A short-put record with no negative month across 2008 and 2020 says more about return construction than trading.
Miao and Sorokina report a Sharpe of 7.81 for an always-on, cash-secured put-writing book. Annualized return reaches 18.7% with 2.4% annualized volatility and a 0.0% maximum drawdown over 300 months from 2000 to 2024. The equal-weight, long-only portfolio drawn from the same universe returned 15.4% at 19.1% volatility, for a Sharpe of 0.81. Its peak-to-trough loss was 47.0%. One dollar compounds to $102 against $24. All 25 calendar years were positive. That includes 2008, when the strategy made +26% and the stock benchmark lost 37%.
The conclusion is built around the Sharpe. The book outperforms the long-only benchmark "on a risk-adjusted basis by a factor of ten on the Sharpe ratio before transaction costs and the risk-free rate." The claim worth examining is therefore 7.81 against 0.81, gross of costs and with collateral earning zero. The 3.3-percentage-point return gap matters far less.
Our data allow a much narrower check. The options history is end-of-day, which prevents verification of intraday stop and take-profit fills at the first qualifying price. We substitute end-of-day marks with modeled slippage. The paper instead uses OptionMetrics best bid/offer mids. Our coverage begins around 2020, leaving the 2000-2024 record beyond reach, and we would use a pre-specified ticker list in place of the paper's fixed 45-name survivor set. Nothing below reproduces their numbers or should be read as having tried.
Options on U.S. equities routinely carry implied volatility above the volatility subsequently realized. A seller collects that wedge while accepting the downside below the strike. Miao and Sorokina ask whether the same wedge appears in nuclear and energy-adjacent single names. The paper cites Bakshi and Kapadia, Carr and Wu, and Bollerslev, Tauchen and Zhou on the premium in equity index options. For individual names, it says, the case "has not been studied."
Their universe contains 45 tickers gathered from NRC plant-ownership records and nuclear ETF constituents, including CEG, NEE, CCJ, UEC, NXE, UUUU, BWXT, NRG, SMR and similar. Usable option coverage exists for 39. A GARCH(1,1) supplies the realized-volatility forecast. Alpha and beta remain fixed at 0.09 and 0.90, while each stock's long-run term is calibrated to its historical variance. The at-the-money implied proxy uses open-interest-weighted IV from puts with delta between 0.40 and 0.60 and 15 to 45 days to expiry.
Each entry date adds one put per name, scored toward a 0.30 delta and 45 DTE target. Eligible contracts have DTE from 35 to 55, absolute delta from 0.20 to 0.40, and open interest above 10 contracts. The first of three conditions closes the trade: premium declines to half the entry mid, premium doubles and triggers the stop-loss, or DTE reaches 15. Every position is cash-secured against the strike. Trade return equals P&L divided by strike, with P&L calculated from entry mid minus exit mid.
This process generates 64,514 trades, with 30 to 43 active names in a typical month. The win rate is 88.3%. Annual figures range from 84.6% in 2008 to 92.7% in 2004.
The premium survives a harder look
The paper's volatility risk premium evidence is worth keeping. Across entry days, mean ATM IV divided by GARCH RV is 1.53 on an equal-weighted ticker basis and 1.44 when pooled. Twenty-eight names have mean IV minus forward realized vol at t above 2. NRG reaches 12.7 from 1,323 observations.
Miao and Sorokina then address a problem many papers leave alone. The 21-day forward realized-vol windows overlap, meaning adjacent observations share about 21 days of data. Effective n can consequently be overstated by up to a factor of 21.
A conservative adjustment divides the naive t-stats by 4.6. NRG remains at roughly 2.8, NXE at 2.6 and UUUU at 2.2. Those three are the only names the paper recomputes, and the rest of the cross-section is not restated after the haircut. NRG has the largest naive t-stat. During the sample, it belongs in the universe as a diversified power producer with nuclear plant ownership stakes.
The result is modest and plausible. Small pure-play and SMR names such as DNN, LTBR, OKLO, UEC have the highest IV/RV ratios. Diversified utilities including SO, AEP, DUK, ETR have the lowest.
