Does unusual options flow actually work?

The honest short answer

Yes, but not in the way most flow-chasing content suggests. Academic finance has studied options order flow for decades, and the consistent finding is that large, aggressive, informed-looking options trades do contain a directional signal about the underlying stock over the next few days. Stocks with unusual call flow tend to outperform; those with unusual put flow tend to underperform. The effect is real and statistically distinguishable from noise.

The catch: "statistically significant" and "tradably profitable" are different thresholds. After accounting for bid-ask spreads, execution slippage, option premium cost, and the difficulty of identifying which prints are signal vs noise in real time, the edge is much smaller than the headline numbers suggest. For a retail trader acting on every EXTREME print they see, the practical result depends heavily on execution, position sizing, and how well they can distinguish the cleaner signals from the murkier ones.

What the research actually says

A series of academic papers have found consistent evidence for informational content in options flow:

• Pan and Poteshman (2006) found that put/call ratios constructed from publicly available options volume contain significant information about next-day stock returns. Stocks with the lowest ratios (most call-heavy buying) outperformed stocks with the highest ratios by a substantial margin over the next week.
• Hu (2014) showed that signed options order flow, distinguishing buys from sells at the trade level, predicts stock returns beyond what other market variables capture, and the effect is strongest for smaller, less liquid stocks with more information asymmetry.
• Ge, Lin, and Pearson (2016) found that non-standard options trades (unusual size, structure, or timing relative to earnings and other events) have significantly higher predictive content than standard flow. The more unusual the trade relative to baseline, the stronger the signal.

The consistent theme: not all flow is created equal. The informational content concentrates in the most unusual, most aggressive, most asymmetric trades, which is exactly the category flow scanners like RadarPulse are designed to surface.

Why flow has a signal: the informed-trading view

Options markets attract sophisticated participants: hedge funds, quant firms, high-frequency traders, and large family offices. Many of these participants spend enormous resources on proprietary research, tracking supply chains, modeling earnings, analyzing satellite imagery of parking lots, monitoring alternative data. When one of these participants has a high-conviction view, they often express it in the options market rather than the stock market, because options provide leverage and defined risk that makes a high-conviction short-term view more capital-efficient.

The result is that the options tape carries what economists call informed order flow, trades placed by people who have done the analysis and believe the stock will move. This is not insider trading (which is illegal and strictly enforced); it's legal, sophisticated analysis that happens to show up as a large, aggressive, unusual options trade.

Market makers know this too. When a large sweep lands, market makers immediately adjust their bids and offers in both the options market and the underlying stock. The stock often starts moving within minutes of a large unusual print, not because of the print itself, but because market participants are interpreting the same signal you are. This means:

1. The market partially discounts the signal fast, which limits how much edge a slower trader can capture.
2. Timing and execution matter more than for slower-moving signals.
3. The most predictive prints are often the most widely noticed, and therefore the most competed for.

What makes a signal actually predictive

Not every print a scanner flags has the same signal content. The characteristics that most consistently correlate with genuinely predictive flow are:

• Aggressive sweep execution: A sweep that crosses multiple exchanges to get filled quickly signals urgency, the buyer is willing to pay up. A patient limit-order fill near the bid is much noisier.
• Out-of-the-money strike: OTM calls or puts are pure directional bets. ATM and ITM trades are more likely to be hedges, spread legs, or covered positions. The further OTM and the shorter the expiry, the more it looks like a directional conviction trade.
• Large premium relative to typical volume: A $2 million call sweep on a name where the typical daily call volume is $500K in premium is meaningful. The same dollar amount on a heavily traded mega-cap might be routine positioning. RadarPulse's scoring adjusts for this baseline, a high score means unusualness relative to that stock's normal behavior, not just large absolute dollar size.
• Short-dated expiry: A trader expecting a move in the next 5–10 days buys short-dated contracts. Someone buying 6-month contracts might be expressing a long-duration thesis, hedging an existing position, or rolling a covered call, none of which is a clean short-term directional signal.
• Single-leg trade: A multi-leg spread (call spread, risk reversal, collar) has a defined risk profile that makes direction harder to read. A single-leg sweep is the clearest statement of directional intent.

