How We Measure Ourselves: Grading and CLV Methodology
This page documents exactly how GamblingSite.com measures its own betting picks: how results are graded, how closing line value is computed, how closing lines are captured, and what happens when we find an error in our own data. Every number on our track record and in The Closing Line Report is produced by the process described here, and this page changes only when the process does.
The short version: flat one-unit stakes at the odds we published, every result graded against the verified final score, closing line value measured only against snapshots we can prove were taken before the game started, and a permanent public log of every correction we have made to our own numbers.
What We Grade and How
Every pick we publish is graded against the final result at a flat one-unit stake, priced at the American odds we quoted when the pick went live. Wins and losses settle to the verified final score; pushes return the stake and are excluded from ROI. A pre-match withdrawal or walkover voids the bet and grades as a push, never as a win.
The headline record counts straight single-game bets only. Player props and parlays are graded the same way but tracked in their own clearly labeled segments: a parlay settles all-or-nothing at its combined odds, a pushed leg drops out and reprices the ticket, and blending that variance into the headline would make the headline meaningless. Nothing is ever deleted; a graded pick stays in the ledger permanently, win or lose.
The Metrics, Plainly
Closing line value is the number we hold ourselves to. Over a few hundred picks, wins and losses are mostly noise, but whether the market consistently moves toward or away from our published number is a durable signal about the analysis. A losing pick that beat the close was still a good number; a winning pick that the market ran away from was still a bad one. CLV percentages are price-implied figures computed from odds, never a claim about anyone’s chance of winning.
How We Capture Closing Lines
We snapshot near-closing consensus lines automatically before each game from The Odds API, storing the moneyline, spread, and total across every book on the board at capture time. Featured player props are captured earlier in the evening, inside their own close window, because prop boards thin out near start time. Where a consensus across books is needed, prices are converted to implied probabilities with the bookmaker margin removed and the median is taken in probability space, never as an arithmetic average of American prices.
The pre-game capture rule is absolute: a snapshot only counts as a closing line if its capture time precedes the event’s final settled start time. Sportsbook boards stay quotable after games begin, so a capture job that trusts the board can store an in-play price as a closer without noticing. Any snapshot that cannot be verified as pre-game is discarded rather than counted, and a pick without a verified closing line is counted in the record but excluded from closing line value.
The Correction Log
A track record is only credible if its measurement errors are public too. This log records every correction we have made to our own published numbers, permanently. Corrections run in whatever direction the data says, and where a value cannot be verified we void it rather than guess; that is the no-guess rule.
An audit of our own capture pipeline found 87 stored closing lines that were actually in-play prices captured after games had started. Every affected snapshot was rebuilt from our first-party odds archive using the last capture before each event’s final settled start time, every affected graded pick was re-scored, and two provably in-play values that could not be rebuilt were voided. Across the matched set of 66 tracked picks, our beat-the-close rate fell from 51.5% to 32.8%. The in-play guard now enforces the pre-game capture rule at capture time. Full anatomy in the June 2026 report.
When a spread or total closed on the same number we took, the old scoring counted it as a miss against the close no matter what the prices said, even though player props already broke that exact tie on price. On August 10 we applied the same price tiebreak to spreads and totals and re-scored the full ledger: the beat-the-close rate moved from 38.6% (76/197 tracked) to 48.7% (97/199 tracked), and 81 spread and total picks that previously carried no closing line value figure now carry a priced one. Corrections run in whatever direction the data says; this one ran in our favor, and it is logged exactly like the one that did not.
Sixteen picks graded before our odds archive existed cannot be verified either way; they are disclosed, annotated in the ledger, and left as they are. One value that is provably wrong but penalizes us stays in place with an annotation, because removing only the corrections that hurt would be its own kind of flattery.
Data You Can Cite
Each issue of The Closing Line Report ships a downloadable CSV of its derived aggregates: record, units, ROI, beat-close rate, and average closing line value by league and bet type. The files contain derived figures only, never raw odds-feed rows. Journalists and researchers are welcome to cite the data with a link back to the issue it came from; the numbers regenerate from the same ledger this page describes, so what you cite is what we grade ourselves on.
Update Policy
This page is a living document: it changes when the measurement process changes, and the correction log above only ever grows. Report issues are the opposite: each is a dated snapshot built from the ledger as graded through its stated date, and once published it is not silently edited. If an issue’s numbers are later found wrong, the correction lands here and the issue gets a visible note, never a quiet rewrite. Today’s picks are on our picks page, and every one of them enters this pipeline the moment it is graded.
Play Safe: Gambling should be fun, not stressful. Set limits, stick to your budget, and never chase losses. If you or someone you know has a gambling problem, call 1-800-MY-RESET or visit ncpgambling.org. For more resources, see our Responsible Gambling page.
Frequently Asked Questions
Quick answers to the questions we hear most about how this measurement works.
Why do you grade yourselves on closing line value instead of just wins and losses?
Because on small samples, wins and losses are mostly luck. Whether the market moved toward or away from our published number is a steadier signal of whether the analysis holds up, and it can be measured on every pick that has a verified pre-game closing line.
What happens when you find a mistake in your own numbers?
It goes in the correction log on this page, permanently. We rebuild what can be rebuilt from our own odds archive, void what cannot be verified rather than guessing, and re-publish the corrected numbers even when they are worse. The July 2026 in-play capture repair is the first entry.
Can I use the numbers from your reports in my own article or study?
Yes. Each report issue ships a CSV of derived aggregates you can download and cite with a link back. The figures are derived from our public graded ledger, so anyone can check them against the track record page at any time.