“Beyond Delta: A Data-Driven Strike System for Safer Weekly Put Selling”

Empirical probability, volatility mispricing, and the 90% strike edge

Table of Contents

Introduction: Strike with Precision
A new framework for high-probability options trading — grounded in data, not delta.

1. The Fundamental Misconception
   Why delta is not a probability — and how retail logic got it wrong.
2. The Mathematical Reality Behind the Illusion
   The structural flaws of delta and the limitations of Black-Scholes assumptions.
3. The Implied Volatility Distortion Engine
   How crowd-driven volatility inflates delta and misprices risk.
4. Proof of Concept: Results First
   2,000+ trades. 90 %+ success. A system that delivers what delta promised.

• Practical Benefits of the 90% Strike System vs. Delta

5. The Three-Strike System
   How historical data and real-time probability combine to set smarter strikes.

• Close-to-Close Historical Strike
• Open-to-Close Historical Strike
• The Band Strike (ML-Weighted)

6. How I Use Three Strikes, Probability and Premium
   Using XIRR to evaluate trade quality — and how intraday price moves adjust the strike.
7. Managing the Inevitable: When Strikes Break
   A disciplined, four-part framework for assignment and recovery.
8. Conclusion: From Model Illusion to Market Reality
   Why the 90% strike system works — and what Delta never understood.
9. References
   Academic and empirical research support the framework.

1. The Fundamental Misconception

For decades, a simple rule of thumb has dominated retail options education: “Sell the 10-delta put to give yourself a 90% chance of success.” It appears everywhere — trading forums, brokerage platforms, YouTube tutorials — and gives traders a comforting sense of mathematical precision. But this guidance rests on a critical misunderstanding of what delta actually represents.

Delta, by definition, measures how much an option’s price is expected to change in response to a $1 move in the underlying stock. A delta of 0.10 implies that the option’s price will rise or fall by $0.10 for every $1 movement in the stock. However, over time, delta evolved in trader lore into something else entirely: a shortcut for probability.

The idea isn’t entirely without basis. For short-dated, at-the-money options, delta can sometimes approximate the probability of finishing in the money. This approximation led many to assume that a 10-delta put carries only a 10% chance of assignment — and therefore a 90% chance of keeping the premium. On the surface, it sounds scientific and straightforward. But this shortcut quickly becomes a trap.

The deeper issue is that delta was never meant to represent probability. Using it this way leads to systematic errors — especially in fast-moving or volatile markets. The assumption that a 10-delta put implies 90% odds of success holds up only under a narrow set of conditions — most of which do not exist in real-world trading.

• Delta is a hedge ratio, not a forecast. It tells a market maker how many shares to hold to hedge an option position — not how likely it is to expire in the money.
• Delta assumes a model world, not a real one. It is derived from the Black-Scholes formula, which relies on idealized assumptions such as constant volatility, frictionless execution, and log-normal price movement.
• Delta relies on implied volatility — a crowd-driven estimate, not a statistical truth. It reflects trader sentiment, skew, and supply and demand pressure — not objective risk.

Yet traders continue using delta as a probability guide — not just near the money, but far out of the money, where its reliability deteriorates sharply. This leads to misplaced confidence, misaligned strike selection, and mispriced risk.

This is not a theoretical concern. It affects real trades, real money, and real outcomes. Moreover, it sets the stage for the deeper structural flaws we examine next — including how delta’s mathematical foundations diverge from real-world probabilities, and how implied volatility further distorts the picture.

2. The Mathematical Reality Behind the Illusion

To understand why delta fails as a probability measure, we need to look at its mathematical roots — and where they diverge from real-world behavior.

In the Black-Scholes framework, the delta of a call option is defined as N(d₁) — the value of the standard normal cumulative distribution function applied to a specific input. Traders often interpret this value as the likelihood the option will expire in the money. But this interpretation is fundamentally flawed.

If we want to assign an actual model-based probability of an option finishing in the money, the correct term is N(d₂) — not N(d₁). The distinction is important:

• N(d₁) reflects the hedge ratio — how much of the underlying asset a trader needs to hold to remain delta-neutral
• N(d₂) is the model’s estimate of risk-neutral probability — that is, the likelihood of finishing in the money assuming the asset grows at the risk-free rate and markets are frictionless.

