The Earnings IV Ramp Is Not the Edge
Everyone knows implied volatility tends to rise into earnings. Far fewer traders realize that most pre-earnings straddles still lose money.
We backtested more than 21,000 pre-earnings straddles, and the result was blunt: the IV ramp by itself is not an edge. It is mostly math.
But a small filtered subset, selected using four earnings-specific signals, changed the picture completely. What looked like a breakeven grind became a seven-figure out-of-sample equity curve without holding through the earnings gap.
This article walks through the trade, the math behind the IV ramp, the signals that mattered, and why most traders who blindly buy pre-earnings volatility are still losing money.
The Trade Setup
The strategy is simple.
We buy at-the-money straddles ahead of earnings using the nearest monthly expiration after the earnings date. Entry occurs roughly two weeks before the announcement, and the position is exited before earnings are released.
That last point matters. This is not a bet on the earnings gap itself. We are not trying to predict whether the realized earnings move will exceed the implied move.
Instead, this is a long-volatility trade designed to capture repricing into the event.
At a high level, the position is long gamma and long vega into a known catalyst. That is attractive because positive-expected-value long-volatility trades are rare. Most of the time, long gamma is expensive. You are fighting theta decay and paying a premium for convexity.
The opportunity exists only when the earnings event appears underpriced. If implied event volatility is too low and gets repriced higher before the announcement, that repricing can more than offset the decay.
Why This Can Help a Short-Vol Portfolio
Most options traders spend much of their time short gamma or short volatility through spreads, short premium, structured products, or similar trades.
That creates a specific risk: if the underlying, the sector, or the broader market experiences a volatility regime shift, the portfolio can be on the wrong side of the move.
Pre-earnings long straddles can act as targeted hedges. Even if the specific earnings event does not reprice significantly, a marketwide volatility spike can lift ambient volatility. Since these positions are long volatility, they may offset some losses elsewhere in a short-vol book.
Again, the key is that the trade exits before earnings. The goal is to monetize volatility repricing, not the post-earnings gap.
Event Volatility vs. Ambient Volatility
To understand the trade, we need to separate two types of volatility.
Ambient volatility is the stock’s normal background volatility. It is what the stock might realize on ordinary non-event days.
Event volatility is the one-day volatility specifically assigned to the earnings announcement. It represents the market’s expected earnings move.
In variance terms, an expiration that includes earnings contains both components:
total variance = ambient variance + event variance
If an option expiry spans the earnings date, its implied volatility includes the regular day-to-day movement plus an extra one-day jump component for earnings.
When there is an expiration before earnings, that expiry can serve as a cleaner proxy for ambient volatility because it does not include the event. If no clean pre-event expiry exists, ambient volatility can be estimated using forward volatility between expirations that both occur after the event, since the shared event component largely cancels out.
Once ambient volatility is estimated, the event component can be backed out from the total implied volatility of an expiration that includes earnings.
That event volatility can then be translated into the implied earnings move traders often quote, such as “the market is pricing a ±6% move.”
Why the IV Ramp Is Mostly Mechanical
Many traders describe the strategy as “buying the IV ramp.”
The story sounds reasonable: implied volatility rises into earnings, so buy a straddle early and sell it later after IV increases.
But that is incomplete.
Imagine an option expiration that is 20 days away from earnings. It contains 19 ordinary ambient days and one high-volatility earnings day. As time passes, the number of remaining ambient days shrinks, but the earnings event remains inside the expiration.
Because the same event variance is packed into fewer remaining days, the annualized implied volatility rises.
That is the IV ramp.
But the ramp does not necessarily mean the earnings event itself is being repriced higher. It may simply reflect the mechanical effect of dropping lower-volatility days from the window.
So even though quoted IV rises, the straddle can still lose money because theta is burning and future variance is being consumed.
The actual edge is not the ramp. The edge exists only if the event volatility component itself is underpriced and later reprices higher.
The Dataset
To test this properly, we built a dataset of pre-earnings straddles.
The universe included liquid stocks with at least 20,000 average daily options volume at entry. Trades were opened as close as possible to 14 days before earnings, with a tolerance of plus or minus four days. The strategy used the nearest monthly expiration after earnings and exited before the announcement.
Commissions and slippage were included.
The final dataset contained about 21,500 trades starting from 2009.
The unfiltered strategy was not impressive. The return distribution was positively skewed, as expected for long volatility, but the median trade lost about 6.3%. The mean return was only slightly positive, around 0.3%, with a win rate near 37%.
That is the classic long-volatility profile: many small losers and fewer large winners.
