Every touting model on the internet sells you a record. "83-42 last season." "Hit 17 of my last 22 NFL plays." Records are theatrically convincing and almost completely uninformative. A 17-of-22 hot streak is consistent with a true 50% bettor 4% of the time — common enough that someone, somewhere, is always running it. A leaderboard ranked by closing line value tells a different story: it shows whose process is actually producing better numbers than the market is settling on, which is the only thing that survives 1,000 bets. This guide walks through why CLV deserves the leaderboard slot, what the thresholds actually mean, and how to read the table on /leaderboards without fooling yourself.
The case against record-based leaderboards
The case is statistical, and it’s settled. Three concrete numbers:
- A bettor with a true edge of +2% needs roughly 1,800 logged bets before their win rate is more than two standard errors above 50%.
- A bettor with no edge (true 50%) has a 2.5% chance of running 56% over a 100-bet sample. With 500 "capper" accounts being tracked, you’d expect roughly 12 of them to look brilliant by record alone, every season.
- The correlation between this-month win rate and next-month win rate, conditioned on no edge, is approximately zero. The correlation between this-month CLV and next-month CLV, conditioned on real edge, is about 0.5.
Record-based leaderboards reward survivorship. CLV-based leaderboards reward process. The two distinguishing properties: (1) CLV stabilizes in a fraction of the sample size, and (2) CLV can’t be backfilled — you either logged the line you got at bet time, or you didn’t. Touts who can’t produce timestamps for their plays can’t produce CLV either.
What CLV actually measures
Closing line value is the difference between the price you locked in and the consensus price at game start. The fuller treatment is in our CLV explainer; the mechanics summary:
- You bet Bills -3 at -110, recorded at 9:14 AM Sunday.
- The line closes Bills -3.5 at -108 at the consensus sharp book.
- You beat the close by 0.5 points; in implied-probability terms, your -3 -110 was roughly 0.524 fair, the close -3.5 -108 devigs to roughly 0.508 fair on Bills. You captured +1.6 cents of fair-line edge.
- Whether the Bills cover or not, the 1.6 cents are real. The settlement of this bet is variance. The settlement of 500 bets at the same average edge is profit.
CLV is the leaderboard-friendly version of the bet because it’s computable per bet, scale-invariant (no need to normalize for stake), and forward-looking (it predicts future profit rather than describing past luck).
The thresholds you should know
The bands we use on /leaderboards for visual coding, based on devigged consensus-close benchmarks:
- Under 0% CLV — you’re paying the market a tax. The only mystery is how long until variance reveals it.
- 0 to 2% — break-even-ish. Could be marginal edge, could be lucky lines. Sample-dependent.
- 3 to 5% — "decent." Real edge over 300+ bets. You’re beating the close enough to survive vig long-term.
- 6 to 8% — good. Most bettors who clear this band over 500+ bets have a discernible process (a model, a steam-chasing routine, or specific market specialization).
- 8%+ sustained — sharp territory. Often gets limited at retail books within a season.
- 12%+ sustained over 1,000+ bets — elite. Vanishingly rare. Usually attached to professional syndicates, not individuals.
One important caveat: the bands shift by market. NFL spreads have the sharpest closes, so 5% CLV there is harder than 5% CLV on UFC fight props, where the consensus close is noisier. The CLV-by-market study breaks out the per-sport variance.
How a leaderboard avoids being gamed
The risks of a naive "highest average CLV wins" leaderboard:
- Tiny-sample exploitation. A user bets one prop early in the week, beats the close by 8 cents, and shows up at the top of the table on a 1-bet sample. Fix: require a minimum bet count per period (we use 50/week or 200/month).
- Cherry-picking which bets to log. Selective logging biases CLV upward. Fix: require every bet placed during the period to be logged, with a timestamp the system can verify (API integrations or screenshot uploads), or settle for self-reporting with public disclosure of the protocol.
- Opener-only specialists. Betting exclusively openers will inflate CLV but limit volume. Fix: rank by CLV × bet-count, or display both metrics side by side.
- Soft-book CLV. Beating a recreational book’s closing line is much easier than beating Pinnacle or Circa’s. Fix: always benchmark against the sharp consensus, not the user’s home book.
The /leaderboards implementation enforces minimum bet counts, timestamped lines, and sharp-consensus benchmarks. The full methodology is published so users can replicate the calculation — which is itself an anti-gaming property.
Reading the leaderboard table
A useful CLV leaderboard shows at least these columns:
- User / model handle — the entity being ranked.
