Cricket statistics glossary: every key term explained

A batting average is a fraction with an argument buried in its denominator, and almost nobody quoting one knows the argument is there.

That is the trouble with a cricket statistics glossary that only defines terms. Cricket has accumulated its vocabulary over a very long time and thrown almost none of it away, so a modern statistics page carries measures invented for a handwritten scorebook sitting beside measures that could not exist before ball-tracking cameras. The two sets disagree about what cricket is. Some of the words mean roughly the opposite of what they look like. Several are computed differently by different publishers. A few sound like jargon and are doing precise arithmetic underneath.

This page does two jobs. The first part explains what the major figures are for, meaning the question each one answers and the question it will quietly get wrong. The second part is the reference list: every term you are likely to meet on a scorecard or a career page, defined properly. The last section is the three questions worth asking of any cricket number before you repeat it.

• 6Legal deliveries in an over
• 10Wickets that end a completed innings
• 5Dismissal types credited to the bowler
• 200First-innings lead needed to enforce the follow-on in a Test

Rule-defined counts rather than performance figures. The follow-on margin varies with the scheduled length of the match; 200 is the figure for a five-day Test.

The batting average and the argument in its denominator

Runs divided by dismissals. Not runs divided by innings. The choice of denominator is not an accounting convenience, it is a claim about what a batter is.

The claim is this: a side has ten dismissals per innings and no more, so a dismissal is the scarce resource a batter consumes. The average therefore measures runs produced per unit of scarce resource. That is a genuinely good idea. It is why a batter who scores 60 and 0 and a batter who scores 30 and 30 are treated identically, which is correct, because they cost the side the same and produced the same.

For a top-order batter the idea holds almost perfectly. An opener's innings ends in a dismissal nearly every time. Being not out at the end is a rare accident: a declaration, a completed chase, the last wicket falling at the other end. The occasional not-out barely moves the number, and the average is a fair estimate of what one innings from that player is worth.

For a number eight it collapses.

A lower-order batter is not out constantly, and the reason is never that they survived. It is that the innings ended around them. The last recognised batter was dismissed, or the fifty overs ran out, or the captain declared, or the target was reached. None of those is evidence about the batter's skill, yet each one adds runs to the numerator and nothing to the denominator. The censoring is not random with respect to the batter. It is systematically produced by their position in the order.

Consider two invented careers with the same number of innings.

• Batting average
• Runs per innings

Batter A, opener4036

Batter B, number eight3217.6

Batter A plays 40 innings with 4 not-outs and scores 1,440. Batter B plays 40 innings with 18 not-outs and scores 704. Both careers are invented to make the arithmetic legible.

Show the numbers

Worked example: what not-outs do to two invented careers

Item	Batting average	Runs per innings
Batter A, opener	40	36
Batter B, number eight	32	17.6

Batter A has 36 dismissals, so the average is 1,440 divided by 36, which is 40.0. Runs per innings is 1,440 divided by 40, which is 36.0. The two figures are four apart, and the average is telling close to the truth.

Batter B has 22 dismissals, so the average is 704 divided by 22, which is 32.0. Runs per innings is 704 divided by 40, which is 17.6. The gap is 14.4, and it is the whole difference between a useful number eight and a batter who averages in the low thirties. Anyone reading the two averages side by side, 40 against 32, would conclude that A is somewhat better than B. In terms of what actually arrives in a given innings, A is worth more than twice as much.

The fix costs nothing. Put runs per innings next to the average. The gap between them is the size of the distortion, expressed in runs. Close together means the average is honest. Far apart means the average is describing a batting position rather than a batter.

This is also what makes cross-career comparison harder than it looks. A player who batted at three early and at six later has one career average describing two different jobs, blended in a proportion nobody records. A nightwatchman's Test average is dragged down by innings that had nothing to do with his batting ability and everything to do with the light. And a specialist finisher in white-ball cricket will carry an inflated average and a deflated one at the same time, because the not-outs push it up while the instruction to attack from the first ball pushes it down.

Strike rate means something different in each format

Runs per 100 balls faced. The arithmetic never changes. What changes is what the number is competing against.

In Twenty20 the scarce resource is not wickets, it is deliveries. An innings ends after 120 balls whether or not anyone is dismissed, so a ball used without a run scored is a permanent loss and a wicket is often survivable. Strike rate is the primary batting statistic in that format and the average is the secondary one. A batter averaging 25 at a strike rate of 150 is usually more valuable than one averaging 35 at 115, and the arithmetic of the format says so plainly.

