The Anatomy of a Collapse Research

Collapse is a word commentators reach for and almost nobody defines. Before any of the numbers below mean anything, the line has to be drawn somewhere and then held there. Ours: a collapse is at least 4 wickets falling for at most 30 runs, read straight off the fall-of-wickets column of a real scorecard. Where several runs of wickets overlap, the longest one is kept, so a single slide is counted once rather than three times.

That is a choice, not a law of the game. Move the line to five wickets, or to twenty runs, and every count on this page changes. A stricter line would find fewer slides and make each one look more fatal; a looser one would find more and dilute them. What would most likely survive such a change is the shape: the ordering of the phases, the drift by decade, and the size of the gap between the sides who collapse and the sides who do not. That is an expectation about this data rather than something tested here. Treat the pattern as the finding and the exact counts as one reasonable way of putting a number on it.

One thing the measure deliberately does not know is time. The fall-of-wickets column records runs, not balls, so 4 wickets for 30 runs across a long, grinding session counts exactly the same as 4 in a dozen deliveries. The typical Test slide is 4.5 wickets for 22.6 runs, and 92.6% of them came in an innings that ended with the side all out. In ODIs it is 4.5 wickets for 23.0 runs, with 78.0% in an innings that ended all out. A collapse, by this definition, is rarely just an interruption in an innings: the innings it happens in usually ends with the side bowled out.

The first surprise is how ordinary it is. At least one collapse turns up in 55.9% of Tests, 1,417 of the 2,537 in the archive. In ODIs the figure is 33.8%, or 1,682 matches. Rather than a once-a-season catastrophe, this is something closer to a feature of the sport: more Tests contain a collapse than do not.

The gap between the formats is mostly about opportunity rather than fragility. Per innings the two are close: a collapse appears in roughly one eligible Test innings in 4 (0.244 per innings) and one ODI innings in 5 (0.214). A Test offers up to four innings and an ODI only two, and the eligible innings bear that out at 3.3 per Test against 1.7 per ODI. Roughly twice the innings, roughly twice the chance that one of them falls apart.

There is a second difference underneath it. An ODI innings can simply run out of overs with wickets in hand, so a good many lower orders are never exposed at all; a Test innings usually ends only when the tenth wicket falls or a captain decides it has seen enough. That shows up in how those innings finish: 92.6% of Test collapses come in an innings that ends all out, against 78.0% in ODIs.

Now the number that carries the piece. Of the 1,732 Test collapses that happened in matches somebody won, the side that collapsed won 369 and lost 1,363: a win rate of 21.3%. Sides that got through a Test without a collapse anywhere in their batting won 67.2% of their decided matches, 1,411 of 2,099. That is a gap of about 46 points. In ODIs the same comparison runs 16.1% after a collapse, from 1,816 decided cases, against 57.6% without one: a gap of about 42 points.

Read that carefully, because it is the easiest number on this page to overstate. It is not a causal estimate, and nothing here says a collapse costs a side roughly 46 percentage points of win probability. Two things are tangled up in it. The first is selection: the sides that collapse are, on average, the weaker sides in the match, and the team table further down makes that plain. The second is contamination of the comparison group. In most matches where one side collapses, the other side did not, and that opponent goes into the no-collapse group carrying a win it would probably have had anyway. The no-collapse figure is therefore inflated by the very matches the collapse figure is drawn from.

What survives both objections is still worth knowing. A collapse is an extremely reliable marker of a match going badly: four Test sides in five and something close to six ODI sides in seven do not come back from one. Whether the slide caused the defeat or merely announced it is a question the fall-of-wickets column cannot answer.

Slides do not begin evenly through an innings. In Tests the middle order is where the largest share of them start, 42.8% of the total against 29.1% in the top order and 28.1% in the tail. ODIs are much the same shape, 45% middle order to 30.2% top and 24.8% tail. That is partly arithmetic, because wickets 4, 5 and 6 sit in the middle of the ladder and a run of four has more room to start there, and partly cricket: the middle order is where a new batter arrives against a ball that has begun to move or turn, with the innings still to be built.

Survivability is where the two formats part company. In Tests a tail collapse is the most survivable kind, 24.4% against 19.7% for a top-order slide and 20.4% for a middle-order one. In ODIs the order inverts: the tail is the worst place to lose a clutch of wickets at 12.6%, and the middle order the least damaging at 19.4%. One reading is that a Test side that loses its last four wickets cheaply has often already banked a respectable score and has three more innings in the match to work with, while an ODI side that loses its tail cheaply has capped its total at exactly the moment the total was meant to accelerate. The tail slides are also shorter, and shorter by construction: a slide that starts at the seventh wicket can run to the tenth and no further, so every one of them is exactly 4 wickets, against 4.8 on average in the middle order.

Most collapses stop at the minimum. 1,331 of the 2,037 Test slides are exactly 4 wickets long, and those are the ones sides most often survive, at 23.8%. Push the slide to six wickets and the win rate falls to 10.5% off 171 cases. The longest category, 7 or more wickets, sits at 12.3%, which is above the six-wicket row and should not be read as a recovery: it rests on 70 slides and 65 decided matches, few enough that a couple of results move it several points.

