What Is Good Cribbage Pegging Strategy?

Good cribbage pegging strategy is a set of counting habits. Lead a card that no reply can make fifteen on, and count ahead so you never leave a count of 5 or 21. Keep cards that answer a ten-card, and set traps only when board position says attack. Pegging supplies about 22% of all points in 25,000 simulated games.

Pegging here means the play, the phase between the cut and the show, as most tables use the word, though the American Cribbage Congress uses it for recording any score with a peg. The rules of the play are on the go and 31 page and the term itself is in the glossary. This page is about choices: what to lead, what to answer with, and when to set a trap.

What is the best card to lead in cribbage?

The lead commonly taught is a 4, because no single card can make fifteen on a lead of 4 or lower. Bicycle's rules page names the 4 among the safest leads. Across the deck, an A to 4 lead can be scored on by three cards, its pair-mates. A 6 to 9 or a ten-card lead can be scored on by seven, and a 5 by nineteen.

Colvert's defensive rule is to lead below a 5, preferably from a pair, and Rasmussen's tip on 2-x-x-x makes the same three-against-seven count. Schell's glossary adds two rules of thumb: from one low card and three mid-cards, the low card is a safe lead when the four total 21 to 24 pips. The second covers a low lead from three low cards and one high. If the dealer answers with a pairable ten-card, pair it when your low cards total eight or less, and otherwise take the fifteen.

The table below is computed from the site's odds data set, src/_data/odds.json, by the Nobs engine. The corpus is 25,000 fixed-seed games of ONYX, the strongest of the three bots in the browser game, playing itself, clamped at 121. The method is on the odds page, and pegging is the mean for the whole deal after the lead. ONYX chose every lead, so each row describes the positions it leads that rank from, not a random lead.

Why should you never lead a 5, and when is it right anyway?

The 5 lead is the one to avoid, and the count from the pone's seat says why. After the deal you have seen six cards, so 46 are unseen. Holding one 5 and no ten-card, 19 of those 46 score for the dealer: the sixteen ten-cards for fifteen and the three other 5s for a pair. That is 41.3% of the unseen cards, against three for an A to 4 lead and seven for any other rank. Chambers makes the same division from 5-10-J-K: 3 of 46 for the other 5s, 13 of 46 for the ten-cards he does not hold.

Colvert's rule is the classic exception: with a 5 and one other card left, your lead, and the opponent showing no ten-card, lead the 5. You concede an occasional fifteen to escape the run that 3-4, 4-6 or 6-7 would make on the 5 later. Schell adds the 5 as prevent defense when one big peg would flip the position. He also leads the 5 from three different ten-cards in an endgame where exactly two points win, because a ten-card fifteen can then be paired nine times in thirteen. Chambers notes that a known 5-lead habit simply gets paired.

The Nobs bots price a 5 lead as a fixed hazard weighed against the play's points and the likely reply. The two rookie bots at the bottom of the ladder carry no lead-safety rating at all.

Which counts should you avoid leaving?

Two counts matter above all, 5 and 21, because from either every one of the sixteen ten-cards scores: for fifteen from 5 and for 31 from 21. From any other count at most four cards can reach 15 or 31 in one play. Bicycle's beginner advice not to aim for 21 is this arithmetic. So is the usual objection to a low lead that Rasmussen answers: a 9 on your 2 makes 11, and a ten-card from you then leaves 21.

Pairs follow a second count: a card cannot be paired once the count is at least 32 minus twice its value. That is the arithmetic behind Colvert's advice to dump a lone jack at 12 or higher.

Past 21 the danger flips to runs, and Colvert warns against being caught with a lone small card once the count is above 21. Play it early, usually as your second card, or send it to the crib. When the opponent needs five or more pegs, he adds, never hold a 4, 5 or 6 as your last card, above all the 5. A 4-5-6 run or three 5s ends games. The Nobs bots add a hazard to any play that leaves the count at exactly 5 or 21, a little over half the penalty they put on a 5 lead.

What is the magic eleven in cribbage?