Even N(d₂), however, is not a real-world forecast. It is a risk-neutral probability — a mathematical construct used to eliminate arbitrage in pricing, not to predict outcomes. It assumes that all assets grow at the same rate (the risk-free rate), that returns are normally distributed, and that volatility is constant. None of these assumptions hold in practice.

Real markets deviate from the model in significant ways:

• Returns are not normally distributed. Stock returns exhibit fat tails, jumps, and negative skew — particularly during market stress.
• Volatility is not constant. It clusters, spikes during periods of uncertainty, and exhibits asymmetric behavior on the downside.
• Prices move in discontinuous paths. Gaps in earnings, macroeconomic events, or geopolitical shocks violate the smooth diffusion process that the Black-Scholes model assumes.

These deviations are not just theoretical concerns; they have been extensively documented in the academic literature. Salami (2024) conducted an empirical examination of the Black–Scholes model using U.S. stock market data and found substantial discrepancies between model-derived prices and actual market prices, particularly for put options — precisely the instruments most retail traders rely on for income generation. Bates (2021) reinforces this critique, demonstrating that traditional models fail to capture the stochastic volatility and jump risks inherent in real markets. His research advocates for empirical modeling approaches that reflect actual market distributions rather than theoretical assumptions.

This empirical evidence exposes a fundamental problem: delta, derived from a flawed pricing model, cannot reliably estimate real-world probabilities. When the underlying mathematical framework fails to capture how markets actually behave, any probability estimate derived from that framework becomes suspect.

The gap between delta’s theoretical meaning and actual market behavior has been well documented:

• Hull & White (2017) proposed an alternative “minimum variance delta” because traditional delta often fails to minimize hedging error.
• Fan & Mancini (2009) found that nonparametric methods far outperformed delta-based models, especially in the tails of the distribution.
• Campuzan (2023) explicitly noted that N(d₁) is a hedge ratio — not a probability — and that even N(d₂) remains anchored in a synthetic pricing world.

This matters because delta does not just give traders a false sense of precision — it gives them the wrong framework altogether. A 10-delta put may look like it has a 90% chance of success on screen, but that number is based on model assumptions that fail under real-world stress.

In reality, delta is a model-derived guess, not a measurable statistic. And when that model breaks — due to volatility, skew, or jump risk — so does the strike selection logic built on it.

Next, we will look at the most unstable part of this equation: implied volatility — the variable that feeds delta its false confidence.

3. The Implied Volatility Distortion Engine

Delta’s failure as a reliable probability measure primarily arises from its dependence on implied volatility (IV). While IV is often seen as a forward-looking prediction of market movement, it is fundamentally different—it represents market sentiment and demand for protection, not an objective statistical forecast.

Why IV Misleads Traders

Implied volatility is not directly observed; it’s derived from market prices. Traders plug current option prices into the Black-Scholes formula to estimate volatility, meaning implied volatility (IV) inherently reflects emotional biases, speculative positioning, and market fear rather than actual future volatility.

This distortion is systematic, not random:

• IV consistently exceeds realized volatility: Research clearly shows IV typically overestimates actual future market movements, particularly during stable market conditions (Christensen & Prabhala, 1998; Dierkes et al., 2024).
• Options are structurally overpriced: High demand for downside protection (puts) artificially inflates IV, making OTM puts seem riskier than historical evidence supports.

The evidence for this mispricing extends well beyond isolated studies. Goyal and Saretto (2007) conducted a comprehensive analysis of option returns and found significant systematic discrepancies between implied and realized volatility across equity options. Their research demonstrates that strategies exploiting these IV-RV mispricings generate substantial abnormal returns — particularly with delta-hedged options — proving that theoretical volatility assumptions consistently fail in practice. This directly validates our approach of abandoning IV-dependent delta calculations in favor of empirical probability data.

Further reinforcing this critique, Chelikani, Kilic, and Wang show that implied volatility spreads between calls and puts serve as predictors of underlying stock mispricing. When IV spreads are distorted by market sentiment, the resulting option prices — and their derived deltas — become unreliable indicators of actual risk. This finding underscores a critical point: IV reflects market psychology and demand imbalances, not statistical probabilities.

