Is Pinnacle's Closing Line Accurate?

Pinnacle's closing line is a very good estimate of the true probability on main football 1X2 markets, not a perfect price. Joseph Buchdahl's 2016 sample of 132,645 matches found almost no gap between opening and closing efficiency (R² 0.996 vs 0.997), and the opening-to-closing ratio predicted later yields. It is weaker on thin limits, props, and any market still moving.

This page is the accuracy question. For the formula, start with closing line value. For opener versus close as a timing choice, opening line vs closing line. For how many bets you need, CLV sample size.

Definition. A closing line is the last pre-event price a bookmaker posts before the match starts. Accurate here means that, over a large sample, the fair (no-vig) close matches how often that outcome actually happens. It does not mean the next bet is a lock.

Key Takeaways

• On 132,645 global football matches from 2007, Buchdahl (Pinnacle, 20 July 2016) measured Shannon entropy of 1.457 at the opener and 1.452 at the close after removing the margin. R² of implied vs observed probabilities: 0.996 open, 0.997 close.
• Across 397,935 home/draw/away outcomes in the same sample, the opening-to-closing odds ratio predicted later yields; the same plot against closing odds showed almost no leftover edge.
• On English league football 2016/17–2018/19 (6,108 matches, 12,216 team scores), Buchdahl (Pinnacle, 11 May 2020) found an average fair-odds score of 0.0030 against an expected 0 if the close were perfect on average.
• A capture is not a close. Leonard −3.5 on 28 October 2026 moved 1.826 → 1.543 (fair 1.631). If you took 1.826 and the market stayed there, that is +12.0% versus the post-move fair price, not a verified close.
• Use the current fair price to decide; use the fair close afterwards to measure. In-play prices are a different market.

What does "accurate" mean for a closing line?

It means the fair close is an unbiased probability, on average, over many matches. If sides priced at a fair 2.00 win about half the time, and sides at a fair 4.00 win about a quarter of the time, the line is doing its job.

It does not mean:

• The close is right on this match.
• Beating the close tonight proves skill.
• A soft book's close is a substitute. Measure against Pinnacle's fair close, not the book that limited you.

The usual benchmark is Pinnacle because it takes large bets from winning players and runs a low margin. That method is Pinnacle as the sharp reference price. Accuracy is a property of a sample, not of one quote.

What does Buchdahl's 2016 football sample show?

Joseph Buchdahl published the numbers on Pinnacle's Betting Resources on 20 July 2016 (re-read 7 November 2026). He used 132,645 global football matches stretching back to 2007, 1X2 market, Pinnacle margin removed.

Entropy. Average Shannon entropy per match: 1.457 at the opener, 1.452 at the close. The close is slightly more certain, but the gap is tiny. His worked example for fair 2.00 / 3.50 / 4.67 has entropy 1.492.

Results vs odds. Implied probabilities plotted against observed win frequencies: R² 0.996 (open) and 0.997 (close). Almost all of the variation in results is explained by the prices.

Opening-to-closing ratio. For 397,935 home/draw/away outcomes, the ratio of opening odds to closing odds predicted later yields in a strong linear way. A side that opened 10% longer than it closed returned about 10% to the opener. The same plot against closing odds showed almost no correlation: once Pinnacle has closed, there is little leftover edge in that sample.

The 2016 article is about football 1X2. It is not a study of tennis handicaps, player props, or in-play.

What did the 2020 English-league test add?

On 11 May 2020 Buchdahl published a second check on the same idea, this time with Pinnacle's closing fair odds on professional English league football, seasons 2016/17 to 2018/19 (re-read 7 November 2026).

He scored each team from the fair close, then looked at the distribution of those scores against actual results.

The average sitting on 0.0030 is the accuracy claim: on this sample, Pinnacle's fair close was not systematically high or low. The six-match spread matching a random draw is the variance claim: a handful of results will not tell you whether a process beats the close. That is why CLV sample size talks in hundreds of bets, not weekends.

Buchdahl notes he had to estimate how to remove Pinnacle's margin. The result is still a T1 public study on 1X2, not a guarantee on every market.

When is Pinnacle's close weaker?

Low limits. A Championship 1X2 or an ITF games handicap can move on a few bets. The close is then the last price a thin market reached, not a deep consensus. See low-limit odds moves and Pinnacle limits explained.

