant Trading walks through the structural decisions most traders get wrong: how to classify market regimes, how to rank trades when you have more signals than capital, and which exits actually matter.
If you missed part one, Cesar covered the core mechanics of mean reversion, universe selection, and how to measure when a stock has pulled back far enough to be worth trading. This episode picks up where that left off. The conversation is practical throughout. Cesar trades these strategies himself, so the advice is grounded in real-world constraints rather than theory.
The episode expanded from a planned two-part series to three parts because listener questions overflowed. That alone tells you how much material there is here.
Watch the full episode below, then read on for the complete breakdown.
Why most traders try to do too much with one strategy
Cesar’s first point is one that trips up a lot of systematic traders. The goal of getting a strategy to work across all market types sounds appealing, but in practice it leads to compromised strategies that do nothing particularly well.
A strategy that tries to perform in both trending and flat markets, or in both high and low volatility environments, ends up needing so many qualifications that it loses its edge in any single condition. Cesar’s approach is to design strategies with a specific regime in mind.
The two primary regimes he uses:
- Bull/bear classification: whether the SPY is above or below its 200-day moving average
- Volatility regime: whether the VIX or 100-day historical volatility on the SPY is above or below a threshold (typically 18 to 22)
He noted that combining these gives you four possible conditions: volatile bull, quiet bull, volatile bear, and quiet bear. In practice, quiet bear markets are rare, so two regimes is usually enough.
One underused option: rather than having separate strategies for each regime, you can run the same strategy with tighter parameters in unfavorable conditions. For example, in a bull market you might enter when RSI drops below 10. In a bear market, you tighten that to RSI below 1. You’re also screening out lower-quality setups and cutting position size in half during adverse conditions.
Measuring volatility: fixed thresholds vs. relative percentiles
On volatility measurement, Cesar prefers simple fixed thresholds over more complex percentile-based approaches. He calculates 100-day historical volatility on the SPY (or the VIX with a moving average to smooth it) and uses values in the 18 to 22 range as the cutoff.
The regime filter always applies to the index, not the individual stocks being traded. Whether or not a specific stock is above its own moving average matters less than what the broader market environment looks like. The question is whether you’re fishing in favorable or unfavorable water.
Ranking trades when signals outnumber positions
Mean reversion strategies generate lots of signals when the market sells off. If you’re running a strategy on a broad universe, you might have 40 or 50 potential trades on a given day but only capital for 10 to 15. This forces a ranking decision.
Cesar’s preferred ranking method is 100-day historical volatility. He favors stocks that are moving more, because higher-volatility stocks tend to generate bigger bounces. His reasoning: a bigger recent pullback usually comes with a bigger snap-back.
Secondary ranking options he uses:
- Recent 3-to-5-day return (prefer stocks that have dropped the most)
- Degree of mean reversion stretch (lowest RSI, furthest from a moving average, smallest percent-B reading)
In practice, 90% of the time he falls back on volatility ranking because it’s the most consistent predictor of which trades will produce the largest returns.
Position sizing: why Cesar prefers fixed percentage over volatility-based
There are two broad approaches to position sizing in a mean reversion portfolio: fixed percentage per position, or volatility-adjusted sizing that scales down more volatile stocks.
Cesar uses fixed percentage. His reasoning is straightforward: in his research, volatility-adjusted sizing rarely improves returns enough to justify the added complexity. With fixed percentage sizing, he typically targets 10% per position (10 positions) or 5% (20 positions).
The number of positions matters more than most traders realize. With only 5 positions, a single lucky or unlucky trade has a disproportionate impact on returns. Two traders using nearly identical strategies but different position counts can show very different results over a year, not because of skill, but because of which specific stocks happened to move. Cesar’s view is that 15 to 20 positions reduces luck’s role enough to make the strategy’s edge more visible. He personally trades between 10 and 12 positions.
Market vs. limit entries: when each makes sense
Cesar currently runs two mean reversion strategies, one entering at the market open and one entering on intraday limit orders. Each has tradeoffs.
| Entry type | Advantage | Disadvantage |
|---|---|---|
| Market open | Guaranteed fill, more opportunities | Slippage risk at the open |
| Limit intraday | Known exact fill price, no negative slippage | Missed trades if price doesn’t pull back to limit |
One observation from his trading: getting filled at the absolute low of the day is actually a bad omen. When the stock continues to the low of your limit and fills you there, it often keeps dropping the next day. He runs both entry types together to balance trade frequency against fill quality.
On ATR-based limits, Cesar has tested this several times and never found it adds enough value to justify the complexity. A simple percentage below the prior close (such as 2%) works just as well in practice.
Why Cesar doesn’t scale in to positions
Scaling into a mean reversion trade when it moves further against you is a common approach. Cesar has tested it and generally avoids it. His finding: overall returns tend to decline when you scale in, because you’re adding exposure to a position that’s already not working.
The one benefit of scaling in is that it raises your win rate, sometimes into the 70s to low 80s percent range. If a high win rate matters to you psychologically, that’s a legitimate reason to use it. But Cesar’s preference is to take the full position at the initial entry and let the trade play out.
Stops in mean reversion: why they don’t do what you think
Stop losses in mean reversion strategies are a contested topic. The intuitive case for stops is that they limit losses. The counterintuitive reality, based on Cesar’s research, is that stops often don’t protect you from the losses you’re most worried about.
The big losses in mean reversion tend to come from gap openings, where a stock opens well below where you placed a stop. The stop triggers at the opening price, which may already be 15 or 20 percent below your intended exit. The stop did nothing to prevent that loss.
At the same time, tight stops increase your trade frequency of getting stopped out on normal intraday noise, then watching the stock bounce without you. This degrades the strategy’s returns without meaningfully limiting catastrophic risk.
Cesar’s general approach is to use wide stops if any, or to manage risk through position sizing rather than stop placement. On exits, he focuses on time-based exits and profit targets, which work better with mean reversion than trailing stops.
Combining exits: what works and what to test
The two exit types that consistently work in mean reversion are time-based exits and profit targets. Cesar typically tests combinations of these rather than relying on a single exit rule.
Time-based exits work because mean reversion trades are designed to be short-duration. If the stock hasn’t bounced within 5 to 10 days, the trade thesis has probably played out or failed. Holding longer just ties up capital.
Profit targets make sense because mean reversion trades have a natural ceiling. You’re not looking for a trend to develop; you’re looking for a return to normal. Taking profit when that happens preserves gains and frees up capital for the next trade.
His recommendation on testing exit combinations: run a grid of different exit parameters and look for regions of stability rather than single optimal values. If a strategy only works with very specific exit settings, that’s a warning sign. If a wide range of parameter values all produce similar results, you have something more robust.
Get the show notes & transcript
Related episodes
- 127: Building Mean Reversion Trading Strategies with Cesar Alvarez (Part 1)
- “How to Build Mean Reversion Strategies” with PJ Sutherland
- “Effective Market Regime Techniques” with Cesar Alvarez
- “Low-Effort Trading Strategies” with Cesar Alvarez
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.