Academic research supports this pattern: Solbakke & Egly (2022) found that historical volatility outperformed implied volatility in predicting future market movements in Sweden’s OMXS30 index, reinforcing that backward-looking data often proves more reliable than forward-looking market sentiment.

A Clear Example: TSLA

Suppose a trader sells a 0.10 delta put, trusting the “90% success” indicator from a brokerage platform. However, suppose IV is elevated due to events such as earnings or geopolitical news. In that case, historical data may indicate that the actual likelihood of TSLA breaching this strike is closer to 18%—nearly twice what delta implies. This mispricing creates substantial hidden risk for traders relying on delta for strike selection.

The Practical Reality

Delta, dependent on IV, becomes unreliable precisely when accuracy is most critical—during market stress or volatility spikes. Traders are left with false confidence, mispriced options, and poorly chosen strike prices. In contrast, our 90% strike system eliminates IV as a variable entirely, using actual historical outcomes and current probability data rather than market sentiment.

4. Proof of Concept: Results First

If delta and implied volatility are unreliable tools for strike selection, what’s the alternative?

We tested that question across over 2,000 real-world trades using Version 2 of Probability Trader Pro, a data-driven approach built on historical probability — not theoretical assumptions. The results:

This wasn’t a paper exercise. These trades were executed live, with real money, across dozens of tickers, using a consistent system: one that starts by asking a simple, measurable question:

“How often has this stock actually closed below a certain level in the past?”

That question led us to build the 90% strike system — a strike-setting framework based not on what option models assume, but on how stocks have actually behaved over time.

Empirical, Not Hypothetical

Instead of guessing probabilities from model inputs like delta or IV, the 90% strike system anchors its logic in reality. It uses historical return distributions — segmented by day of the week — to identify the 10th percentile closing return. This forms the baseline strike: the level a stock has breached only 10% of the time over the past 10–15 years.

This empirical approach aligns with recent academic findings. Wu, Hsu, and Wang (2021) demonstrate that probability-based frameworks significantly outperform traditional delta-based approaches for options trading. Their research shows that empirically derived probabilities provide superior accuracy in identifying market trends and directional moves compared to theoretical model assumptions. Critically, they argue that static approaches like delta-based strikes fail to account for time decay, short-term volatility clustering, and non-linear price responses — exactly the real-world factors our historical probability system captures.

It doesn’t rely on volatility assumptions. It doesn’t care what delta says. It simply asks:
What has happened before — and how often?

Why It Works

• Market behavior is asymmetrical. Stocks fall faster than they rise, and downside moves are often clustered in certain timeframes or on certain days of the week. This observation aligns with decades of research on day-of-week effects, including studies by Drogalas et al. (2007) and Lakonishok & Smidt (1988), which found persistent weekday return patterns across global markets.
• Delta doesn’t see that. However, historical data does — and it can be quantified.
• Real-world probabilities are measurable. With enough data, we can know exactly how often a stock has breached a given level within a 5-day period.
• That’s what our system does. And it does so without a single line of option math.

From Version 2 to Version 3

In July, we decided it was time to expand our probability acumen and improve our predictive ability on the probability side of the equation (see our white paper on Probability).  This major revision was completed from August 2024 to May 2025, over 9 months. We now have and are actively testing Version 3 of our probability predictions. Why? Because, as accurate as the 90% strike is, a higher probability stock will outperform a historical 90% strike. And, for those times when the stock is assigned to us, we have a better post-management system to regain profits in an otherwise lost trade quickly.

In Version 2 of our probability system:

• We used five base models (for the weekly probability predictions) and a one-strike structure (90% strike probability system), based on historical deciles.
• The results, as provided above, were robust, but as model calibration improved, we identified room for further refinement.

In Version 3 (current):

• We expanded to 8 base models, three stacked meta-models, and one final weighted ensemble model, and introduced a three-part strike system:
  1. Historical 10th percentile by day-of-week (end of day)
  2. Historical 10^th percentile by day-of-week (opening of market day)
  3. Probability overlay based on machine learning (one 4-day prediction)
• This provides us with even more precision, especially when market conditions shift midweek.