The market is still moving. If the price is changing in the last minutes, the fair number you computed a minute ago is already stale. Accuracy is a statement about the final pre-match price.

Derived and niche markets. Totals, BTTS, player props and same-game combinations are not the 1X2 sample above. They can be internally consistent with the main line and still mispriced relative to each other; that is correlated markets.

You used the quoted close. Raw CLV is flattered by Pinnacle's margin. No-vig CLV is the stricter figure. Worked methods: devig methods.

You used another book's close. A soft closer can be shaded. The test is Pinnacle's fair close, which is also how you judge a value bet.

How do you measure a real drop against that close?

Pinnacle Terminal does not keep odds history, so a capture is the price at that moment, not a verified kick-off close. Treat the figures below as CLV versus the post-move fair price on the capture. If the line later rebounded, the true close would be different.

ITF Women Poitiers, Leonard v Hesse, games handicap Leonard −3.5. 28 October 2026, about 09:37 Madrid (CET). Pinnacle 1.826 → 1.543, labelled Limit-backed. Fair price (Power): 1.631.

ITF Women Taby, Bernstein v Pulchartova, handicap Bernstein −1.5. Same morning. Pinnacle 3.52 → 2.88, labelled Sharp drop. Fair (Power): 3.138.

The raw column is larger because it still includes Pinnacle's margin. Report the no-vig column. A bet365 price that looked long against the new fair Pinnacle number is judged the same way: against the close, not against whether the selection won. That is the FAQ we added on bet365 dropping odds.

How Pinnacle Terminal handles this

Pinnacle Terminal streams Pinnacle's prices and limits live and shows the fair price (Power is the default devig method) on each line. Every pre-match move is graded 0–10 (Big move, Sharp drop, Limit-backed, Downtrend, Rebound, Standard). Live in-play football drops are shown and filtered, without a grade. The tracker records Pinnacle's closing price for every bet you log and computes CLV against the fair close with four devig methods plus a worst case. The terminal keeps no odds history of its own: it shows moves as they happen. Product: Pinnacle dropping odds. Tracker: CLV tracker guide.

FAQ

Is Pinnacle's closing line accurate?

On main football 1X2 markets, public samples say it is a very good average estimate. Buchdahl's 2016 set of 132,645 matches had R² 0.997 at the close; his 2020 English-league set of 6,108 matches had an average fair-odds score of 0.0030 versus 0. It is not a claim about every prop or every limit.

Is the opening line almost as accurate?

In that 2016 sample, yes: entropy 1.457 vs 1.452, R² 0.996 vs 0.997. The leftover edge sat in the move from opener to close, not in a wild opener.

Does beating the close mean I will profit?

It is the best early evidence that your prices were good. Results still take a much larger sample. See ROI vs CLV. Blind longshots at that same close still lose more than blind favourites (favourite-longshot bias); beating the close is a statement about your price, not about backing every short number.

Should I use Pinnacle's quoted close or the fair close?

The fair close. Raw CLV is always flattered by the closing margin.

Are in-play closing prices the same test?

No. The studies above are pre-match 1X2. In-play is a different market, and Pinnacle Terminal does not grade in-play moves.

Sources

• Joseph Buchdahl, "The maths behind Pinnacle's 'winners welcome' policy", Pinnacle Betting Resources, 20 July 2016 (re-read 7 November 2026): 132,645 matches; entropy 1.457 / 1.452; R² 0.996 / 0.997; 397,935 H/D/A outcomes.
• Joseph Buchdahl, "Another look at randomness and efficiency in soccer betting", Pinnacle Betting Resources, 11 May 2020 (re-read 7 November 2026): EFL 2016/17–2018/19; 6,108 matches, 12,216 scores; average score 0.0030; six-match SD 0.1866 vs SEM 0.1860.
• Leonard −3.5 and Bernstein −1.5 prices: Pinnacle Terminal captures, 28 October 2026 (same figures as how to track sharp money and arbitrage vs value betting). CLV arithmetic on this page, 7 November 2026. Not verified kick-off closes.

Try it

Log each bet against Pinnacle's fair close, with the close recorded from the same feed that graded the move: start the Pinnacle Terminal 7-day free trial. Nothing is charged before day 8. Plans: pricing.

18+ · Bet responsibly. Only bet money you can afford to lose.