In a fifty-over match both resources bind. There are 300 balls and ten wickets, and a side that loses the wickets cannot use the balls. Strike rate matters, but only read against the phase of the innings it was accumulated in. A strike rate of 90 through the middle overs against spin and a strike rate of 90 in the last five are not the same performance, and a career figure merges them.

In a Test, strike rate is mostly a description of intent and situation. The same figure can be a virtue or a crime depending on whether the side is setting a target on the fourth evening or saving a follow-on on the third. What a Test strike rate does tell you, over a large enough sample, is how much time a batter's runs cost his own side, which is the resource a Test match is actually played with.

The cleanest way to state the difference: in a Test a wicket costs far more than a dot ball, and in a Twenty20 a dot ball costs nearly as much as a wicket. Everything about how the two figures trade off follows from that one exchange rate. The longer argument about which of the two should be read first turns on exactly this point, and it is the reason the same batter can be badly underrated in one format and overrated in another without changing anything about the way they play.

One more thing about strike rate, which nobody prints. It has no not-out distortion at all, because balls faced are counted the same whether the innings ended in a dismissal or not. That makes it structurally cleaner than the average. It also has no duration information whatsoever, so it needs volume before it means anything.

Three bowling figures, three different questions

Bowling produces three headline rates, and the common mistake is treating them as three versions of the same judgement. They are answers to three separate questions.

Economy rate is runs conceded divided by overs bowled. The question is: what does an over from this bowler cost?

Bowling strike rate is balls bowled divided by wickets taken. The question is: how often does this bowler break through?

Bowling average is runs conceded divided by wickets taken. The question is: what does a wicket from this bowler cost?

The three are locked together by an identity worth memorising, because it lets you recover any one of them from the other two:

bowling average = economy × bowling strike rate ÷ 6

Which one matters depends entirely on the constraint the format imposes. In a Test the side must take twenty wickets inside five days, so the binding constraint is time. Bowling strike rate is literally the rate at which that constraint is discharged, and it is the primary measure. Economy matters in a Test mainly as a proxy for pressure and for how many overs the captain can afford to give a bowler before the attack tires.

In Twenty20 the reverse. There are no twenty wickets to take and the innings ends on schedule, so wickets matter chiefly because they slow the scoring down. Economy is the primary measure, and a death bowler with an unremarkable average and a very low economy is doing the job the side picked him for.

Fifty-over cricket sits between the two, and the phase decides. In the middle overs a wicket genuinely reduces the eventual total, because the incoming batter has to start again with plenty of balls left; at the death it barely does, because there is no time for the loss to compound.

Now the loss of information in the average. Two invented bowlers, both averaging 30:

Bowler	Economy	Strike rate	Average
A	3.0	60 balls	30
B	5.0	36 balls	30

Identical price per wicket, completely different bowlers. A gives a captain control and takes a wicket every ten overs. B leaks runs and takes a wicket every six. In a Test where the fourth innings is coming and the clock is the enemy, B is the more valuable bowler and the average hides it. In a Twenty20 A is worth vastly more and the average hides that too. The bowling average merges containment and wicket-taking into a single figure, and merging is a loss you cannot undo without the other two numbers. The full treatment of how the three interact is worth reading before you rank any attack by average alone.

An all-rounder's difference figure is crude, and it works

Subtract a player's bowling average from their batting average. If the result is positive, each wicket they take costs fewer runs than each dismissal they cost. That is the whole test.

It is crude in at least five ways. It treats a run scored and a run conceded as equal, which is true in aggregate and false in situation. It ignores volume completely, so a batter who bowls six overs a match and a genuine second seamer who bowls twenty-five are assessed identically. It ignores format, and the two averages behave differently in each. It is unstable, because bowling averages have fatter tails than batting averages and a poor run with the ball moves the difference further than an equivalent run with the bat. And it says nothing about whether the two skills are available on the same day, which is the real selection question for a player carrying a bowling workload and a top-order batting position at once.

It survives all of that because it is the only single number that asks the right question. The question about an all-rounder is never "is this player good at both". It is "does the side profit from picking this player instead of a specialist", and the difference figure is a rough answer to exactly that. It is also hard to fake. Both halves need volume before they mean anything, so a genuinely positive difference sustained across a long career is difficult to accumulate by accident.

Read it as a screening test, not a verdict. A large positive difference is a reason to look closer. A small one tells you nothing at all.

Aggregates, rates, and why an average of 60 over ten innings is worthless

Every cricket statistic is either a count or a ratio, and the two fail in opposite directions.