A magic eleven is a set of cards whose pips total eleven, kept because two ten-cards make 20 and the eleven then makes 31 for two. Schell's glossary defines the whole family, and the two-card elevens are A with any ten-card, 2-9, 3-8, 4-7 and 5-6. Colvert's tip adds three-card elevens for desperate defense, 5-3-3, 9-A-A, 8-2-A, 7-2-2 and 6-3-2. He shows a dealer tossing J-K from A-2-6-9-J-K to keep the 9-2.

The seat rule is parity: answering a lead you play the second and fourth cards of a series, so two cards can finish it. Leading, you play the first, third and fifth, so a lead needs three. Rasmussen, who calls the doctrine Elevens from Heaven, gives the two-card elevens to the dealer and the three-card elevens to the non-dealer.

The elevens pay because sixteen of the 52 cards are ten-cards. Colvert notes that ten-cards with a 5 is the commonest holding, so with nobody leading the 5 the ten-cards come out first.

The Nobs bots keep the idea at discard time only, as a bonus for a four-card keep totalling exactly eleven that arrives from the second ladder tier upward. The app's own copy calls it holding cards that add to eleven when the deck is thick with tens.

What are losers and winners?

In Schell's glossary a loser is any card the opponent can play to score on the card you just laid. A winner is any card the opponent might play that lets you score with your next one. In Schell's worked endgame, the pone at 105 against a dealer at 119 held 8-9-10-10 after a 10 was cut. The 10 lead had four losers, the three 5s and the last 10, the 9 lead six and the 8 lead seven, so he led the 10.

Winners matter when you attack, and Colvert's lesson is a hand of 4-5-10-J needing two points. The safe 4 lead had six winners, the three 10s and three jacks, against ten for the 10 lead, and he led the 4 and lost. His rule for offense is never to lead the card that cuts your winners: needing three with 8-9-10-J, you lead the 8 or the jack, never the 9 or 10.

The percentage play, a phrase Colvert, Rasmussen, Chambers and Schell all use, means the card with the fewest losers or the most winners rather than a hunch. The Nobs bots compute exactly that: for each card they could play, they average the best immediate score the opponent could take with every card he might still hold. They weigh that expected loss against the play's own points and the hazards above.

What are Colvert's 5-card traps?

Colvert ranks the 5 as the easiest card to trap, ahead of the jack, the ace, the 2 and the 3. His 5-card traps all steer the count to a spot where only the 5 fits.

The simplest needs 6-6-4 in the non-dealer's hand. Lead a 6: the dealer cannot answer with a 5, since your 4 would take a run and fifteen together, so he plays a ten-card for 16. Your second 6 makes 22, a ten-card no longer fits, the 5 comes for 27, and your 4 makes 31 with a 4-5-6 run: five points to nothing.

The dealer's version keeps two cards that run with a 5, meaning 3-4, 4-6 or 6-7, as the last two. Holding 3-6-7-8 against a king lead, play the 8 for 18. A pone hoping for the go plays a ten-card for 28 rather than dumping the 5, and your 3 takes 31 for two. He leads his last ten-card, your 7 makes 17, the trapped 5 must come, and your 6 scores the run and the go. A shrewd pone escapes by dumping the 5 at 18, so it works best against two 5s.

Colvert's warning covers every trap: never set one while defending, since even a seven-point profit is worthless if it pegs the opponent into winning range.

What are sleeper cards and the rattle trap?

A sleeper card, in Colvert's first tip, is the one card that does not fit an otherwise combined hand, such as the 7 in 7-J-Q-K. Lead it, and the opponent cannot read what remains. With 10-10-J-Q, play the jack rather than a 10 and then the queen, which exposes the jack-shaped gap.

The jack trap, which Rasmussen calls the rattle trap, keeps a pair of jacks as your last two cards. From J-J-Q-K Colvert leads the king, because most players answer a ten-card lead with a 5 rather than pair it and risk a six-peg. A jack lead against 5-J draws the 5 and wastes the trap.

A lone jack is the opposite case: dump it at 12 or higher, where it cannot be paired. Schell's frequencies add a caveat: the jack is the pone's second most-kept card, but the dealer holds jacks, queens and kings about equally and duplicates the 10 most often. Against a dealer, a pair of 10s is the better trap.