Since testing version 3 with real trades from May 23, 2025, to June 6, 2025, we have 53 new trades. The results?

• 53 ‘PUTs’
• 4 assigned
  1. These occurred after Trump announced a new 50% tariff on the EU.
  2. The stocks were sold the next trading day for over 100% more CAGR on the rebound.
  3. This indicates that the system has potential during periods of market stress.
• 5% success (100% if not for the political volatility)
• 36% CAGR

Our version 3 results, ‘PUTs’ and ‘Swing Trades’, are updated weekly at Probability Trader Pro

The Takeaway

Model-based probability estimates, such as delta, give the illusion of precision.
The 90% strike system delivers actual probability, based on actual outcomes.

Next, we will walk through how it works — step by step — and how we integrate historical data, model confidence, and market timing into a high-probability, high-discipline strike selection framework.

Why the 90% Strike System Outperforms Delta-Based Approaches

Our empirical, data-driven strike selection method provides tangible advantages over traditional delta-based approaches:

• Improved Accuracy and Reliability:
  Our system bases strike selection on measurable historical outcomes rather than volatile market sentiment (IV). This ensures strikes accurately reflect real probabilities, not theoretical guesses.
• Enhanced Risk Management:
  By clearly identifying historical performance boundaries, traders can confidently gauge true risk, reducing exposure to surprise volatility-driven events that delta-based strikes consistently underestimate.
• Greater Responsiveness:
  Frequent updates and simple real-time adjustments ensure traders can swiftly adapt to intraday market shifts, unlike static delta strikes, which quickly become outdated.
• Transparent, Practical Guidance:
  Traders benefit from explicit, structured rules on strike selection and trade adjustments, eliminating the guesswork and ambiguity inherent in interpreting delta and IV signals.

In short, our 90% strike system offers traders a practical edge—empirical accuracy, transparent rules, and disciplined responsiveness—where delta-based approaches provide only the illusion of precision.

5. The Three-Strike System: Historical Data Meets Real-Time Probability

The 90% strike system employs three distinct methods to identify a strike level that the

stock is unlikely to breach, each offering a different perspective on downside risk.

• 10+ year stock and day of week specific close-to-close strike,
• 10+ year stock and day of week specific open-to-close strike,
• 5-year, stock-specific, probability-determined strike.

The first two are based on 10+ years of historical returns, while the third reflects the current probability generated by our probability models. For more information about our stock-specific probability models, see our white paper, Our Machine Language Method

The goal is simple:

The 90% strike system employs three distinct methods to identify a strike level that the

stock is unlikely to breach, each offering a different perspective on downside risk.

• 10+ year stock and day of week specific close-to-close strike,
• 10+ year stock and day of week specific open-to-close strike,
• 5-year, stock-specific, probability-determined strike.

The first two are based on 10+ years of historical returns, while the third reflects the current probability generated by our probability models. For more information about our stock-specific probability models, see our white paper, Our Machine Language Method

The goal is simple:

Example: BSX

Here, the probability is high and the band strike price is close to the stock price, above the historical open and close strikes (representing all historical probabilities combined).

In contrast, look at CAH and GOOGL. These stocks show a low probability of success, and as a result, the band strike is much lower than historical strikes. This is the bottom of the strike barrel. The probabilities where historical new lows are made, which we want to avoid, and the band’s strike level reflects that.

20% premium: This represents the minimum amount of option premium needed to generate a 20% annualized return, calculated using XIRR (Extended Internal Rate of Return). This accounts for both the time to expiration and the capital at risk if assigned. For a Wednesday-to-Friday trade lasting just 2 days, BSX might require $0.16 to hit this threshold. A longer trade, such as Monday to Friday, would require $0.32.

6. How I Use Three Strikes, Probability and Premium

Here are the rules I trade by:

• I seek a 20% or larger premium for each trade.
• I only trade high probability stocks with a win rate above 70 – 80% and a Sharpe ratio above 2-3.
• I want the ‘Band Strike’ to be higher than the ‘Close Strike’.
• I use the ‘Open’ and ‘Close’ Strikes as a strike window. Early in the day, I trade near the ‘Open Strike’ and later in the day, I gravitate towards the higher ‘Close Strike’.