Counts are aggregates: runs, wickets, matches, catches, hundreds. They cannot be inflated by selective sampling, they reward longevity, and they are the honest measure of career weight. They are a poor measure of quality, because a long ordinary career outscores a short brilliant one.

Ratios are rates: average, strike rate, economy, conversion rate. They measure quality and they are wildly unstable when the denominator is small.

Take a batter averaging exactly 40, dismissed every innings, and give them one century.

An invented batter averaging exactly 40 with no not-outs, who then scores 100 and is dismissed. The plotted value is the difference between the new career average and 40.

Show the numbers

What one innings of 100 does to a career average

Item	Rise in career average
After 10 innings	5.5
After 25 innings	2.3
After 50 innings	1.2
After 100 innings	0.6
After 200 innings	0.3

At ten innings the century moves the career average by more than five. At two hundred it moves it by three-tenths. Nothing about the innings changed. The only thing that changed is how much the figure was already carrying.

Which is why an average of 60 over ten innings is close to information-free. Ten innings with, say, four not-outs is six dismissals. Six is the entire denominator. One dropped catch on nought, one declaration, one rain-shortened innings, and the number moves by ten in either direction. The batter may be excellent. The figure simply cannot tell you, and the width of the uncertainty around it is greater than the gap between a great player and an ordinary one.

There is a second trap sitting on top of the first, and it catches out more people than the arithmetic does. Every "highest career average" list has a qualification threshold, a minimum number of innings or wickets, and that threshold is doing most of the work in producing the list. Move the cut-off by ten innings and the order changes. When a ranking's contents depend that heavily on an arbitrary boundary, the ranking is a statement about the boundary.

The practical rule is short. Find the denominator before you read the ratio. Dismissals for a batting average, wickets for a bowling average, balls faced for a strike rate, overs for an economy rate. If a page shows you the ratio without the denominator, it is not being careless, it is choosing what you can check. Career pages that show both, like the ones behind our player records, let you do this at a glance.

What ball-by-ball data counts that the scorecard never did

A scorecard records outcomes. It knows that a batter made 40 from 30 balls and was caught at deep midwicket. It does not know whether that was a controlled 40 or a fortunate one, whether the boundaries came off the middle or off the toe, or how many times the bowler beat the bat without reward.

Ball-by-ball recording is an attempt to capture the process instead, and four measures come out of it.

Boundary percentage is the proportion of a batter's runs that arrived in fours and sixes, or in some presentations the proportion of balls faced that reached the rope. Check which definition a page is using, because the two numbers are nowhere near each other. A batter taking seventy per cent of their runs in boundaries is not rotating the strike, which means their contribution is lumpy: the side gets a burst or it gets nothing, and the fielding captain can plan for exactly one of those outcomes.

Dot-ball percentage is the proportion of deliveries from which no run was scored. In limited-overs cricket it is the most direct measure of pressure, and the reason is a transfer mechanism rather than a psychological one. The runs not scored off those balls do not disappear from the requirement. They have to be found later, off fewer deliveries, usually by somebody else. Two batters with identical strike rates and different dot-ball percentages have handed their side very different problems.

Control percentage is recorded by an analyst watching every delivery, not derived from the card. A ball that beats the bat, takes a thick edge, pops up short of a fielder, hits the pad or is badly mistimed is logged as not controlled, regardless of whether it produced runs. False-shot percentage is the same measure inverted, and the two sum to 100. Publishers use one or the other.

Control is the most useful of the four and the least standardised. Its value is that it correlates with what happens next rather than with what has just happened: a batter scoring quickly with poor control is usually about to be dismissed, and a bowler inducing false shots without wickets is usually about to get some. Its weakness is that it is a judgement made in real time by a person, which means two providers watching the same innings will produce different figures, and neither publishes a full rulebook for the call.

The idea underneath all four is the one other sports arrived at independently: judge an action against what was expected in that situation rather than against nothing. Golf did it most completely with strokes gained, which measures every shot against a baseline for that exact position on the course. Cricket's expected-runs and impact measures are attempts at the same trick on a harder problem, because a golf shot has one player and a lie, while a cricket delivery has two players, a field setting, a match situation and a deteriorating pitch. That is also why the analytics departments inside international and franchise sides spend most of their effort on situational splits rather than on new career-level numbers.

Why partnerships are attributed to a pair and not a player

Batting is the only part of cricket where two players are simultaneously responsible for the same unit of production, and partnership statistics are the only routinely kept figures that record it.