What flow does not tell you

This matters as much as what it does tell you:

• The trader's full position. A large call sweep might be one leg of a collar, a delta hedge on a large short position, or a spread with a leg you can't see. The tape shows individual prints, not books.
• Whether the trader is right. Even the most sophisticated funds are wrong on individual trades. Large premium is evidence of conviction, not evidence of correctness.
• What the precise move will be. Flow can hint that someone expects a move in a particular direction; it cannot tell you the magnitude, timing within a day, or exact catalysts. A stock with aggressive call flow might move the next morning, the next week, or after a news catalyst the flow buyer anticipated.
• Insider information. The signal from options flow is legal and based on public data analysis. It is not a leak of earnings results, M&A announcements, or other material non-public information. Treating it as a tip-off sets you up for poor risk management decisions.

The hit-rate problem: what a "win rate" actually means

Flow scanner marketing often leads with win-rate claims. Before trusting any number, ask:

• What is the base rate? Stocks trend up over time. A random long-biased strategy already beats 50% directional accuracy. Any claimed win rate needs to be benchmarked against this drift.
• How is "win" defined? Did the underlying move in the right direction by the expiry? Did the option itself profit (much harder, due to premium cost and IV crush)? Did the trader hold to expiry or exit early? Underlying directional accuracy and options P&L are very different metrics.
• What is the sample size? A 70% win rate on 15 trades is statistically meaningless. You need hundreds of trades to distinguish skill from luck, especially when outcomes are highly correlated (all signals that fire in a bear market will have lower hit rates regardless of quality).
• Is survivorship bias present? Scanners that show historical signals only for stocks that moved are not showing you the full picture. You need to see all signals fired, not just the ones that looked smart in hindsight.

How the RadarPulse Smart-Money Scorecard works

The Scorecard is our answer to these questions. Every signal flagged at EXTREME, ELEVATED, or NOTABLE is recorded at the moment it fires, not retroactively, not cherry-picked. We then measure the underlying asset's move over a 1-day, 3-day, and 5-day window from the signal time. What the Scorecard shows:

• Average underlying move by flag tier (EXTREME vs ELEVATED vs NOTABLE), so you can see whether higher-scored prints lead to larger subsequent moves.
• Directional accuracy by signal type (CALL vs PUT), measured against the implied direction of the flow.
• Sample size displayed throughout, the Scorecard is explicit when the sample is too small to be statistically meaningful, and withholds conclusions until it is.

The measured metric is the underlying's price change, not options P&L. This is an intentional design choice: options P&L depends on premium paid, IV crush, bid-ask spread, and timing, too many variables to isolate the signal. By measuring the underlying move, the Scorecard isolates whether the flow correctly anticipated direction, which is the cleanest question.

You can view the live Scorecard at radarpulse.io/scorecard →

Practical implications: how to actually use flow

Taking everything above seriously, here is how flow is best used:

1. Use it as a screening tool, not a trade signal. Flow is best for building a watchlist of names where sophisticated money is active. It narrows the universe; it doesn't replace the work of understanding why.
2. Prioritize the cleanest signals. EXTREME prints in short-dated OTM single-leg sweeps are the category with the most historical predictive content. NOTABLE prints in longer-dated multi-leg structures are much noisier.
3. Check for obvious alternative explanations. Is there an earnings event coming up that would explain hedging activity? Is the stock in a heavily shorted sector where covering could mimic call buying? Has the company recently made M&A news that would explain protective puts?
4. Size appropriately for a signal, not a certainty. If you're using flow as one input among several, size the position accordingly. Treating a single print as a "guaranteed move" and over-sizing is the most common way flow-following goes wrong.
5. Track your own outcomes. Keep a log of the signals you act on and what happened. Your personal sample, even if small, will tell you whether the specific type of flow you're using is actually working in the current market environment.

The bottom line

Unusual options flow works, in the specific sense that the most extreme, aggressive, unusual prints contain measurable informational content about short-term stock direction. The effect is real, academically documented, and concentrated in the highest-quality signals. It is not a reliable machine for generating profits without further work, and it is not insider trading or a magic tip-off.