I want the open/close historical strikes to be LOWER than the band strike.
That tells me the model is confident, and history gives me a buffer. That’s where I trade.

If historical strikes are higher than the Band Strike — I skip the trade. These trades are where new lows are made.

The Strike Table: Updated Three Times Daily

To keep things actionable, I update the strike table:

• End of day, after market (based on today’s close)
• Moring, after the first 60–90 minutes of the market day
• Afternoon, the last 1-2 hours of trading

Real-Time Strike Adjustments

To ensure the system remains actionable throughout the trading day, we provide three daily updates to our strike table:

• End-of-day (post-market close)
• Morning (60–90 minutes after market open)
• Afternoon (last 1–2 hours of trading)

However, intraday price movements require timely adjustments. Here is our straightforward process to quickly adapt the published strikes to real-time prices:

Example Adjustment:

1. Calculate Current Price Difference
   • Published stock price: $101.00
   • Current market price: $102.00
   • Price difference: +$1.00
2. Apply Adjustment to Published Strike
   • Published strike: $96.62
   • Adjusted strike calculation: $96.62 + $1.00 = $97.62
3. Round to Nearest Tradable Strike
   • Options trade in increments (typically $0.50 or $1.00)
   • $97.62 rounds conveniently to $97.50 or $98.00, depending on available strikes.

This quick method maintains high accuracy, typically within pennies of the precise strike calculation.

When This Method Works Best:

• Normal market movements (±1–3% intraday).
• Moderate market conditions without major news-driven shocks or earnings reactions.

With frequent updates and this simple intraday adjustment method, traders remain precisely aligned with market reality, something delta-based strikes, relying on static IV estimates, cannot match.

7. Managing the Inevitable: When Strikes Break

Even with a system built around 90% probability, breaches will still occur. That’s the nature of trading: stuff happens! The key is not trying to eliminate them — it is having a clear, unemotional plan for what to do when they occur.

The strike, premium, and probability must all align. If they don’t, I pass.

The Strike Table: Updated Three Times Daily

To keep things actionable, I update the strike table:

• End of day, after market (based on today’s close)
• Moring, after the first 60–90 minutes of the market day
• Afternoon, the last 1-2 hours of trading

Real-Time Strike Adjustments

To ensure the system remains actionable throughout the trading day, we provide three daily updates to our strike table:

• End-of-day (post-market close)
• Morning (60–90 minutes after market open)
• Afternoon (last 1–2 hours of trading)

However, intraday price movements require timely adjustments. Here is our straightforward process to quickly adapt the published strikes to real-time prices:

Example Adjustment:

1. Calculate Current Price Difference
   • Published stock price: $101.00
   • Current market price: $102.00
   • Price difference: +$1.00
2. Apply Adjustment to Published Strike
   • Published strike: $96.62
   • Adjusted strike calculation: $96.62 + $1.00 = $97.62
3. Round to Nearest Tradable Strike
   • Options trade in increments (typically $0.50 or $1.00)
   • $97.62 rounds conveniently to $97.50 or $98.00, depending on available strikes.

This quick method maintains high accuracy, typically within pennies of the precise strike calculation.

When This Method Works Best:

• Normal market movements (±1–3% intraday).
• Moderate market conditions without major news-driven shocks or earnings reactions.

With frequent updates and this simple intraday adjustment method, traders remain precisely aligned with market reality, something delta-based strikes, relying on static IV estimates, cannot match.

7. Managing the Inevitable: When Strikes Break

Even with a system built around 90% probability, breaches will still occur. That’s the nature of trading: stuff happens! The key is not trying to eliminate them — it is having a clear, unemotional plan for what to do when they occur.

This section outlines exactly how I manage failed trades, including what typically causes them, how I respond, and when I walk away.

Why Strikes Break (The 10% Zone)

In my experience, breaches usually fall into three buckets:

1. Geopolitical or Macro Shocks
   War headlines, central bank surprises, unexpected regulation, tariffs — the kind of things that don’t show up in price structure or momentum but move entire sectors or markets overnight.
2. Earnings Misses or Guidance Cuts
   Even strong companies sometimes disappoint. If a stock gaps down 7–10% on poor earnings or soft forward guidance, it can blow through even a very conservative strike. (I never sell ‘PUTS’ near earnings reports)
3. Narrative or News-Driven Moves
   Short-seller reports, social media rumors, analyst downgrades, or regulatory headlines can send a stock sharply lower without any prior technical signal.