The pair is the correct unit because the pair makes the decisions. One faces, both run, and the choice to take a single is made jointly and instantly. More importantly, each batter determines what the other faces. A batter who refuses the single on the fifth ball leaves his partner on strike for the next over. A batter who farms the strike to protect a tailender changes both players' figures without either of them playing differently. None of that appears in an individual column.

Partnerships are also recorded by wicket number, and that is deliberate. A third-wicket stand and an eighth-wicket stand are different jobs performed under different conditions against different bowlers with different fields. Aggregating them would produce a number describing nothing.

Two details catch people out. A partnership total includes extras conceded during the stand, so the two individual scores will often not sum to it. And any partnership can be reconstructed from a scorecard by subtracting consecutive fall-of-wickets figures, which is worth knowing because most cards do not print the partnership column at all.

The weakness is obvious once stated. A stand of 140 is credited identically whether it was split 70 apiece or 125 and 12, and the record does not show which without going back through the innings.

The fielding statistics that barely exist

Catches taken. Run outs, sometimes, and only where the scorer named the fielder. Stumpings, for keepers. That is very nearly the complete list of fielding statistics kept as a matter of course, and it has barely changed in a hundred and fifty years.

What is missing is almost everything: chances offered, catches dropped, half-chances, boundaries saved, misfields, runs conceded or saved by ground fielding, the strength and accuracy of the arm, and positioning, which is the skill that determines how many chances a fielder gets in the first place.

Two structural reasons explain the gap. The first is that fielding is the discipline where the counterfactual matters most and the scorebook records only events. A catch taken is an event. A catch not offered, because the batter declined the shot he would have played against a worse fielder, is not an event, and it may be the more valuable contribution. The second is that a fielder's opportunity count is not his own doing. It is set by where the captain puts him and by what the bowlers make the batters do. Counting catches rewards a slip fielder in a side with a good seam attack and penalises an outstanding fielder stationed where nothing goes.

The dropped catch is the single largest recorded gap in the game. It changes a bowler's figures, extends a batter's innings, can decide a match, and leaves no trace whatsoever on the card. A bowler who has had four chances put down has the same figures as one who induced nothing.

Ground fielding is worse, because the run saved is invisible by construction: the run never happened, so there is nothing to record. Some providers now derive runs saved and conceded from tracking data, which is the most promising fielding measure cricket has produced. It is also proprietary, inconsistently published, and computed differently by each supplier, so it cannot yet be compared across sources.

Even a careful slip-catching count runs into sample size. A specialist slip may be offered a handful of genuine chances in a season, which is not enough to separate a great catcher from a lucky one.

Wicketkeeping figures, and the one stat that punishes a keeper for a bowler's mistake

A keeper's dismissals column is catches plus stumpings, sometimes split and sometimes summed. It is the most-quoted wicketkeeping statistic and it is largely a record of who the keeper kept to.

Catches behind the stumps depend on edges, and edges depend on the bowlers. A keeper standing back to a fast attack that beats the outside edge repeatedly will accumulate catches that a keeper of equal skill standing up to spinners never gets the chance to take. Stumpings depend almost entirely on standing up, which depends on the attack and the surface. Neither number is under the keeper's control in the way the column implies.

Then there are byes, which are unique in cricket. Byes are runs taken when the ball passes both the bat and the keeper untouched. They are charged to the team as extras, not to the bowler, and by convention they are read as the keeper's failure.

They are the only figure in the game that debits one player for another's action. A ball speared down the leg side, a delivery that keeps unexpectedly low out of a rough patch, a bouncer that climbs over the keeper's head: all of them are recorded as byes, and all of them originate with the bowler or the pitch. A keeper standing up to a spinner bowling into footmarks on the fourth day will concede byes that no glovework could have prevented.

And yet byes per innings, or byes per hundred deliveries, remains the best routinely available proxy for glovework, because it is the only one that responds to takes rather than to chances created by others. It is a crude instrument pointed roughly in the right direction, which is more than the dismissals column manages.

What is still missing: takes offered against takes completed, the stumping chances not taken, the leg-side take, and the difference between standing up and standing back. The reasons those matter, and why the discipline resists measurement so stubbornly, are worth a proper look at what keeping actually involves.

Era adjustment: why raw averages across a century compare almost nothing

Put a batting average from the 1930s next to one from the 2020s and the comparison is close to meaningless, because almost every input to the number has changed.

Consider what a fair adjustment would have to account for.

