Over 4.5 Cards Hit Rates 2026: 25,298 Matches Measured

A bookmaker quotes you over 4.5 cards at 2.40 and nothing on the coupon says whether that is generous or robbery. The answer is the gap between 2.40 and the 2.52 the real frequency demands, a quiet loss of five per cent a bet. This page gives the hit rate for every total cards line, the odds you need to break even, and how far the league, the season and the referee move it.

We counted the cards rather than quoting anyone. The sample is every completed fixture in the nine leagues behind our referee pages, 2018/19 to 2025/26, from football-data.co.uk records downloaded on 1 November 2026: 25,298 matches and 104,784 cards, 100,020 yellow and 4,764 red.

Three findings do not appear in any card guide ranking for this market. The league beats the referee as a signal, the baseline window matters more than the current season, and the referee effect survives an honest test at a quarter of its advertised size.

Over 4.5 cards hit 39.7 per cent of 25,298 matches

Start with the frequencies, because everything after this is a correction to them. A match produced 4.14 cards on average, and the five lines a bookmaker offers broke as follows.

The break-even column is the useful one and nobody publishes it. A hit rate alone cannot tell you whether a bet is good. Over 4.5 cards needs 2.52 to be a coin flip, and the typical quote on a mid table fixture sits between 2.20 and 2.60. This market is close enough to fair that the corrections below decide which side of it you are on.

Read the under column too. Under 4.5 at 1.66 is the same bet as over 4.5 at 2.52, and the two quotes are rarely consistent once the margin is split. Checking both sides here is the fastest way to find the half a bookmaker priced lazily.

The shape of the distribution, and why 4.5 is the pivot

A hit rate hides what actually governs a card bet: how tightly the outcomes cluster. The median match produced exactly four cards, and over half landed on three, four or five.

The 4.5 line sits almost exactly on the mode, where 17.6 per cent of matches pile up on four cards alone, so one extra booking in the 88th minute flips roughly one match in six. That is why 4.5 is the most traded card line, and where a small error in your estimate costs the most.

Then the tail. Matches of ten cards or more are 2.15 per cent of the sample and the worst reached nineteen, so the under has bounded upside and a fat tail of disasters. Our booking points guide prices that tail directly, a red weighing 25 and a yellow 10.

The league is the biggest signal in the market

Here is the number that should reorder most people's card betting. The gap between the quietest and rowdiest league is 38 points of over 4.5 frequency, larger than any referee, team or home advantage effect we have measured on this site.

In prices, over 4.5 cards in the Primeira Liga is a 1.63 shot and the same bet in the Eredivisie is 4.24. Carry a Premier League instinct into a Portuguese fixture and you are wrong by 30 points of probability, which no referee research rescues.

The over 2.5 column matters too: a Primeira Liga match clears it 86.8 per cent of the time, 1.15 in decimal terms, so that line is dead money in a high card league whoever referees it. Why Portugal rather than Italy sits on top is our league card volume guide.

The league ordering has held for eight straight seasons

A spread is only useful if it is stable, and that is where the one page ranking here with real card data falls down: it ranks leagues on the current season, about 100 matches each in early November. Split ours into its eight seasons instead.

The top is close to immovable. Portugal and Spain finished top three in eight seasons out of eight, the Eredivisie bottom three in seven of eight. A league baseline is something you can carry into next weekend. A 100 match leaderboard is not.

One row is moving, and it is the one with the loudest reputation. Serie A has fallen from 62.1 per cent to 35.3 in six seasons and finished bottom three in 2025/26. Italian football's disciplinary reputation is now the most reliably mispriced assumption in card markets, because folklore updates slower than officiating.

Card rates drifted up, so your baseline window matters

The Premier League row above does what no row should do if a long average is safe: it nearly doubles, from 25.5 per cent in 2018/19 to 44.2 in 2023/24.

That jump has a cause and a date. For 2023/24 the Premier League, the FA, the EFL, the PFA, the LMA and PGMOL adopted a Participants Charter that lowered the threshold for a dissent booking and told officials to card time wasting, with dissent cautions trebling across the opening 16 game weeks. The Law 12 offences did not change; the threshold for applying them did.

Use the eight season Premier League figure of 31.7 per cent and you are pricing a 2020 league. The last three completed seasons give 40.7 per cent, which is 2.46 rather than 3.15. Pooled across all nine leagues the last three seasons give 39.9 per cent, almost identical to the long run figure, but that stability is a coincidence: England rose 17.2 points and the Championship 11.1 while Serie A fell 13.4 and La Liga 7.2. The 2020/21 row, played in empty stadiums, is the lowest season on record in five of the nine leagues, for the reasons set out in our home bias guide.

Premier League referees ranked on over 4.5 cards

Now the signal every card guide calls the dominant one. Of the 23 Premier League referees with 50 or more fixtures from 2018/19 to 2025/26, here are the eight highest and seven lowest by over 4.5 cards. Note what the raw table appears to say.

