VRP Short Put Spreads: Honest 7-Year Backtest, 5 Symbols (2026)

Search "volatility risk premium strategy" and you will find a hundred posts with an equity curve going up and to the right and a win rate north of 80%. They are almost all built on three quiet lies: mid-price fills, clean profit-target exits, and parameters that were chosen on the same symbol they are advertised on. This study removes all three.

We took a single short-put-spread VRP harvest, tuned it once on SPY in prior work, froze every parameter, and ran it unchanged across QQQ, IWM, AMZN, and NVDA from 2019 to 2026. We then added SPXW as a flagship out-of-sample check via a completely separate daily-resolution data path. Execution was modeled the way a real account experiences it: post-and-wait limit orders, exits that have to cross the spread, a stale-quote guard, and a signal that can only see the past.

This is the most honest thing we can publish about VRP harvesting. The 70% win rate the internet sells is real. The risk-adjusted edge that survives honest execution and out-of-sample deployment is close to zero.

TL;DR - the cross-sectional table

Period: 2019-02-01 to 2026-04-02 (7.16 years). $100k notional, unlevered Half-Kelly.
† SPXW is sourced from the Historical API at daily-close resolution, not 1-minute. Its fill rate is 100% by construction (it transacts at daily-close mid less the measured −$0.04 haircut, not a post-and-wait limit). It is a valid out-of-sample direction confirmation but its Sharpe and fill rate are not methodologically comparable to the 1-minute rows. See §5.7.

1. Why this study exists

The volatility risk premium is real. Implied volatility has, on average, exceeded subsequently realized volatility for decades. The economic rationale (insurance demand, dealer inventory) is sound. The open question was never "does VRP exist." It is: how much of that premium survives honest execution when you don't get to re-fit the strategy to every new market?

The genre of VRP content you find online almost always rests on three assumptions that real accounts never enjoy.

This study answers exactly that question, and nothing more. The takeaway is not that VRP is untradeable. The takeaway is that the honest, transferable, retail-executable edge is much smaller than the genre advertises - and the win rate is a misleading place to anchor your expectations.

2. The strategy (frozen specification)

A short put credit spread harvest on the underlying, gated by a market-regime VRP signal. Every number below was fixed before the cross-sectional run and never changed per symbol.

The signal in detail. A daily market-regime classifier built from VIX, VIX9D, VVIX, the VIX term structure, a realized-vs-implied VRP proxy and its 252-day percentile, high-yield spreads, and stress z-scores. Crucially, this is a market-wide signal computed from SPY/VIX. Applying the identical signal to QQQ, IWM, AMZN, and NVDA is what makes those four a genuine cross-sectional out-of-sample test rather than four separately fitted strategies. The signal construction is leak-free by design - for the methodology see Historical VRP Percentiles Without Lookahead Bias.

3. The honesty controls (what makes this citable)

This is the entire point of the study. Each control below removes a specific way backtests lie.

1. Leak-free signal

   The signal CSV is shifted forward one calendar day with a 7-day walk-back lookup. Any trading day's decision uses only data fully observable before that day's open. There is no way for the strategy to "know" today's VRP when deciding to trade today.

2. Honest fills (post-and-wait limits)

   We do not fill at mid. We post a limit at the offer plus a few cents (ask_edge) and wait for someone else's order to cross our price. If nobody does within the wait window, the order is cancelled and we re-rank. This is why the fill rate is 3-14% and not 100%.

3. Stale-quote guard

   When a cross is detected, we re-check that the mid hasn't moved through our limit by more than a floor (−$0.05). A one-tick bid blip during a vol spike does not get to manufacture a phantom fill.

4. Patient-then-cross exits

   Profit-target and stop-loss exits are not free. We post a buy-to-close limit at the trigger and wait a few bars. If it fills, great. If it doesn't, we cross the spread and pay the offer - and that exit is tagged pt_x / sl_x so the execution tax is auditable. It is large. See §5.3 and §5.5.

5. EV-blind tiebreak

   When multiple candidate spreads cross on the same bar, the winner is chosen by a timestamp-seeded random shuffle - never by which one had the best expected value. Any EV-aware tiebreak is a look-ahead oracle that silently inflates win rates.

6. Frozen, never re-fit

   The configuration in §2 was tuned once, on SPY, in prior work. For this study it was frozen. QQQ, IWM, AMZN, and NVDA received zero per-symbol optimization. What you see on those four is what a trader who tuned on SPY in 2019 and walked away would actually have experienced.