- Bets in period — sample-size confidence indicator.
- Avg CLV — mean closing line value per bet, in cents or points.
- Standard error — the noise band. CLV ± 2 standard errors is the realistic range.
- Market mix — what fraction of bets were NFL sides vs MLB props vs whatever else. Some bettors only look good in one market.
- Hit rate — secondary metric, useful for sanity but not the primary.
- ROI — third-tier; correlated with CLV but variance-heavy.
The most informative bettors to follow are the ones with moderate CLV (4 to 7%) over large samples (500+ bets) in a focused market mix. They’re running a process. The flashy top-of-table players with 12% CLV on 80 bets are almost always going to regress — variance pulls extreme samples back toward their true edge.
What to do with someone else’s leaderboard rank
Following high-CLV bettors makes more sense than following high-record cappers. Two ways to use the leaderboard:
- Watch the process, not the play. Why did this bettor jump on this number? Their edge replicates if you understand the trigger.
- Compare your own rank against the cohort. Are you keeping pace on the same markets they are? If yes, your process is at least directionally correct.
The leaderboard is a feedback loop, not a tip sheet. The bettors at the top who matter aren’t trying to sell you their picks — they’re showing you what CLV looks like when the process works. That signal travels in the format the data exposes, not in marketing.
Why public CLV is rare
Most bettors and most touting services hide their CLV because it would expose them. The honest ones who do publish — a small number of independents, a handful of model accounts — provide the proof of concept that CLV is teachable and reproducible. The hostile narrative against CLV ("it’s academic," "variance still rules," "I’m a contrarian, the close is wrong") is consistent only with people who haven’t logged enough bets to compute it themselves. The framework holds up when checked. The framework also doesn’t care about anyone’s self-image.
The personal version: your own CLV table
The leaderboard mechanic scales down to one user. The /desk bet log computes per-bet CLV automatically when you supply the price you got and the closing line. After 200 to 300 bets you have a table that tells you:
- Which sports you actually beat the close in.
- Which markets you’re paying a tax on (often props you took because you "liked it," not because the price was right).
- Whether your high-confidence plays beat your routine plays on CLV — they often don’t.
- Whether your CLV is improving over time, or flat (the latter is the common case unless you’re actively iterating on your process).
Pair the table with our bet-tracking guide and the spreadsheet does most of the heavy lifting.
The flip side: when CLV is misleading
Two cases where CLV breaks down as a metric:
- Markets that don’t have a sharp consensus close. Niche props (low-volume player props, certain college bowl markets, exotic UFC props) often close at noisy numbers because no sharp book pushes back. CLV in those markets is mostly random.
- Markets you helped move. If you bet a number early enough that your stake itself moved the line, your CLV is calculating against a line you caused. This usually only matters for syndicate-scale bettors, but it’s a real edge case.
The Bayesian framing in our Bayesian CLV calculator handles both cases by widening the credible interval when sample sizes or sharp-close confidence are low.
Building the leaderboard: the technical bits worth knowing
What an honest CLV leaderboard system has to handle, in order of operational difficulty:
- Timestamped bet capture. Either an API integration with the user’s sportsbook, or a manual log that records the timestamp when the bet was created. Without timestamps, you cannot prove which line was available when. This is the single hardest engineering problem in CLV tracking.
- Sharp closing-line ingestion. Polling Pinnacle (or Circa, or a sharp consensus aggregator) at game start, snapping the line, and storing the devigged probability. Needs to handle late line moves, postponed games, and stale-line periods.
- Devigging. Two methods, both acceptable: multiplicative (divide implied probabilities by their sum) or additive (subtract half the vig from each). They produce slightly different numbers on heavy favorites. Pick one and use it consistently.
- Per-bet CLV computation. implied_prob(user_price) minus implied_prob(devigged_close). Aggregate across bets with weights by stake if you want a stake-weighted average, or unweighted for the "per-decision" CLV.
- Rank stabilization. Use a Bayesian shrinkage estimator or a minimum-sample filter so single-bet flukes don’t dominate the top of the table.
The minimum-sample filter is the single most important fairness lever. A leaderboard that ranks anyone with at least one bet rewards luck. A leaderboard that requires 50 bets per period filters out the noise and surfaces process.
Leaderboard cadence: weekly, monthly, all-time
Different timeframes serve different purposes:
- Weekly — useful for catching heaters, but mostly variance. Helpful for short-term confidence checks and for tracking when a new model first ships.
- Monthly — the most useful default. Long enough for CLV to stabilize on a focused bettor, short enough to detect process drift.