Pitches. Uncovered pitches meant rain could turn a surface unplayable within an hour, and a batter's career could contain innings against conditions no modern player has ever faced. Covering removed that entirely.

The lbw law. It was widened in the 1930s to allow a batter to be given out to a ball pitching outside off stump, and interpretation has shifted repeatedly since. Every widening moved runs from batters to bowlers, and none of it appears in an average.

Over length. Overs have been four, five, six and eight balls long in different countries and periods. Economy rate and maidens are defined per over, so both are directly incomparable across those changes, and bowlers' workloads differ in ways that affect what they could sustain.

The no-ball law. The change from judging the back foot to judging the front foot altered where a bowler could land and therefore how close to the batter the ball was released from.

Equipment and dimensions. Bat pressing and edge thickness, protective equipment that changed how short bowling may be played, and boundary ropes brought in for advertising hoardings and player safety all move the run rate without anybody batting better.

Playing conditions in white-ball cricket. Fielding restrictions, the number of overs per innings, the number of balls used and the rules on bouncers have all changed more than once, and each change resets the baseline.

Opposition. The number of Test-playing nations has grown, so the spread of quality faced differs by era. So does the balance of home and away matches, and the proportion played on seam-friendly against spin-friendly surfaces.

Volume and scheduling. More matches per year, more travel and less recovery change what a career average is averaging over.

Officiating. Neutral umpires and then ball-tracking review changed the rate at which lbw decisions are given, which is a direct transfer between the two disciplines.

A fair adjustment does not correct a player's average. It recentres it. You compare a player against the scoring level of the cricket they actually played, meaning the overall runs conceded per wicket in matches of their era and conditions, and express their figure as a distance from that baseline rather than as a raw number. That is the same reasoning behind every rate-versus-baseline measure in sport, and it is the only approach with any chance of being fair.

Two honest caveats. The baseline for a short era is computed from few matches, so it is noisy. And where a player is one of a handful of dominant figures in a small population, they are part of the baseline they are being measured against.

No era adjustment is standard. Several serious attempts exist, they disagree with each other, and no governing body uses any of them. The defensible position is therefore that a raw cross-era ranking of averages is a ranking of eras, not of players.

Where the data itself runs out

Statistical honesty in cricket eventually comes down to admitting what was never recorded.

Ball-by-ball coverage does not reach back evenly. It is dense for recent international cricket and for the major franchise leagues from their beginnings, and it thins quickly going backwards. Any analysis built on ball-by-ball data is therefore an analysis of the recent past wearing the clothes of a general claim, unless it says otherwise.

Balls faced were not routinely recorded for much of Test history. A batting strike rate simply cannot be computed for a great many older innings. Where a career strike rate is shown for a player whose career spans that boundary, it has been computed over the subset of innings where the data exists, and it is very rarely labelled as such.

The women's game has thinner records going back, both in ball-by-ball data and in scorecard completeness, which makes cross-era work in women's cricket harder still and makes any confident historical comparison there worth checking twice.

Domestic first-class cricket outside the major boards is patchy, so career first-class figures for players who spent time in less-covered competitions are assembled from sources of uneven quality.

Control and false-shot data exist only where somebody was paid to record them, which means televised matches involving well-resourced teams, using definitions that differ between providers.

Fielding tracking data is proprietary, held by suppliers and by the teams that buy it, and mostly not published at all.

Historical scorecards still contain errors. Reconciliation projects continue to amend the record, which means a career aggregate for a nineteenth-century player is not a fixed quantity.

None of this makes the numbers useless. It makes the provenance part of the number, which is why the most useful habit in reading cricket statistics is asking what a figure was computed from before asking what it says.

One innings, five figures moved

These measures resist tidy summary because a single passage of play updates several of them at once, in different directions, and only one of those updates is visible on the scoreboard.

1. 1The innings is played47 not out from 39 balls, two fours, two sixes. The scoreboard records one line. Five separate career measures are about to change, and three of them move for reasons that have nothing to do with how well he batted.
2. 2Batting average rises, wrongly47 runs enter the numerator. No dismissal enters the denominator, because the innings ended around him rather than because of him. His average climbs, and his runs-per-innings figure climbs by less.
3. 3Strike rate is updated honestly47 from 39 is a strike rate of about 120 for the innings, and balls faced are counted the same whether he was out or not. This is the one figure the not-out does not distort.
4. 4Boundary percentage jumpsFour boundaries produced 28 of the 47 runs, so roughly sixty per cent of his scoring came from four deliveries. The other 35 balls yielded 19. A selector reading the total sees an innings; a selector reading the split sees a scoring pattern.
5. 5The partnership is credited to two playersThe 88-run stand goes into the seventh-wicket partnership record for the pair. His 47 and his partner's contribution will not sum to 88, because extras conceded during the stand belong to the partnership.
6. 6His control percentage may have fallenIf ten of the 39 deliveries beat the bat or took the edge, his control was under seventy-five per cent, which is a warning regardless of the runs. The card cannot show it. Only an analyst watching every ball can.