The best approach: treat flow as a powerful but imperfect signal, weigh it alongside fundamentals and technicals, demand a large audited sample before trusting any win-rate claim, and track outcomes over time. The RadarPulse Smart-Money Scorecard exists to give you exactly that transparency, not to sell a fantasy, but to show the honest numbers as they accumulate.

How market conditions affect options flow signal accuracy

The single most underappreciated variable in options flow analysis is the market regime in which a signal fires. Flow printed in a low-VIX bull market means something quite different from the same print in a high-volatility bear market. Understanding these regime effects is not a subtle nuance, it is central to distinguishing real informational content from noise.

Bull markets (VIX below 15, uptrending indices): In sustained low-volatility bull markets, call flow is the dominant signal type by volume. The problem is that retail participation also spikes in these environments, producing more raw call activity across the tape. When everyone is buying calls, the signal-to-noise ratio deteriorates because a meaningful portion of call sweeps comes from momentum-chasing retail activity rather than informed institutional positioning. To compensate, the threshold for a meaningful signal should be higher in bull regimes: require Vol/OI above 5x (not the standard 2-3x) and a premium floor above $500K. The base rate of stocks moving up is already elevated in a bull market, which means a bullish call sweep needs stronger structural characteristics to stand out from ambient upward drift. A 50% directional accuracy in a bull market for call flow is actually worse than it looks, random bullish bets would also beat 50% given the drift.

Bear markets (VIX above 25, downtrending indices): Put flow becomes more prominent in bear markets, but portfolio hedging also increases dramatically, creating more false bearish signals. Large institutions running equity long books buy protective puts as insurance, not as directional short bets. In high-VIX environments, EXTREME put sweeps are more likely to represent hedging activity than informed directional conviction. Bearish flow in a bear market requires the same strict hedge-filter analysis every time: is this a fund protecting a long equity position, or is it a dedicated directional short bet? The presence of the print in a stock that also shows up in institutional 13F long disclosures is a strong indicator of hedging rather than direction.

The sweet spot, medium-VIX environments (15–25 VIX): Research consistently finds that options flow is most predictive in medium-volatility environments. Below 15, complacency produces noisy carry and yield-enhancement strategies (covered calls, cash-secured puts) that contaminate the tape. Above 25, hedging volume overwhelms the directional signal. The 15–25 VIX range is where informed institutional conviction is expressed most cleanly in the options tape, volatility is high enough that options are expensive enough to deter casual speculation, but not so high that every large print is a hedge. This is the regime where EXTREME signals historically carry their highest informational content.

Earnings season effects: During the peak earnings waves, the weeks surrounding January, April, July, and October reporting for S&P 500 names, option flow in individual names is heavily distorted by event-driven positioning. Earnings directional bets, straddles, strangles, and protective puts all surge ahead of reports, creating enormous volume that has nothing to do with a sustained directional thesis. Signal quality is at its lowest for single-name flow during earnings season and at its highest in the 2–4 weeks between earnings waves, when the tape returns to longer-duration conviction trading rather than short-term event speculation. If you are evaluating an EXTREME print in a name with earnings within 10–15 days, apply a substantial haircut to its informational content.

Sector-specific regime effects: Not all sectors respond identically to market regimes. Technology options flow is least predictive during periods of broad risk-off selling, when all tech names move together regardless of fundamentals. In those environments, even informed institutional positioning in an individual tech name gets overwhelmed by sector-wide de-risking. Technology flow is most predictive in stock-picker markets, periods when tech names diverge based on company-specific fundamentals rather than macro. Healthcare flow behaves differently: it is most predictive when it is company-specific (tied to an FDA PDUFA date, an M&A rumor, or a drug trial readout) and least predictive when driven by sector-wide healthcare policy shifts, which affect all names indiscriminately. Recognizing the regime your sector is trading in is the first filter before evaluating any individual print.

What "works" means for different trader types

The question "does options flow work?" has different answers depending on your time horizon, trading approach, and what you are measuring as the outcome. Here is an honest breakdown by trader type.

