If you spend any time on options trading Twitter or in alert-service marketing emails, you've seen this claim: "Our trades are 87% winners last 90 days."

Almost without exception, that number is not something you can verify.

We built our unusual options activity pipeline differently — and we publish what it actually shows, including the trades that didn't work. This page is about the principles behind our scoring, the questions to ask any options service whose win rate looks too good to be true, and why our willingness to publish losses is the most useful number on our entire site.

Five questions to ask any options track record

Before walking through how we approach scoring, here are the questions we'd want any service to answer about its numbers — including ours. If a service can't answer all five clearly, the win rate isn't worth much.

What were the actual entry and exit prices? If a service shows trade tickets ("AAPL 250C 5/17 alerted at $4.20") but never the price you'd have realistically been able to fill — or what price the trade ultimately closed at — the gap between alert and reality routinely runs 20%+.
What time window? "Last 30 days" or "Q1 only" are easy to cherry-pick. For any directional bias, there's a 30-day window in which it looks brilliant.
Where are the losers? If you can scroll the historical feed and see only winners, that's not honesty — that's selection bias.
What are the exit rules? When does an open trade count as a winner or a loser? What happens to a 0DTE that expired worthless overnight? What about partial fills? If the rules aren't published, the win rate isn't replicable.
How big is the sample? "87% over 23 trades" is wildly different from "87% over 230 trades." Small samples are noise.

If a service can't answer those five, the headline number is marketing, not measurement.

How we approach scoring

A few principles drive how we audit every print on the tape:

Settled exchange prices, not "alert prices"

When a print appears on the public options tape, the option had a real, observable mid quote at that moment. That's our entry. The exit is whichever real, observable settled price applies — the matched closing print if the institution unwound the position, the price at expiration if the option expired, or the most recent settled close if the position is still open. No service-defined "we sold here" hand-waving.

Multi-leg trades scored as one position

When an institution prints a bull call spread, a put ladder, or a strangle, the whole structure has one outcome — not separate per-leg outcomes that can be cherry-picked. Our matcher reconstructs multi-leg structures from same-time same-execution prints and scores the group's net P&L. (For more on how the matcher reconstructs these from the raw tape, read How We Decode Whale Trade Intent.)

We exclude what we can't price fairly

A small percentage of prints get recorded but don't count toward the headline:

prints that paired and cancelled before any open interest moved (the institution crossed and pulled out)
contracts that don't have settled exchange data (illiquid strikes, off-tier listings)
single legs of multi-leg structures where we never observed the other side

These trades remain visible in the public dataset with their exclusion reason — they just don't affect the win rate. Hiding them entirely would be cleaner-looking but less honest.

Every trade carries an opening-vs-closing classification

Whale-following only works when you know whether the institution is putting on a position or closing one. Closing prints don't tell you what whales are about to do — only what they're unwinding. Our internal classifier tags every print on a confidence scale, and that tag rides through to the public track record so you can filter to opening prints with the highest conviction signal.

What you'll actually see in the numbers

We're holding back the formal published win-rate number while we extend the methodology, but we can already tell you the shape of what you'll see when we publish:

The headline is materially below "80% winners." The institutions printing UOA tape are not collectively running a 60% win rate, and any service claiming they are is selectively reporting. The honest aggregate for raw, unfiltered UOA flow is closer to a coin flip than to a signal — and the median trade return is negative. This is exactly what you'd expect from a feed that includes long-premium bets that decay to zero, premium-seller positions that get run over, and short verticals that get assigned.
Median is more honest than mean. A few trades that doubled or tripled drag the mean up; a few that printed -100% drag it down. We anchor on the median because it reflects the typical trade better.
The edge — when there is one — is in the subset. Defined-structure spreads (bull call, bear put, etc.) outperform naked single-leg flow. Opening prints with enough time to expiration to survive a chop outperform short-dated expiry plays. Specific multi-leg block trades printed by single institutions — the kinds of cases we cover in our flashback series — carry multi-bagger payoffs to expiration. Averaging across all UOA flow obscures that subset; the right framing is to slice, not average.

The takeaway: a published "win rate" that doesn't disclose what was excluded, what the median (not mean) was, and which subset out-performs is doing performance theater.

A more honest metric for retail traders

The standard win-rate number measures trade outcomes from the institution's perspective — entry to whichever close anchor we found. If you're a retail trader following the print, that's not always the most useful number. You don't know when the institution is closing; you do know when the print appeared, and you can monitor it from then.

So for every trade we also surface the highest peak the option mark reached after the print appeared on the tape. Not a price you could realistically have realized — that requires perfect timing — but the ceiling for what a retail trader could have captured if they entered on day 1 and exited at the local high.

For the ARM 170C printed on April 20, 2026: entry $42, peak $95 four trading days later. A retail trader who saw the print on the morning of 4/20 had a ~120% peak available to them. That's the ceiling. The price the institution actually exited at is the floor. Both numbers belong on the page.

Why we publish the losers

Every UOA scanner has losers. Most don't show them. We do — by name, by size, by dollar P&L — on the same page as the winners.

A track record without losers isn't a track record; it's a marketing page. If a trade went badly, knowing why is more valuable than not knowing it happened.

Browse the Top losers section of the public track record. You'll find named, sized, dollar-quantified entries — short call positions on names that ripped, defined-risk spreads that went the wrong way, premium-buyer positions that decayed to zero. These are not buried behind a "show losers" toggle. They're visible on the same page as the winners, with the same data, in the same format.

Our biggest single-leg loser in the recent window — an ARM short-call print from earlier this year — sits on the public ledger at roughly −$88M of mark-to-market loss. Knowing that helps you calibrate. So does knowing that it was the biggest loser, not a typical one.

You can verify any number we publish

Every aggregate number on this site traces back to specific underlying prints. The public track record page lets you spot-check those prints directly — entry print details, exit anchor, and exclusion reason if applicable.

If a number you see on this site doesn't reconcile to the underlying rows, that's a bug. Tell us and we'll fix it. (That's a sentence very few alert services would print, and there's a reason.)

The bottom line

We don't run an 87% win-rate alert service because no honest UOA scanner does.

What we run is a transparent audit of every institutional print we surface, including the losers and the inconclusive ones, with every entry and exit traceable to settled exchange prices.

If your benchmark for an options signal is "make me 80% winners every month," nothing legitimate will satisfy that — and you should be skeptical of anything that claims it can. If your benchmark is "show me what institutional flow is doing, in a way I can verify and use as one input among several alongside my own work," that's what we built.

Browse the public track record → · See today's unusual flow → · Read a flashback case study →

How We Score Every Unusual Options Trade — Honestly