4. Data

All five intraday symbols carry a dense weekly/daily expiry ladder, so the 14-DTE entry window is continuously reachable. SPXW is handled separately (§5.7): its 1-minute weekly chain was not available locally, so it was sourced from the FlashAlpha Historical API at daily-close resolution and is reported as an out-of-sample direction check, not a sixth execution-comparable row. The reasons minute-level options analytics are this hard to assemble are catalogued in Why Historical Options Analytics Are Rare.

5. Findings

5.1 The win rate is stable. That is the headline everyone stops at.

Across five very different underlyings - a broad-market ETF, a tech ETF, a small-cap ETF, and two megacap single names - the win rate clusters tightly at 64-72%. The VRP edge in probability terms is robust and travels well cross-sectionally. If you stop reading here, VRP harvesting looks like a free lunch. This is exactly where most articles stop.

5.2 The risk-adjusted edge does not survive honest, frozen, out-of-sample deployment.

Two of four out-of-sample symbols lost money. The best non-tuned result (NVDA, Sharpe 0.22) is still a Sharpe you would not trade. The premium is real; the tradeable, honest, transferable edge is close to zero.

5.3 Why a 70% win rate nets to nothing: the payoff is brutally asymmetric.

A 70% win rate at a 1:2.3 win/loss ratio is, by arithmetic, a coin flip. This is the defining feature of short-vol: you win small, often, and lose big, occasionally - and it is precisely why the win rate is the wrong number to look at. The strategy is selling insurance; the premium and the claims very nearly cancel.

5.4 Only 3-14% of orders filled. And the fills were adversely selected.

A backtest assuming mid-fills books all of the proposed trades at a better price. Reality fills a single-digit-to-low-teens percentage of them, and the ones that do fill cross at roughly 4 cents worse than mid - because the orders that fill are disproportionately the ones the market is running through (adverse selection). Per spread, on 100-multiplier contracts, that is a structural ~$4 headwind on every fill before the trade even begins.

5.5 The "profit target" is mostly a forced spread cross.

Close-reason breakdown (pt = clean limit fill at target; pt_x = target hit but had to cross the spread to actually get out):

On every symbol, more profit-target exits required crossing the spread than filled cleanly at the limit. The idealized "close at the 50% target" that naive backtests book is, in practice, the minority outcome. This single effect - invisible in any mid-fill backtest - is a primary driver of the gap between the seductive win rate and the unimpressive Sharpe.

5.6 Year by year: the signal helps in crises, and the "full-risk" regime doesn't earn its name.

Net P&L by entry year (selected):

The signal's risk_off gate genuinely cushioned 2020 (it sidestepped the worst of the COVID crash and re-entered into the rebound). But 2022 - a slow grinding vol-elevated bear - punished the single names hard even with the gate, because the strategy still traded in neutral/risk_on days while drift went against short puts.

The most uncomfortable finding: the signal's risk_on ("deploy full size") regime did not outperform neutral. In fact it underperformed on most symbols:

5.7 The flagship instrument, fully out-of-sample, confirms it

SPXW - SPX weeklies, the single most-traded options complex in the world and the canonical home of the VRP trade - is the cleanest out-of-sample test available: the strategy was never tuned on it, and it had to be sourced from a different data path entirely (the FlashAlpha Historical API, at daily-close resolution, because no 1-minute SPX weekly chain was available locally).

It lost money too.

Two things matter here. First, the same "risk_on underperforms" pattern from §5.6 reappears independently - on SPX weeklies, sourced from a different feed, the "deploy full size" regime is again where essentially all the losses are (−$5,266 of −$4,641 net came from risk_on; neutral was net positive). A pattern that survives a complete change of instrument and data pipeline is not noise. Second, the win/loss asymmetry is even worse (1:4.4) because narrow 5-point SPX spreads collect tiny credits against a wide max loss - the short-vol payoff shape in its most unforgiving form.

6. What this means

1. VRP is real; the free lunch is not.

   The premium shows up reliably as a high win rate everywhere. It does not reliably show up as money once you (a) pay realistic execution and (b) refuse to re-fit per symbol.

2. The win rate is the most misleading number in options content.

   A 70% win rate with a 1:2.3 payoff is structurally breakeven. Any post leading with win rate and not showing the loss distribution is selling you the setup, not the result.

3. Execution is the strategy.

   The single biggest gap between the hyped version and this one is not the signal - it is fills. Mid-fill assumptions, clean-target exits, and 100% fill rates are where the fictional returns live. See the companion piece VRP Backtest: The Fill Model Is the Edge for the controlled fill-model sensitivity study.