At face value this is an enormous effect. John Brooks at 47.3 per cent and Martin Atkinson at 12.4 are 35 points apart, which in price terms is 2.11 against 8.06. If that were real, current and usable it would be the most valuable table on this site, and it is roughly the table every tipster presents when it calls the referee the dominant signal.

It is not usable, and the reason is the previous section. Per referee over rates for the current squad sit on each referee page, and our strictest referees guide ranks the current group on volume. Here is why neither should be read the way that table invites.

Most of the referee spread is the era, not the referee

Look at the four most lenient officials in that table: Jonathan Moss, Graham Scott, Andre Marriner and Martin Atkinson. Then check when they worked.

Three of those four refereed not one Premier League match after the 2023/24 change, and the fourth took six. Kevin Friend took none, and nor did Mike Dean. They are not lenient officials. Their entire sample was drawn from a league that booked at roughly two thirds of today's rate, while four of the six strictest worked most of their matches after the reset.

The fix is to compare each referee against the seasons they actually worked: expected rate is the league's average over 4.5 rate across that referee's own fixtures, and the edge is what is left.

The spread collapses from 35 points to 23 and the standard deviation across the 23 officials falls from 9.1 points to 6.8. Restrict it to the 18 referees with 30 or more matches in the last three seasons, the group taking fixtures now, and the entire active range is 32.1 to 52.5 per cent around a 40.7 per cent baseline: real, worth money, two thirds the size the raw table advertises. Michael Oliver is the one reputation that survives, minus 5.8 points across 224 matches. Sam Barrott reads more lenient still at minus 7.9, but on 62 matches, and that difference decides everything.

What a referee's record actually predicts

Every guide that tells you to bet cards off referee averages makes a claim it never tests: that a referee's past predicts their next match. Take every official with 80 or more matches in the Premier League, Championship and Scottish Premiership, split their fixtures in half chronologically, and compare the halves.

Across all 46 officials the correlation between first and second half edge is 0.27, so a track record explains about 7 per cent of what a referee does next. The sign held in both halves for 31 of 46, 67 per cent against a coin flip's 50. The effect is real. It is nothing like the effect the tables imply.

The reason is sample size, and it is arithmetic rather than opinion. A hit rate near 40 per cent carries a standard error of nine points on 30 matches.

Set that against an active range of 20 points. A 50 match sample carries an interval of about 14 points, wider than the whole genuine spread between the strictest and most lenient official working today. Sam Barrott's minus 7.9 on 62 matches is indistinguishable from zero. Only 150 and 200 match samples, four or five seasons in one division, measure anything.

Pricing a total cards bet with all three signals

So here is the method the numbers support, ranked by the evidence.

1. Start from the league, over the last three seasons, not the last eight. Largest effect, most stable, cheapest to get right. Premier League over 4.5 cards is 40.7 per cent, not 31.7.
2. Adjust for the referee, then throw most of the adjustment away. At a correlation of 0.27, keep about a quarter of the gap between the referee's era adjusted rate and the league baseline.
3. Do nothing until the appointment is published. An unappointed fixture has a league number and nothing else, and the appointments page is where that changes.

Worked through on the most favourable case available. Stuart Attwell runs at 52.5 per cent over 4.5 across his last three seasons against a 40.7 baseline, a gap of 11.8 points on 61 matches. Keep a quarter and you get 43.7 per cent, so the fair price is 2.29, not the 1.90 his raw recent rate suggests. Take 2.00 believing the raw number and you are not getting value, you are losing 12.6 per cent a bet. That correction is the practical content of this page.

The same logic drives the neighbours. Distribution rather than volume decides both teams to be booked, so how the lines are built is worth reading alongside this.

How we counted, and what this does not cover

Every completed fixture in nine domestic leagues, 2018/19 to 2025/26, from football-data.co.uk result files downloaded on 1 November 2026. A card is one yellow or one red as the source records it. Referee names are the source's abbreviated strings mapped to our canonical entities, and the methodology page sets out that resolution.

Four limits.

A second yellow counts as two cards here, one yellow and one red, because that is how the source records it. Bookmakers differ, and some settle a second yellow as one card. Check the settlement rules on the specific market, because that rule difference on one match in twenty is larger than most of the edges above.

Referee figures exist only for England and Scotland. The source files carry a referee column for the Premier League, Championship and Scottish Premiership and not the other six, so the league rows are complete while the referee analysis is not. Our league referee pages carry per referee over rates where we hold that data.

Break-even odds ignore commission and margin. They are the price at which a bet is a coin flip, not the price at which it is worth placing. Beat them, do not match them.

This is a baseline, not a prediction. It accounts for nothing about the two teams, the stakes or the derby. It tells you what the distribution has done, which is a starting point and not a finish.

For reference, 2026/27 to 1 November 2026 stands at 785 matches, 3.89 cards per game and 35.5 per cent over 4.5. Serie A sits at 26.7 per cent and the Bundesliga at 23.8, both well below their recent form, and on 90 and 63 matches that is exactly the sample this page warns you not to price off.