The ODI column has no such wobble at the bottom. Of the 52 ODI slides of 7 or more wickets, the side went on to win 1 of 51 decided matches, 2.0%. Losing seven or more wickets for under 31 runs in a fifty-over innings is, on this evidence, close to a result in itself.

The raw count of collapses climbs, unevenly, through the archive, from 22 in the 1920s to 347 in the 2010s, but that mostly tracks how much Test cricket was played rather than how badly it was batted. The interesting movement is in what happened next.

In the 1960s, 32.1% of Test collapses ended in a draw or a tie, and of the ones that were decided the batting side still won 27.5%. By the 2010s only 7.5% finished undecided, and the win rate among decided matches had fallen to 18.7%. The draw was the escape hatch. As Test cricket has produced fewer of them, a collapse has turned from a setback you could survive by batting out time into something very close to a result. The 2020s bar rests on 146 collapses and a Test archive that stops in 2024-03-30, so read it as a partial decade rather than a trend.

Every collapse has two sides to it. Ranking teams by collapses inflicted minus collapses suffered gives a rough register of who has spent the archive applying the pressure and who has spent it absorbing. In Tests Australia lead on 413 inflicted against 314 suffered, a net of 99, and Bangladesh sit at the other end with 27 against 91. The ODI ranking runs the same way, and the table below carries both formats so the figures can be checked rather than taken on trust.

This is the softest table on the page and should be read as such. Net collapses reflect era and opposition at least as much as any quality of a side: teams that played most of their cricket long ago appear with large counts in both columns simply because they played the most matches, and teams admitted to the top level recently have faced established sides far more often than they have played each other. A deficit here is a statement about the fixtures as much as about the batting.

Each format ranked by net, within its own archive. A side needs at least 40 collapses across both columns to appear, which is why the lists are short, and the two formats are not comparable to each other: a Test side has two innings a match to collapse in and an ODI side has one.

At the extreme, the slide and the innings are the same thing. The heaviest in the Test archive is New Zealand losing 10 wickets for 26 runs against England at Eden Park, Auckland in 1955, from 0 to 26, all out for 26. The worst in the ODI archive is Afghanistan against Scotland at Sheikh Zayed Stadium, Abu Dhabi in 2015: 9 wickets for 25, from 38 to 63, all out for 63.

None of the eight heaviest Test slides was suffered by a side that went on to win. One did escape defeat: Australia lost 9 for 27 at Edgbaston, Birmingham in 1902 and the scorecard still reads "Match drawn", a perfectly respectable way out of an innings that finished on 36.

Three things come out of the fall-of-wickets column. A collapse is common, not freakish: more than half of all Tests contain one and a third of ODIs do. Where it starts matters, and it matters differently by format, with the tail the safest place to lose wickets in a Test and the most expensive in an ODI. And length is close to destiny, with each extra wicket in the slide pulling the win rate down until, at the far end of an ODI innings, recovery all but disappears.

You can feel the same arithmetic from the other side of it in The Follow-On, CricArcade's Test game, where two quick wickets change what the next hour is worth. Our guide to settling in, and the myth of the nervous nineties covers the other half of the story: why the batter walking in during a slide is the most vulnerable player on the field. How a collapse compounds inside a one-day chase is in our guide to the required run rate and collapses.

• The Test archive runs to 2024-03-30 and the ODI archive to 2026-08-13. Matches played since then are not included, and the last decade bar on the chart is a partial one. The ODI archive also holds no Afghanistan ODIs after 12 March 2024, which its source withholds.
• Men's international cricket only. Nothing here covers women's cricket, first-class or list A domestic cricket, or any Twenty20 format.
• The definition is a choice. 4 wickets for 30 runs is one defensible line among many, and a different line would change every count on this page. The ordering of the findings is more robust than the figures themselves.
• The measure has no clock. Runs conceded between wickets are all it sees, so a slow slide across a session counts the same as a sudden one inside a few overs.
• Innings without a clean fall-of-wickets ladder are skipped, so a small number of innings, from every era of the archive, never get the chance to qualify.
• The win-rate comparison is not causal. Sides that collapse tend to be the weaker side in the match, and the no-collapse group is inflated by the opponents of the sides that collapsed, who usually won those same matches. The gap measures association, not the cost of the collapse itself.
• Several cells rest on small samples: the longest Test slides number only 70, the longest ODI slides 52, and the earliest decades a few dozen each. Percentages built on those move a long way on a single result.
• Draws, ties and no-results are left out of every win rate, so a collapse that turned a likely defeat into a draw does not appear as a partial escape anywhere in the figures.
• Net collapses by team reflect how many matches a side has played and who it played them against, not batting quality alone.

Figures computed from CricArcade's match archive with our own analysis scripts. Drafted with AI assistance, checked against the scorecards and edited by Vishal Jaiswal. Found a number that looks wrong? Tell us and we will check it against the scorecard.

• CricArcade Test match archive: 2,537 Tests, 1877-03-15 to 2024-03-30
• CricArcade ODI match archive: 4,981 ODIs, 1971-01-05 to 2026-08-13

More research: Does Winning the Toss Win Tests?, Batting First or Second?, How Strong Is Home Advantage?, How Often Is 300 Chased Down?.