The pooled premium was negative in six of 25 years (2000, 2002, 2008, 2011, 2018, 2024), and its worst shortfall was 6.2 percentage points in 2008.
How the curve becomes so smooth
All six negative-premium years still delivered put portfolio returns between 12.1% and 30.8%. The paper credits positions opened during earlier high-premium periods that expired profitably through those episodes. The return definition deserves to be read beside that explanation.
A monthly portfolio return is the equal-weight mean of ticker-level average trade returns for trades closing during the month. P&L gets averaged twice, first within each ticker and then across 30 to 43 names. Every per-trade percentage uses its own strike as the denominator. Averaging thousands of those percentages compresses dispersion, making 2.4% annual volatility arithmetically unsurprising. The same double averaging produces the 0.0% drawdown.
Losses are truncated twice over. A stop at twice the entry premium limits a losing trade to about one premium unit. The take-profit captures half of one. The rough payoff is therefore 1:2 against an 88.3% hit rate.
The authors explain the smooth curve in their own words: "Individual trade losses are hard-capped by the 2x stop-loss; combined with the 88% win rate and cross-sectional averaging, no negative monthly portfolio returns occur in this sample." All three ingredients operate at trade level. Open positions receive no mark. The paper also acknowledges that exits assume a fill at the first daily mid meeting the condition, "which may understate gap risk on adverse overnight moves."
There is one inconsistency. The win-rate section counts 64,514 trades "across all 45 tickers" while the data section says 39 tickers enter the backtest.
A richer signal loses 5.2 Sharpe points
Requiring IV over GARCH RV to exceed 1.1 reduces the trade count from 64,514 to 11,404. Return drops to 10.9%, volatility climbs to 4.2%, the Sharpe falls to 2.63, and a 7.7% drawdown appears. Raising the threshold to 1.2 makes the result weaker again: 6,693 trades, a Sharpe of 2.03 and a 9.4% drawdown. Volatility reaches 5.3%, compared with 4.2% at the 1.1 threshold and 2.4% unconditionally. The higher per-trade edge leaves the strategy 5.2 Sharpe points behind the always-on book.
The authors attribute the decline to lost diversification, which explains part of it. I see a diagnostic. A headline Sharpe generated by the premium should survive some concentration in the richest signals. Here it loses 5 points. The drawdown arrives when the cross-section contracts from 30 to 43 names to just 5 to 15.
We have previously covered the mirror image, a high-Sharpe book whose abstention gate never bound out of time (note). This gate binds, and the Sharpe disappears with it.
The paper states the weaknesses plainly
The abstract says performance is measured before transaction costs and on a fixed universe. The conclusion is unusually direct. It identifies mid-price fills throughout, collateral earning zero, survivorship in the universe, and design choices that were never jointly optimized. The universe, the authors write, is "fixed at today's list of survivors." They add that a point-in-time reconstruction "is needed before the performance claims can be treated as investable." Their warning should govern any reading of the headline figures.
In the paper's phrasing, realistic bid-ask execution costs for short puts "can be substantial for less liquid names" and "would reduce reported returns." The strategy crosses that spread twice for every trade across 64,514 trades. Using a zero risk-free rate flatters the Sharpe rather than the return.
Capacity receives no treatment. Open interest above 10 contracts is the sole liquidity screen in the eligibility rules, while the largest IV/RV ratios occur in the small pure-plays DNN, LTBR, OKLO and UEC. Whether those option books can take institutional size is our question, and the paper does not answer it. We also did not find a comparison with a generic short-put benchmark or a matched non-nuclear universe. Such a comparison would separate the sector claim from the equity VRP already documented elsewhere.
One test would change my mind. Build a capital-based monthly equity curve, mark open positions to market, charge spread-crossing, and fill stops at the next available print. If 2.4% volatility survives, the premium is real and larger than I think. If the drawdown rises from 0.0% to something a risk committee recognizes, the 7.81 came from the average of averages all along.