An invented number-eight batter comes in at 96 for 6, makes 47 not out from 39 balls with two fours and two sixes, adds 88 for the seventh wicket, and is stranded when the innings closes. Every figure below moves from that one innings.

Every one of those movements is defensible on its own terms. Read together, they are the argument for never judging a player on a single figure: the same innings makes him look better on one measure, unchanged on another, and worse on a third.

The reference list

What follows is every term you are likely to meet, defined properly. Where a term appears on a match card rather than a career page, the walkthrough of an actual scorecard shows it in position.

Batting terms

Runs. Runs scored off the bat. This excludes byes and leg byes, which are credited to the team and to nobody in particular, and excludes the penalty run charged for a wide or a no ball. Runs scored off the bat from a no ball do count to the batter.

Balls faced. The number of deliveries a batter has dealt with. A no ball counts as a ball faced, because the batter had something to play at. A wide does not, because by definition it was out of reach. Balls faced is the denominator of the strike rate and the single most useful column for judging whether any other batting figure is worth reading.

Not out. A batter still at the crease when the innings ended, marked with an asterisk after the score. An innings can end not out because the overs ran out, because the last wicket fell at the other end, or because the captain declared. It describes how the innings finished, not how well the batter played.

Batting average. Runs scored divided by the number of times dismissed. Not divided by innings. See the first section of this page for what that does and when it is safe to trust.

Runs per innings. Runs divided by innings played, ignoring dismissals. Rarely printed as a column, always computable, and the fastest check on whether an average is honest.

Batting strike rate. Runs divided by balls faced, multiplied by 100. A strike rate of 130 means 130 runs per 100 deliveries. Higher is better. Note that bowling strike rate is a completely different quantity measured in the opposite direction.

Boundary percentage. The proportion of runs scored in fours and sixes, or in some presentations the proportion of balls faced that went to the boundary. Check which.

Dot-ball percentage, batting. The proportion of deliveries faced from which no run was scored.

Control percentage. The proportion of deliveries met as intended, recorded ball by ball by an analyst.

False-shot percentage. The inverse of control percentage. More often quoted from the bowler's side, as the rate of false shots a bowler induces, which is a better measure of a good spell than the wicket column when the catches have gone down.

Duck. Dismissed for nought. The name comes from the shape of the zero on the scoreboard, a duck's egg. A batter not out for nought has not made a duck.

Golden duck. Dismissed for nought off the first ball faced. A diamond duck is a dismissal without facing a delivery at all, which almost always means being run out at the non-striker's end, though some publications reserve the term for a first-ball dismissal without facing. A pair is a duck in both innings of the same match. A king pair is a golden duck in both innings.

Fifty, hundred, double hundred. Milestone scores in a single innings, appearing on career pages as counts. The fifty count usually excludes innings that went on to a hundred, so the two columns are read together.

Conversion rate. The proportion of fifties that became hundreds. A batter who reaches fifty often and rarely goes on is producing innings that look better on a card than they were worth to the side.

Carrying the bat. An opener who bats through a completed innings in which the side is bowled out, finishing not out. It is recorded separately because it means facing the whole innings while ten partners were dismissed.

Nightwatchman. A lower-order batter promoted late in the day, usually in a Test, to face the remaining overs so a specialist does not have to start in poor light against a hard ball. Nightwatchman innings drag an average downwards for reasons unconnected to batting ability.

Anchor, finisher, pinch hitter. Role descriptions rather than measures, and they explain figures that otherwise look wrong. An anchor bats through a limited-overs innings at a moderate rate while others attack. A finisher enters in the closing overs under instruction to attack immediately, and will carry a low average and a very high strike rate because both are products of the job. A pinch hitter is promoted to attack the new ball, an assignment in which failure is the expected outcome most of the time.

Top order, middle order, lower order, tail. Positions one to three, four to six, seven and eight, and nine to eleven. Averages are not comparable across these groups.

Bowling terms

Overs. Six legal deliveries. On a card, a figure of 8.3 means eight complete overs and three further balls, which is eight and a half overs. It is not eight point three, and any economy rate calculated from the decimal as written will be wrong.