- Season / all-time — the trust signal. A bettor who’s positive CLV over 1,500+ bets across multiple seasons has shown durable edge.
The leaderboard surface at /leaderboards defaults to monthly with weekly and all-time as toggles. The monthly view is where the action is — it’s long enough to require process, short enough that improving bettors can climb visibly when they tighten their workflow.
How CLV connects to the rest of the bettor stack
CLV doesn’t sit alone. It’s the connective tissue between bet selection, bankroll, and feedback. The handoffs:
- Bet selection — your model on /tinker outputs a probability. If your CLV across bets sourced from that model is positive, the model is real. If it’s flat, the model is no better than the market.
- Bankroll — Kelly sizing assumes you know your edge. CLV is the empirical check on your assumed edge. If your assumed edge is 4% but your CLV is 1%, scale Kelly down accordingly. The /desk bankroll panel exposes this comparison directly.
- Process iteration — when CLV drops in a particular market category, that’s the signal to adjust: either retrain the model, narrow the market focus, or stop betting that slice.
The bettors who actually compound use the leaderboard as a diagnostic, not a scoreboard. Their rank matters less than the trend line of their CLV inside their own table.
Bottom line
A CLV leaderboard is the only ranking that survives contact with statistics. It exposes who’s actually beating the market and who’s riding variance, in a fraction of the sample size a record-based leaderboard would need. Track your own CLV on the /desk bet log, watch the public table on /leaderboards, and use the 3 to 5% "decent" band as your first benchmark. If your CLV is positive over 300+ bets, you’re building a real process — even if the dollar line still wiggles. The metric exists to tell the truth, especially the parts of the truth that any record-based view would happily hide for another month.
Bet responsibly — set limits, never chase losses.
Named example board
Keep the page grounded with actual decisions. Josh Allen rushing props, Bijan Robinson usage, Puka Nacua target volume, Amon-Ra St. Brown reception stability, and Travis Kelce touchdown equity are all different cases even when they sit on the same fantasy or betting screen. The point is to map the name to the input that matters most.
- Role example: routes, carries, targets, and red-zone work before highlights.
- Market example: spread, total, team total, or prop price before prediction.
- Fantasy example: ADP, roster build, and scoring format before ranking.
- Review example: compare the final result to the original input, not only the box score.
Price examples and pass rules
Use names as evidence, not decoration. The useful SEO win is that Josh Allen, Ja'Marr Chase, Bijan Robinson and Puka Nacua and Bills, Chiefs, Eagles and Lions appear inside decisions, thresholds, and internal links instead of being dumped into a keyword list.
- Spread example: if Chiefs-Broncos opens Chiefs -3.5 and your fair number is -2.8, +3.5 is the bet, +3 is a pass, and the moneyline needs roughly +155 or better before it replaces the spread.
- Total example: if a Bills outdoor total opens 46.5 and wind moves from 8 mph to 21 mph, an under projection at 42.8 still needs a playable number; under 45 or better is different from chasing 43.5.
- Futures example: Bengals AFC North +280 is 26.3% before hold. If your fair number is 30%, stake modestly, track portfolio correlation, and avoid stacking every Burrow, Chase, and Higgins bet into the same thesis.
- CLV rule: a good write-up is not enough. Track whether the spread, total, prop, or futures price closed better than your entry before grading the process.
Use closing-line value guide to keep the examples attached to measurable prices.
Research note board
Use this table to turn the guide into a decision note. The point is to know when the idea is actionable and when it is only context.
| Angle | Input to verify | Example application | Pass when |
|---|---|---|---|
| Market price | Spread, total, moneyline, prop price, or futures hold | Bills and Chiefs compared through PPR | The price has moved past the number that created the edge |
| Football or sport context | Role, pace, weather, injury status, opponent style | Josh Allen role news mapped to the relevant market | The original input changes or remains unconfirmed |
| Review loop | Entry, close, result, and reason code | closing line value logged with a clear thesis | You cannot explain whether the process beat the market |
Model calibration: predicted vs observed
Predicted win probability bucket vs the empirical win rate inside that bucket on the test set. Points on the y=x reference line are perfectly calibrated; points below mean the model is overconfident in that bucket.
Expected bankroll growth at 55% edge
Expected geometric growth of a $100 bankroll under different Kelly multipliers across 1000 bets at p=0.55, decimal=2. Full Kelly maximises long-run growth but produces the deepest drawdowns; fractional Kelly trades growth for variance.