4. Out-of-sample is brutal and necessary.

   SPY (tuned) was the best result. Every symbol the strategy had never seen did worse, two of them losing. Honest deployment is the non-tuned result, not the tuned one.

5. Don't lever the confidence.

   The risk_on regime - the one a leveraged variant would size into hardest - was the weaker regime. This is the precise mechanism by which "+5,000% backtest" leveraged VRP posts blow up.

This is not an argument that VRP is untradeable. It is an argument that the honest, transferable, retail-executable edge is small, regime-fragile, and nowhere near what the genre advertises - and that anyone trading it should size for breakeven-with-tails, not for the win rate.

7. Limitations and caveats (read before citing)

• Out-of-sample in symbol, not in time. The signal's thresholds were originally chosen on a 2018-2026 SPY window. Deploying the frozen config on four other symbols is genuine cross-sectional OOS, but it is not walk-forward-in-time OOS. We do not claim it is.
• This is the unlevered honest reference. Half-Kelly, default 0.05, cap 0.25 deploys little capital, so absolute CAGR and drawdowns are small (1-8%). A leveraged variant of this exact strategy shows a far larger in-sample CAGR - that figure is explicitly not reported here because it is in-sample and not OOS-validated. Do not cite a leveraged number against this study.
• Sharpe is trade-level (per-trade Sharpe × √trades-per-year), internally consistent across these rows but not directly comparable to a daily-return buy-and-hold Sharpe.
• IWM is statistically thin (58 trades / 7 yr; 3.1% fill rate). Treat its negative Sharpe as directional, not conclusive.
• SPXW is a daily-resolution, out-of-sample direction check, not an execution-comparable row. It is sourced from the FlashAlpha Historical API at daily-close cadence, transacts at mid-less-measured-haircut (100% fill by construction), and its exits are all forced crosses. Cite it for "the frozen strategy also loses on the flagship instrument and the regime pattern reproduces," never for a comparable Sharpe or fill rate. Full treatment in §5.7.
• No commissions/fees are modeled; a real account's results would be worse, not better. This strengthens, not weakens, the central finding.

8. Reproduction

Everything in this article regenerates from two commands:

bash run_cross_symbol.sh                              # 5 symbols, 1-min local chains
python spxw_api_backtest.py --start 2019-02-01 \
       --end 2026-04-02 --label vrpfrozen             # SPXW, daily-close via API (§5.7)
python aggregate_cross_symbol.py                      # cross_symbol_summary.{csv,json}

Per-symbol engine invocation (identical except --symbol):

python intraday_bt_ev_rank.py --symbol <SYM> \
  --start 2019-02-01 --end 2026-04-02 \
  --delta 0.10 --dte 14 --pt 0.50 --sl 1.0 \
  --vrp-signal vrp_signal_v2.csv \
  --kelly-default 0.05 --kelly-mult 0.5 --vrp-on-mult 1.0 \
  --kelly-max 0.25 --max-drawdown 0.30

Raw artifacts: evrank_<sym>_vrpfrozen_{summary.json,trades.csv.gz,equity_curve.csv}. The SPXW row consumed the same Historical API surface a paying user would call. See the Historical API documentation and our broader replay infrastructure write-up: Historical Options Analytics: Replay GEX, VRP, Dealer Positioning.

Appendix A: Full metrics table

Generated 2026-05-17 from the frozen-config cross-sectional run. $100k start, unlevered.
† SPXW: daily-close resolution via the Historical API; 100% fill by construction (transacts at mid less the −$0.04 measured haircut, not a post-and-wait limit); exits are all forced crosses so PT-hit rate is structurally 0. Out-of-sample direction check only - see §5.7. The other five are 1-minute MM-fill from local chains.

Frequently Asked Questions

Related Reading

• VRP Backtest: The Fill Model Is the Edge - the companion fill-model sensitivity study that re-frames the result in this article as the pessimistic bound of a range
• Historical VRP Percentiles Without Lookahead Bias - how the leak-free signal in §3.1 is constructed
• Historical Options Analytics: Replay GEX, VRP, Dealer Positioning - the replay infrastructure behind every number above
• Effective Open Interest Methodology - our standard for surfacing the methodology behind a derived signal, end to end
• Gamma Exposure (GEX) Explained - why dealer hedging is the structural reason VRP exists in the first place
• Why Historical Options Analytics Are Rare - the build-vs-buy math behind minute-level greeks history
• Historical API documentation - the same data path the SPXW row consumed