Day trader (holding periods: minutes to hours): For day traders, the most relevant question is whether a large EXTREME print moves the underlying stock meaningfully within the next 30–60 minutes. Research and practitioner observation both suggest that stocks do show above-average intraday moves in the first hour after a large unusual sweep, market makers reprice almost immediately when a large sweep lands, and other sophisticated participants interpret the same signal simultaneously. But this rapid discounting is also the day trader's core problem. By the time a flow scanner surfaces the print, a retail trader receives the alert, and the trader places an order, a meaningful fraction of the initial market-maker reaction is already in the price. Flow "works" for day traders primarily as an intraday directional bias tool, it tells you which names are in play that session and helps you bias your entries in the right direction, but not as a standalone trade trigger with reliable timing precision.

Swing trader (holding periods: 5–20 days): This is the time horizon where flow signals are best calibrated by design. Contracts with 15–45 days to expiry are explicitly targeting a move in the coming weeks, which is exactly the swing trading window. Academic research measuring predictive content in options order flow typically uses measurement windows of 1–5 days, and the full effect on the underlying often unfolds over 2–3 weeks for high-conviction setups. Flow works well as a screening tool for swing traders who combine the signal with technical confirmation (a breakout level, a trend support retest) before entering. The print identifies the name; the technicals identify the timing. Together they are considerably more effective than either alone.

Position trader and investor (holding periods: months): For longer-duration investors, individual options prints are too noisy to be useful as primary signals. A single sweep tells you that one participant had a short-term view on a given day. More relevant to a position trader is a sustained pattern: institutional call flow accumulating in a name across multiple sessions over 4–6 weeks, particularly when it coincides with 13F filing data showing accumulation and congressional trading data pointing in the same direction. Multi-session accumulation in the same name over an extended period is a fundamentally different signal from a single EXTREME print, it suggests that sophisticated capital is building a sustained thesis, which is the kind of confirmation a position-level investor actually wants. Single prints are noise; extended multi-session accumulation is the signal for this time horizon.

What "works" should not mean for any trader type: The profitability question cannot be answered by directional hit rate alone. A signal that is directionally correct 55% of the time with a 1:1 average risk-reward is profitable by a meaningful margin. A signal that is directionally correct 70% of the time but where the average losing trade loses twice the average winning trade is a slow drain. The expected value calculation, win rate multiplied by average win size, minus loss rate multiplied by average loss size, minus transaction costs, is the only honest measure of whether any signal "works." Options flow traders who focus only on directional win rate while ignoring the size distribution of their wins and losses are systematically miscalculating their edge.

The academic evidence in depth: what the papers actually found

The three core papers cited earlier deserve a closer look. The headline findings are frequently quoted; the methodological details that determine how to apply them in practice are rarely discussed.

Pan and Poteshman (2006), The Journal of Finance: This is the foundational paper in options flow research, and it has an important methodological feature that is almost always omitted when the findings are cited. Pan and Poteshman used proprietary CBOE data that included buyer-versus-seller identification at the trade level, the tape indicated whether the initiating party was a buyer or a seller, not just the direction of the option. This is fundamentally different from what public flow scanners can access. Public scanners use bid-ask side inference: if a trade prints at the ask, the scanner assumes the buyer was the aggressor; if it prints at the bid, the seller was the aggressor. This inference is correct the majority of the time but introduces noise, particularly in fast-moving markets where the spread is tight and mid-price fills are common. Pan and Poteshman's result, that stocks in the lowest put/call ratio decile outperformed those in the highest decile by approximately 40 basis points per day over the next week, is real and significant. But the magnitude likely does not translate fully to strategies based on public-tape inference alone. The signed trade data used in the study is cleaner than anything available to retail flow scanner users, so the practical edge is lower than the 40bp-per-day figure implies.

Hu (2014), Journal of Financial and Quantitative Analysis: Hu's contribution is the finding that predictive content in options order flow is not uniformly distributed across stocks. It is concentrated in smaller, less liquid stocks with greater information asymmetry, and notably absent (or much weaker) in heavily traded mega-cap names. The intuition is straightforward: in names where thousands of sophisticated traders are active simultaneously, information advantages are competed away rapidly. In mid-cap names where coverage is lighter and institutional information advantages are greater, the options tape carries more informational content. The practical implication for flow traders is significant: if you are exclusively watching Apple, Nvidia, and Tesla options flow, you are watching the least predictive segment of the market in terms of informational content per dollar. The best flow signals are more likely in names outside the most crowded, most-scrutinized tier of the market, exactly where a well-designed scanner provides the most value by surfacing names you might not otherwise be watching.