Other edge-case contributors:

• Index rebalancing
• ETF liquidity squeezes
• Sector-specific rotation or sentiment collapse

The point is: these are real events, not system errors. That’s why the strike broke. The question is what to do next.

The Four-Part Response Framework

When a strike level gets breached, I follow a decision tree with four options, depending on the model’s current outlook and the stock’s behavior.

1. Roll and Re-Center

If the stock has dropped below the strike but the model still sees strength (high probability), I will roll the position:

• Push the expiration out 1 week
• Lower the strike for more buffer
• Target a new premium, ideally aiming for a 20%+ XIRR, but primarily to roll down to a lower strike price without paying a premium.

This resets the trade and works best on clean pullbacks where the model remains bullish. IF I cannot roll down to a lower level without sacrificing earned premium, THEN I will do the following.

2. Convert to EDTL Swing Trade

If I cannot roll to a lower strike with some additional premium, or if the stock is assigned but still shows strong model support, I shift to a swing trade using my EDTL framework (Entry, Discount, Target, Levels): (see separate white paper on the EDTL method)

• The assigned price becomes the swing base
• Use predefined levels to add additional lots (usually via new puts with the strike price set at the predetermined discount level)
• Sell the stock at a clearly defined profit target
• Risk is capped by the number of potential additional lots as determined by the EDTL method.

If the stock still has a positive probability, these post-breach swings typically outperform the original ‘put’ profit, especially after sharp but temporary drawdowns and subsequent rebounds.

3. Sell Covered Calls

If the stock looks neutral (based on probability) and no obvious rebound or breakdown is apparent, I will take an assignment and sell covered calls:

• Use short-term calls (1–2 weeks) to generate premium
• Sell calls near the EDTL exit price

This strategy reduces drawdown without chasing recovery.

4. Exit and Reallocate

If the stock appears ‘broken’ and the probability turns bearish, I exit the trade:

• No revenge trades
• No trying to “get even”
• Capital gets reallocated to a better setup

This is the discipline filter. Not every position deserves or can be fixed, and not every loss is a mistake.

Conclusion: From Model Illusion to Market Reality

For decades, delta has been treated as a probability — a comforting but dangerous myth repeated across trading platforms, forums, and even professional education. Yet delta is not a forecast. It’s a hedge ratio wrapped in theoretical math, driven by implied volatility that reflects fear, not fact. As we have shown, this illusion of precision can lead to mispriced risk, misplaced strikes, and unnecessary losses — especially in volatile or fast-moving markets.

The 90% strike system offers a fundamentally different approach: it does not estimate probability from a model — it measures it from actual market behavior. Built on over 10 years of historical returns and reinforced with calibrated, machine-learning-based probability bands, the system delivers what delta never could — consistent 90 % %+ success rates based on what stocks actually do, not what models assume they should do.

This isn’t about chasing perfection. Even the best probability system will fail 10% of the time. But when it does, the response is clear and structured: roll down, convert to a swing trade, sell covered calls, or exit and reallocate. No guesswork. No revenge trades. Just disciplined action.

The core insight is simple yet powerful: in markets, as in life, the most dangerous errors are often disguised in mathematical language. Delta feels scientific. Implied volatility sounds sophisticated. However, both are built on fragile assumptions that crumble under real-world stress.

The 90% strike system works because it starts with the right question — not “What should happen?” but “What has happened before?” That shift — from theoretical elegance to empirical truth — is what transforms a trading idea into a trading edge.

This perspective isn’t just philosophically sound — it’s academically validated. Fan and Mancini (2009) demonstrate that nonparametric, empirically-grounded models consistently outperform parametric delta-based approaches, particularly in volatile market conditions where precision matters most. Their research confirms what our trading results prove: adopting an empirical framework isn’t merely practical — it’s demonstrably more accurate and profitable.

For traders ready to move beyond delta’s false promises, the path forward is clear: trade what you can measure, not what you can model. In the end, the market’s history remains the only teacher that never lies.