Win Rate Lies.

Saturday morning, my trading system flagged a fake winner.

I run a paper-trading stack at home — a signal scanner, a Bayesian grader, an executor, and a nightly calibrator that re-weights every signal based on its track record. The calibrator's job is to find what's working and lean into it. That morning, the first run did exactly what you'd expect. It boosted a signal called scanner_bull_high from a weight of 1.0 to 1.1. The reason was clean: a 90% win rate over its full history.

Two hours later I re-ran the calibrator with one extra check turned on — a payoff-aware override. It looked at the same signal and reversed the boost. New weight: 1.0. The override fired with a flag I'd named TROJAN HORSE when I built it weeks earlier, half-joking, never expecting it to actually trigger on something I owned.

Here's what the second pass saw that the first one missed.

Ninety percent of the time, this signal made me a small amount of money. Ten percent of the time, it took back more than it gave. Across the full history, every single trade carried a negative expected value of just over one percent. The wins were a friendly handshake. The losses were a punch in the face. Every dollar I ran through it was bleeding.

If I had stopped at the win rate, I would have boosted that signal and lost more money faster. The same trap is hiding in every sales pipeline I've ever managed.

Win rate is a confidence indicator, not a profit indicator.

Most sales orgs I've worked with track three numbers obsessively: pipeline coverage, average sales cycle, and win rate. Win rate is the one that gets the most leadership attention, because it's the easiest to compare across reps, segments, and quarters. Lisa is at 28% closed-won, James is at 41% — what's James doing differently?

The trap is that win rate, on its own, tells you nothing about whether a rep, a segment, or a product line is actually making the company money. It's a measure of how often you say yes. It is silent on what you got and what you gave up to get it.

This is the equation every RevOps lead should stare at on Monday morning, and almost none do:

The second term is what gets ignored. The cost of a lost deal isn't zero. It's the time the rep spent in discovery, the SE hours burned on a POC, the deal desk cycles, the legal review, the executive sponsor's calendar, and the demand gen spend that fed the lead in the first place. When you account for all of it, a 60% win rate on small deals can lose to a 25% win rate on large ones — by a wide margin.

Run the math on your own pipeline and watch which reps quietly flip from heroes to bottom-decile.

Three Trojan Horses I've seen in real sales orgs.

Once you know what to look for, the same pattern shows up everywhere. Here are three that I've personally watched eat into ARR while looking great on a dashboard.

1. The high-velocity SDR with a leaky funnel.

An SDR books meetings at twice the team's average rate. Calendar full, MQL→meeting conversion looks elite, manager celebrates her in the all-hands. Six months later you look at the cohort: of every ten meetings she booked, one closed. The team average is one in four.

She isn't booking meetings. She's booking interruptions to AE time. The cost of every dead meeting is roughly an hour of AE prep, an hour on the call, thirty minutes of internal write-up, and a CRM hygiene tax. At 200 meetings a quarter, that's 500+ hours of AE capacity going to deals that don't close — and crowding out the deals that would have. High win rate on a bad metric (booked meetings) is a Trojan Horse for the metric that pays the bills (closed-won pipeline).

2. The product line that closes well — and bleeds margin.

One product in your portfolio has a 62% close rate. The marketing team funds it heavily because it converts. The leadership team praises it because the velocity is short. Then someone runs gross margin by SKU and the product is the lowest-margin thing the company sells, by a wide margin. Every dollar booked is a dollar the rest of the company has to subsidise.

This is the textbook Trojan Horse: the wins look frequent and clean, but the unit economics are upside down. The fix isn't to kill the product. The fix is to stop treating closed-won as the terminal metric and start tracking contribution margin per closed deal alongside it. Once you do, the comp plan tells the truth.

3. The "champion" deal that closes — at 70% off list.

Senior reps know how to sense a stalled deal and rescue it with a discount cliff at quarter end. The deal closes. The rep hits quota. The CRM logs a win. But the price-per-seat baseline you set with that customer becomes the anchor for every renewal, every expansion, and every benchmark a future buyer hears about through their procurement network. You traded one quarter's number for three years of compressed ACV.

This one's the hardest to spot, because it presents as resilience on a slide. The fix is to track discount severity by deal as a first-class metric and treat anything past a threshold as a partial loss in the comp plan, not a clean win.

The five-minute payoff audit any RevOps lead can run on Monday.

You don't need a new tool to find this. You don't need a Snowflake migration. You need fifteen minutes with last quarter's pipeline export and a calculator. Here's the exact sequence I'd run if I dropped into a new RevOps seat.

1. Pull the last 12 months of closed-won and closed-lost. Both, not just the wins. The lost column is where the cost lives.
2. For each segment (rep, product line, industry, ACV band — pick whichever lens matters most for your motion), compute four numbers: count of wins, count of losses, average ACV won, and average fully-loaded cost of a lost deal — including SE, deal desk, exec sponsor time. Your finance team can give you a per-hour blended rate.
3. Compute expectancy per opportunity: (WR × AvgWonACV) − ((1 − WR) × AvgLossCost). The number can be negative. When it is, you have found a Trojan Horse.
4. Compute payoff ratio: AvgWonACV ÷ AvgLossCost. Anything below 1.0× is a segment where every loss costs you more than every win is worth.
5. Sort the segments by expectancy, descending. Anything in the bottom quartile is where your team is bleeding. Anything with a high win rate in the bottom quartile is the most dangerous: it looks healthy on the dashboard, and it's the hardest to kill politically because someone is being celebrated for it.

I have never run this audit on a real pipeline and not found at least one Trojan Horse. Often three. Almost always with a high-performing rep, a popular product, or a strategically loved segment sitting on top of it.

What changed in my system.

The fix on the trading side took eight lines of code. The calibrator now computes payoff ratio and expectancy on every signal, not just hit rate. Anything with a payoff below 1.0× and negative expectancy has its weight forced back to neutral, even if its win rate is 90%. The override has a name (TROJAN HORSE), a reason field, and a log line every time it fires. The point isn't to never trade those signals — it's to never over-trade them on the strength of a number that flatters them.

The same fix works in revenue operations. You don't need to fire the SDR, kill the product, or stop discounting deals. You need to make sure your dashboard and your comp plan stop quietly rewarding the version of the metric that loses you money.

The honest takeaway.

Most pipeline reviews ask the wrong question. What's our win rate? is a question about confidence. What's our expectancy per opportunity? is a question about whether the work is worth doing.

Confidence and economics aren't the same thing. The market doesn't care how often you're right. It cares about the size of the cheque on the way in versus the size of the cheque on the way out. Run that math on every segment of your pipeline, and you'll find a Trojan Horse. Run it on your trading system, and you'll find one there too.

I'd rather find them on a Saturday morning, in a paper-trading log, than on a Monday board call.