Ge, Lin, and Pearson (2016), Review of Financial Studies: This paper studied non-standard options trades, those with unusual size, structure, or timing characteristics relative to a stock's own baseline behavior and to corporate event calendars. The key finding: unusualness relative to baseline is more important than raw dollar size. A $1 million sweep in a name where the typical daily call premium is $100K is a 10x event. The same $1 million in a name that does $10 million per day is unremarkable. Ge et al. found that non-standard trades show 2–3 times higher predictive content than standard-volume flow. This is the quantitative validation for what RadarPulse's scoring model does: the score penalizes large absolute prints in high-baseline names and rewards moderate prints in low-baseline names, because unusualness relative to the stock's own history is the right measuring stick.

Additional research on put flow and delta-hedging: A strand of more recent research has found that put flow specifically predicts negative returns, but part of this predictive content is mechanically driven. When market makers sell large put positions to institutional buyers, they hedge their exposure by shorting the underlying stock. This delta-hedging effect itself creates downward price pressure on the underlying. Some of the "predictive content" in put flow is therefore partially self-fulfilling in the short term: the initial move is driven by the market maker's hedge, not by a subsequent catalyst the options buyer knew about. This does not make put flow less useful as a signal, the market maker hedge still creates a tradable move, but it does suggest that the causal mechanism is different for puts versus calls, and that the timing window for capturing the put-flow effect (closer to the print time) may be narrower than for call flow.

Building your own options flow accuracy tracking system

No published study, no backtested model, and no aggregate scorecard can tell you whether options flow works specifically for you, in your account, with your execution speed, in the specific names and market conditions you actually trade. The only way to answer that question is to track your own outcomes systematically. Here is a practical six-step system for doing that.

Step 1, Log every signal at the moment you act on it, not retroactively. Hindsight bias is the biggest enemy of honest signal tracking. If you log signals after the fact, you will unconsciously include more of the ones that worked and forget the ones that quietly expired worthless. The discipline is to log at the moment of decision: before you enter a trade based on a flow signal, record the signal. Required fields: ticker, signal tier (EXTREME / ELEVATED / NOTABLE), direction (call or put), premium, Vol/OI ratio, days to expiry, time of day, your entry price on the underlying or option, and a one-sentence thesis, why you acted on this specific print. The thesis field is more important than it looks. It forces you to articulate your reasoning before you know the outcome, which prevents post-hoc rationalization of both good and bad trades.

Step 2, Define your outcome rule before you enter. Consistent outcome measurement requires a consistent definition of "correct." Before acting on any signal, decide what counts as success for this specific trade. A reasonable default for short-term directional signals: "the underlying moves more than 2% in the direction of the flow within 10 calendar days of the print." The definition must be agreed to before entry and applied consistently across all logged trades, if you change the definition based on what the stock actually did, you are gaming your own scorecard.

Step 3, Record outcomes systematically at the defined window. At your predetermined outcome window, record the underlying's actual price change. Calculate two metrics: the raw directional hit rate (what percentage of your logged signals had the underlying move in the right direction), and the magnitude-adjusted hit rate (what percentage moved enough to produce a profitable options trade at your typical entry price, accounting for IV and premium). These two numbers will often diverge, particularly in high-IV environments where an option requires a larger underlying move to break even due to elevated premium cost.

Step 4, Segment your outcomes to find what actually works for you. After accumulating at least 50–75 signal outcomes, break down your hit rate by: signal tier (EXTREME vs ELEVATED vs NOTABLE), market condition at signal time (VIX quartile), time of day (first 30 minutes of session vs midday vs final hour), sector, and DTE range (0–15 days, 15–45 days, 45 days plus). Most traders who do this analysis discover that their hit rate varies dramatically across these segments, not because flow itself is inconsistent, but because certain signal types in certain conditions suit their particular trading context far better than others. NOTABLE prints might show 47% directional accuracy for you while EXTREME prints show 64%. If so, eliminate NOTABLE from your active filter entirely.

