CLOSING-LINE VALUE — THE FORECAST METRIC THAT MATTERS.

If you follow sports forecasts — or anyone who publishes them — there's one number you should care about more than any other. It's not win rate. It's not profit. It's not return on investment. It's closing-line value, or CLV. Here's what it is, why it's the standard metric, and how to use it to evaluate any prediction source (including ours).

The problem with win rate

The most common thing people brag about: "I'm hitting 56% on my picks!" Sounds great. Probably means nothing.

Win rate is noisy. Over a small sample, even genuinely random picks can show win rates above or below their true value by a wide margin. A coin-flipper has a 50% true win rate, but in any given 30-pick stretch they could easily hit 60% — or 40%. The variance is huge.

The math:

• 30 picks at a true 50% win rate → 95% confidence interval is 32–68%.
• 100 picks at a true 50% win rate → 95% CI is 40–60%.
• 1,000 picks at a true 50% win rate → 95% CI is 47–53%.

So a tipster showing you a 60% win rate over 30 picks could just be on a lucky streak with no actual edge. Even a 100-pick sample at 60% (CI 50–70%) leaves room for the underlying skill to be exactly 50% — pure noise.

You'd need thousands of picks for win rate alone to be statistically meaningful. Almost no public forecaster or tipster has that record — and the ones who do are usually selectively reporting.

The problem with units profit

"Up +18.4 units last month!" Also potentially meaningless. Profit is path-dependent. A forecaster who got lucky on three big-priced longshots can show a positive ROI for years before the variance catches up. A forecaster with genuine edge who hit a cold streak can show negative ROI for a year.

Profit is the outcome people care about, but it's not the metric you can measure skill with. Skill shows up over thousands of forecasts. Most public records aren't long enough.

What CLV actually is

CLV stands for closing-line value. It's the difference between the line at the time of the forecast and the line that closes (when the game starts).

Example: A forecast goes out on the Lakers at -3 on Friday morning. By Sunday afternoon at tipoff, the market has moved to Lakers -4. The forecast identified the side at -3, but the market closed at -4. The forecast beat the closing line by 1 point.

That's positive CLV. Whether the Lakers actually win or lose by 5 doesn't matter for the CLV calculation — what matters is that the forecast identified a price the market eventually concluded was off (in the forecaster's direction).

Why CLV is the standard

The closing line is, statistically, the most efficient price the market produces for any given game. By the time the game starts:

• Every public participant has acted on their view.
• Every sharp participant has acted on their view.
• Every injury, weather change, and lineup announcement has been priced in.
• Operators have rebalanced their exposure based on flow.

The closing line is, in effect, the wisdom-of-crowds answer with everyone's information aggregated. Beating it consistently means you have access to (or have computed) information the market doesn't have yet.

This is why professional forecasters and institutional operators both treat CLV as the only meaningful accuracy metric. Operators track CLV per participant — forecasters who consistently beat the close get flagged as sharp. Forecasters who lag the close are noise.

How to compute CLV

For a single forecast, the simple formula:

• For sides/totals: CLV (in points) = forecast line − closing line. If the forecast was -3 and the close was -4, CLV is +1 point. If the forecast was +6.5 and the close was +5, CLV is +1.5 points.
• For moneylines: Convert both prices to implied probability. CLV = closing implied probability − forecast implied probability. If the forecast was +200 (33.3% implied) and the close was +150 (40%), CLV is −6.7%. (The forecast was worse than the close — negative CLV.)

For a sample of forecasts, take the average. A forecaster or model averaging +2% CLV over 200+ picks has demonstrated edge that's almost statistically certain to be real.

What "good" CLV looks like

Pick1's 30-day rolling CLV is +3.8% at a standard -110 reference price. That's in the strong-edge band — and the model has generated enough forecasts for that number to be statistically meaningful.

Why CLV beats every other metric

Three reasons.

1. It's hard to fake. Win rate can be cherry-picked. CLV requires comparing your forecast line to the closing line on every forecast — a record only available to people who actually published (or recorded) the forecast at a specific time. Tipsters showing screenshots of verified multi-event combos can't fake CLV without making up data.
2. It converges fast. Win rate needs thousands of picks to be statistically meaningful. CLV converges on edge within hundreds — often within months for an active forecaster.
3. It's the metric efficient markets use. Institutional operators track CLV. Professional syndicates track CLV. If you want to know if a forecast source is more accurate than the consensus, the metric efficient markets use is the metric you should use.

How to use CLV when evaluating any prediction source

If you're trying to decide whether to follow a tipster, an AI model, or any other sports-prediction source, ask three questions:

1. Do they track and publish CLV? If no, dismiss them. Anyone serious about edge tracks this. Tipsters who only show "5-1 last week!" don't because they can't.
2. Is their average CLV positive over a meaningful sample? Look for at least 200 picks with a positive average. Below 200 picks, even a +2% CLV could be variance.
3. Is the sample current? A 2018 record of +5% CLV doesn't tell you what they're doing in 2026. Markets evolve; models drift. The relevant CLV is the last 3–6 months.

This single filter eliminates 99% of paid sports-prediction services. Most don't survive it because most don't have the edge they claim.

What we do at Pick1

Every Pick1 prediction is logged with the line we got and the closing line. CLV is computed automatically per pick, and the rolling 30-day average is displayed publicly in the app. If our CLV drops, you'll see it. If it rises, you'll see that too.

The model is recalibrated nightly on CLV results — picks that beat the close get more weight, picks that lag the close get less. This is the same online-learning approach high-frequency trading firms use to keep their models honest in non-stationary markets.

It's not a marketing claim. It's how the engine works.