Maiden. An over from which no runs are charged to the bowler. Byes and leg byes conceded during it do not spoil a maiden, because they are not the bowler's runs. A wide or a no ball does spoil it. Wicket maiden is a maiden in which a wicket also fell to the bowler.

Runs conceded. Everything scored off the bat from the bowler's deliveries, plus wides and no balls, plus runs resulting from those illegal deliveries, plus overthrows arising from the ball. Byes, leg byes and penalty runs are excluded. This is why the bowlers' runs column never sums to the team total: the difference is exactly the byes, leg byes and penalties.

Wickets. Dismissals credited to the bowler, being bowled, caught, leg before wicket, stumped and hit wicket. A bowler is not credited for a run out, obstructing the field, hitting the ball twice or timed out, because the delivery did not cause them.

Economy rate. Runs conceded divided by overs bowled. Convert part-overs to sixths before calculating.

Bowling average. Runs conceded divided by wickets taken. The price of a wicket.

Bowling strike rate. Balls bowled divided by wickets taken. A strike rate of 48 means a wicket every forty-eight deliveries, or once every eight overs.

Five-for. Five or more wickets in a single innings, written 5/42 or "five for 42". Six, seven or more all count in this column. Also called a five-wicket haul.

Ten-for. Ten or more wickets in a match across both innings. Ten wickets in a single innings, meaning all ten, is a separate and far rarer event, and a page meaning that will say so. The ten-for column is empty for limited-overs cricket.

BBI and BBM. Best bowling in an innings and best bowling in a match. Both are written wickets first: 6/28. BBM appears only for multi-innings formats.

Hat-trick. Three wickets in three consecutive deliveries by the same bowler. The deliveries may span two overs, two innings, or in first-class cricket two days, as long as they are consecutive balls bowled by that bowler.

Dot-ball percentage, bowling. The proportion of a bowler's legal deliveries from which no run was scored. In white-ball cricket it is often a better guide to containment than economy, because economy can be wrecked by a single over while a high dot-ball rate reflects sustained pressure.

Spell. An unbroken sequence of overs bowled from the same end. Spell figures rarely appear in the summary table, and a bowler's day is frequently two or three quite different spells averaged into one line.

Wide. A delivery judged out of the batter's reach as they stood. One run is charged to the bowler and to extras, the ball does not count towards the over, and it is bowled again. White-ball cricket applies the wide line far more strictly than Test cricket, which is one reason economy rates are not comparable across formats.

No ball. An illegal delivery. The commonest cause is the bowler's front foot landing beyond the popping crease, but a delivery above waist height on the full, too many bouncers in an over, a fielding infringement and various other breaches also produce one. One run is charged to the bowler, the ball is bowled again, and in white-ball cricket a free hit usually follows, on which the batter can only be dismissed by the means that apply when a no ball has been bowled.

Team and innings terms

Run rate. Runs scored divided by overs faced, expressed as runs per over.

Required run rate. In a chase, runs still needed divided by overs still available. The most watched number in a run chase and the most over-read, because it takes no account of wickets in hand. A side needing seven an over with nine wickets standing and a side needing seven an over with three are in different matches.

Net run rate. A tournament tie-breaker: the run rate a side has scored at across all its matches minus the rate it has conceded. Two details trip people up. A side bowled out before using its full allocation has its rate calculated over the full allocation, not the overs actually faced, which makes being dismissed cheaply doubly damaging. And matches with no result are excluded entirely.

Partnership. The runs added while a particular pair was at the crease, along with the balls used. Includes extras conceded during the stand, which is why the two individual scores do not always sum to it.

Fall of wickets. The team score at each dismissal, usually with the batter's name and the over. It is the only part of a scorecard that preserves the shape of an innings.

Extras. Runs conceded by the fielding side and credited to no batter: byes, leg byes, wides, no balls and penalty runs. The extras line explains the entire gap between the sum of individual scores and the team total.

Byes and leg byes. Byes are runs taken when the ball passes both bat and keeper untouched. Leg byes are runs taken when the ball hits the batter's body rather than the bat, and are only allowed if the batter was attempting a stroke or trying to avoid the ball. Neither is charged to the bowler.

Declaration. A captain voluntarily closing the innings with wickets in hand, available only in multi-innings cricket and marked (dec). Every batter unbeaten at the moment of a declaration is recorded as not out.