The Signals That Mattered
We tested a range of signals, including IV percentile, term structure slope, skew slope, and several earnings-specific metrics.
The strongest signals were relative-value measures comparing the current implied move to prior earnings behavior.
The final model focused on four types of relationships:
-
Current implied move versus the previous earnings implied move
-
Current implied move versus the previous realized earnings move
-
Current implied move versus the average historical implied earnings move
-
Current implied move versus the average historical realized earnings move
The consistent pattern was simple: lower current implied moves relative to history tended to produce better returns.
Raw implied move also had some predictive value, with smaller implied moves generally performing better, but it was less stable across regimes. Because relative signals provide better context across names and market environments, the raw implied move was dropped from the final model.
Skew slope, by contrast, was weak and noisy. Its decile plot was not meaningfully monotonic, which made it a poor candidate for the model.
Stability and Out-of-Sample Testing
A signal is not useful if it only works in one period.
To test stability, we split the dataset into four-year blocks and reviewed signal behavior across time. The relative implied-versus-history signals remained reasonably monotonic across regimes.
We also checked for collinearity. The features had some moderate correlations, but variance inflation factors were benign, with the highest around 1.85. That suggested the model was not simply using duplicate versions of the same signal.
The main test was walk-forward optimization.
The process worked like this:
Use 2009–2010 as the first in-sample period, fit the regression, then trade 2011–2012 out of sample. Next, expand the training window to 2009–2012 and trade 2013–2014 out of sample. Repeat this process in two-year blocks.
This mirrors live trading: fit the model on available history, then trade forward without leaking future data.
Across walk-forward windows, the coefficients changed slightly but generally kept the same signs and similar magnitudes. Out-of-sample correlations between predicted and realized returns were consistently positive, and top-minus-bottom decile spreads were also positive.
The model was not perfect, but it showed stable predictive power.
Filtering for Positive Expected Return
Next, we filtered the out-of-sample trades to include only those with a predicted return greater than zero.
The improvement was meaningful.
The mean return increased from about 0.3% to 3.3%. The median improved from roughly -6.3% to -3.6%. The win rate rose from around 37% to about 42%.
The strategy still lost more often than it won, but the winners became large enough to create a positive edge.
That is exactly what a long-volatility strategy should look like when it works.
Kelly Sizing and Risk
Using the filtered out-of-sample trades, the estimated Kelly fraction was about 48.5%.
That sounds enormous, but it is less surprising for a long-debit strategy where the maximum loss is capped at the premium paid.
Still, full Kelly would be far too aggressive in practice. A much smaller fraction of Kelly is more realistic, especially given the variance, drawdowns, and uncertainty around future performance.
In practice, single-digit percentages of Kelly are more sensible.
Backtests using different Kelly fractions showed the expected trade-off: higher Kelly produced higher raw returns, while lower Kelly improved Sharpe ratios and reduced drawdowns.
Implementation
A practical daily workflow would look like this.
For each stock with an upcoming earnings event in the target window, compute the four earnings-relative inputs. Feed those values into the regression model to estimate the expected return of the pre-earnings straddle.
Then filter for trades with positive predicted returns, rank them by expected return, and apply liquidity and capacity constraints.
Once a name passes the filter, enter a long at-the-money straddle using the nearest monthly expiration after earnings. Entry should be around 14 days before the announcement, with some flexibility depending on liquidity.
Exit before the earnings release, ideally on the announcement day before the event or the day before if liquidity and spreads make that preferable.
Optional delta hedging or gamma scalping can be added during the holding period, but it is not required to benefit from volatility repricing. It also introduces additional complexity and transaction costs.
Important Caveats
This is a long-volatility strategy. Expect many small or moderate losers and occasional large winners.
The edge appears over a large number of trades, not on any single position. Drawdowns can last for months before a large winner pushes the equity curve to new highs.
The backtest also did not assume holding through earnings. Holding through the event changes the strategy entirely because it introduces direct exposure to the realized earnings move.
Conclusion
The earnings IV ramp is real, but it is not automatically profitable.
Much of the ramp is mechanical: as ordinary ambient days fall out of the option’s remaining life, the same earnings event variance is compressed into fewer days, causing annualized IV to rise.
The real opportunity is not buying the ramp blindly. It is identifying cases where the current implied earnings move is cheap relative to that stock’s own earnings history.
In our test, a simple model using relative implied and realized earnings-move signals produced a positive out-of-sample edge across more than 21,000 trades.
The takeaway is straightforward: the ramp is math. The edge is mispricing.