Step 5, Adjust your filter based on your actual data. Once your segmented analysis identifies which combinations consistently underperform, remove those combinations from your trading filter. If your log shows that midday flow prints perform materially worse than opening prints (a common finding, since opening sweeps are more likely to be informed positioning established before the session while midday prints more often reflect reactive or momentum-chasing activity), apply a midday filter and stop acting on prints that arrive between 11am and 2:30pm ET. Let your own empirical data, not the design of any scanner, determine what you trade.

Step 6, Review and recalibrate quarterly. Market regimes shift substantially over 3–6 month periods. A filtering strategy that achieved a 63% hit rate in a low-VIX bull market may drop to 51% in a high-VIX environment. Review your full tracking log quarterly, compare your current 90-day hit rate to your cumulative baseline, and recalibrate your filters when hit rates have moved materially downward. A 5-point hit-rate decline sustained over 90 days is a meaningful signal that the current regime no longer suits your filters, not a reason to abandon flow, but a reason to adjust thresholds and tighten criteria.

Case studies: three signals examined, correct, incorrect, and ambiguous

Abstract characteristics of good and bad flow signals are easier to internalize through concrete examples. The following three case studies illustrate how the same analytical framework produces very different outcomes, and what the difference between a clean signal and a contaminated one actually looks like in practice.

Frequently asked questions

Does unusual options flow actually predict stock moves?

Unusual options flow has a measurable signal, academic research consistently finds that large, aggressive options trades (especially sweeps with high open-interest-relative-to-volume ratios) are followed by above-average underlying moves more often than chance. But "more often than chance" is not the same as reliably profitable. The signal varies by stock, market regime, and signal type. EXTREME-scored sweeps in short-dated OTM contracts tend to show the strongest subsequent moves; generic high-volume prints are noisier. Flow is an input, not a prediction engine.

What is a good hit rate for options flow signals?

There is no single benchmark, but context helps. A coin flip gives you 50% on directional accuracy. Market drift (stocks tend up over time) already pushes random long-biased signals above 50%. A signal that scores 55–60% directional accuracy over hundreds of calls, net of this base-rate drift, is genuinely informative. Anything claiming 70–80% win rates without a large, audited sample and disclosed methodology should be treated skeptically. RadarPulse's Smart-Money Scorecard tracks outcomes on a per-flag basis (EXTREME, ELEVATED, NOTABLE) so you can see how each tier performs over time.

What makes an options flow signal more predictive?

The most predictive signals share several traits: aggressive sweep execution (not patient limit fills), large premium relative to the stock's normal options volume, short-dated expiry (the trader expects a move soon), out-of-the-money strikes (directional bet, not hedging), and no obvious alternative explanation like an earnings hedge or index rebalance. Single-leg sweeps in OTM short-dated contracts on names with unusual relative volume are the cleanest signal type. Multi-leg structures (spreads, collars) are harder to read directionally and tend to be noisier.

Can large options flow be wrong?

Yes, and it is wrong often enough to matter. Even the largest institutional traders make incorrect directional calls. A $5 million sweep in call options can represent a hedge against a short position, a portfolio insurance overlay, a spread with a leg you can't see, or a simple directional bet that fails. Flow tells you there was conviction and size; it does not tell you the trader was right. Treating flow as a tip-off rather than a signal is one of the most common mistakes retail traders make when following smart-money activity.

How does RadarPulse measure options flow accuracy?

RadarPulse's Smart-Money Scorecard records every flagged signal (EXTREME, ELEVATED, NOTABLE) at the moment it fires, then measures the subsequent underlying asset move over a fixed window. The measured outcome is the underlying's price change, not options P&L, which is affected by IV crush, bid-ask spreads, and timing, so the scorecard shows directional accuracy of the signal, not a simulated trading result. A disclaimer is displayed throughout: past signal accuracy does not guarantee future results.

Is options flow the same as insider trading?

No. Unusual options flow is not insider trading. Insiders are legally prohibited from trading on material non-public information under SEC Rule 10b-5. Large options prints reflect positioning by hedge funds, institutions, market makers, and retail traders using publicly available information and their own analysis. The SEC actively monitors and prosecutes insider trading. Treating flow as insider trading signals both misunderstands the law and sets an unrealistic expectation for the signal's reliability.

Related guides