Follow-on. In a match of two innings a side, a first-innings lead above a set margin lets the leading captain require the opposition to bat again immediately. The margin depends on the scheduled length of the match, and in a five-day Test it is 200 runs. Enforcing it saves time and commits the fielding side to bowling consecutively, which is why a captain may decline. A following-on innings is marked (f/o).

All out. An innings ended by dismissals. Ten wickets is the usual figure, but an innings can be all out with fewer if a batter is absent hurt or absent ill, because an innings also ends when nobody is left to come in.

Target. The score a chasing side must reach to win, which is one more than the score needed to tie. In a rain-affected match it may be revised.

Par score. In a rain-affected chase, the score the chasing side should have reached at the current point, given overs used and wickets lost, for the match to be even. Par accounts for wickets; the required run rate does not, which is why a side can be ahead of the rate and behind par at the same time.

Powerplay. A defined block of overs in limited-overs cricket during which fielding restrictions limit the number of fielders outside the inner circle. The specifics vary by format and competition, and powerplay figures are the most commonly separated phase split.

Result terms. A side batting second that reaches the target wins by wickets, counting the wickets left. A side batting first that defends wins by runs, counting the margin. A tie is a match with level scores and a completed second innings. A draw is a multi-innings match that ran out of time, and it is not the same as a tie. A no result is a limited-overs match abandoned before the minimum overs required for a decision.

Dismissal and fielding terms

The credited dismissals. Bowled, caught, leg before wicket, stumped and hit wicket, all credited to the bowler. On a card they appear as b Jones, c Smith b Jones, lbw b Jones, st Smith b Jones and hit wicket b Jones. In c Smith b Jones the catcher is always named first and the bowler second; the order never reverses.

The uncredited dismissals. Run out, obstructing the field, hit the ball twice and timed out, none credited to a bowler. A run out is written with the fielder or fielders responsible in brackets.

Retired hurt and retired out. Retired hurt, or retired not out, means the batter left injured or unwell and may return later in the innings. No wicket falls, and if the batter never returns the innings is treated as not out. Retired out, where a batter leaves for any other reason and does not return, does count as a dismissal, but no bowler is credited.

Catches and stumpings. Catches taken, and for keepers stumpings effected, usually summed into a dismissals column. Dropped catches are not recorded at all.

Analytical terms

These come from ball-by-ball data rather than from the scorebook. They vary between providers and are not always available.

Phase splits. Batting and bowling figures broken down by period of the innings: powerplay, middle overs and death in white-ball cricket, or by the age of the ball in Tests. A death bowler's career economy read against a powerplay bowler's compares two different jobs, and the phase split is what makes them comparable.

Average and strike rate by bowling type. A batter's figures separated by pace and spin, and often further by right-arm and left-arm, or wrist and finger spin. The most practically useful split in modern analysis, because a weakness against one type is the commonest way a batter is planned against.

Average by position. A batter's figures split by where they batted, which is the direct answer to one career average describing several different jobs.

Balls per boundary and balls per dismissal. Two frequency measures that are more stable over small samples than the ratios built from them. Balls per boundary describes scoring pattern; balls per dismissal is the batting equivalent of a bowling strike rate.

Expected and impact measures. Modelled figures that weight runs by the difficulty of the situation in which they were scored, or estimate how much a performance changed the probability of winning. They are the only common measures that account for context, which is precisely what the classic figures throw away. They are also model-dependent, and different providers will disagree about the same innings.

Three questions to ask of any cricket statistic

What is the denominator, and how big is it? Every ratio in cricket hides one. Dismissals sit under a batting average, wickets under a bowling average, balls faced under a strike rate, overs under an economy rate. A small denominator means the figure will move on the next innings, and a figure that moves on the next innings is not describing a player.

What did it throw away to become one number? A career average has discarded the format, the phase, the opposition, the conditions, the batting position and the era. Each of those was a real difference, and the single figure was produced by pretending it was not. Ask which of them was doing the work.

Who else is inside it? Byes carry the bowler's mistakes and the pitch's. A catch count carries the captain's field placings and the attack's ability to induce edges. A partnership carries a partner. A not-out carries a batting position. Very few cricket statistics belong to one player, and the ones treated as though they do are the ones most often misread.

Three habits follow from those questions. Put runs per innings beside any batting average. Put the phase split beside any white-ball strike rate or economy. Put the denominator beside everything.

Do that and most arguments about cricket statistics resolve themselves before they start, because nearly all of them turn out to be arguments about which question the number was answering. More of this sort of thing, on this sport and others, is collected in the cricket archive.