Colvert's low-card traps run the count high with a small pair. Holding two aces, take the count to 29 with one of them, hoping the opponent drops a lone ace for 30. Your last ace makes 31 with three of a kind, eight points against his two. With two 2s run to 27 first, with two 3s to 25.

Should you pair your opponent's lead?

Barlow's answer is to count the worst case in your own hand before you pair a lead. The feared cost is a net four, their six for three of a kind less your two, but it is not always four. Holding 10-9-8-5 against a 9 lead, pairing risks 27 for six plus a go, a net five lost. Holding 9-4-4-2 against the same lead, a third 9 is answered by your 4 for 31 and two, so the most you can lose is two. The leader has to be holding the third 9 half the time just to come out level. Pair where your hand caps the loss and the gain is steady.

Barlow's second tip warns that an opponent who notices you always lead from a pair stops pairing, so keep the pair back now and then. The Nobs bots play the pair trap as a bonus. When a bot leads a rank it still holds, it adds the six points of the three-of-a-kind reply it hopes for, scaled by how likely you are to hold a copy of that rank. Absent from the Rookie and Club tiers, the term first appears with Ray in the Card Shark tier and reaches full strength with Iris and Judy at the top of the Master tier.

How do you read your opponent's cards?

Barlow's rule is that what your opponent holds is guesswork, but what he does not hold is nearly certain. Holding 5-9-10-J, you lead the 10 and he answers with a 7 for 17: play the 9 to 26, because a 5 for 31 would already have taken the fifteen. Colvert's logic starts from the lead: a defensive queen lead rules out A-4 and 2-3.

A go is the strongest evidence, because playing is compulsory: a go at 22 proves only ten-cards remain, at 25 nothing below a 7, at 27 nothing below a 5. Your own two discards cannot be in his hand either. In Schell's card-frequency table the pone holds a 5 in 38.5 to 39.0% of hands after the discard and a king in only 12.3 to 14.7%. The dealer's rates are flat, 21 to 29% for every rank. Suits can lie, too: Barlow plays three cards of one suit first, so an opponent who fears a flush plays the card the flush could not pair.

The bots in our cribbage app Nobs see only their own hand, the played cards and the starter. With the memory skill they strike their own discards from the cards you could hold. With the inference skill a go by you proves a floor under your remaining cards, and lower cards leave the set they reason over.

How many points does pegging really earn?

The corpus behind this page is 25,000 fixed-seed games of ONYX, the strongest of the three bots in the browser game, playing itself. In it the dealer pegged 746,521 points over 224,426 hands, a mean of 3.33 per hand, and the pone pegged 488,661 over the same hands, a mean of 2.18. Both roles divide by every hand played, cut-short final hands included; the odds page explains the method. These are bot-versus-bot figures, not a claim about human play.

Pegging is 21.9% of the dealer's 15.17 points per hand and 21.8% of the pone's 9.98, about 22% of all points scored. The dealer's 0.5% of blank hands is a rule effect: the pone must lead, so the dealer gets a go or the last card unless the game ends first.

Can you win by pegging? Yes, and often: 4,978 of the 25,000 games, 19.9%, ended on a pegging score. The rest ended on a counted hand, 61.1%, on the crib, 18.4%, or on his heels, 0.5%. Colvert's claim is that more games are won or lost pegging the last few points than by all the earlier play, and that most average games are decided by six points or fewer.

How does board position change your pegging?

Board position decides whether the advice above is offense or defense. In Schell's vocabulary, playing on means offense first, to keep a positional surplus or recover a deficit, and playing off means defense first. Prevent defense plays off to avoid a big score while conceding small ones, and cautious offense is the balanced default. Desperation offense and desperation defense are last-deal modes that ignore the other side entirely.

The same hand leads differently by position, as Chambers shows with 2-6-7-8 and a 4 starter. At 12 against a dealer at 10, he leads the 6, which keeps the choice of playing on into a run or off with the 2. At 119 against 119 he leads the 2, because only a pair can score on it.

His checklist: count the hand, fix the hole you want and how far it is, and note the opponent's position. Then work out the minimum you must peg, and only then choose on or off. Traps belong to playing on, and the opening 5 lead to prevent defense.

The Nobs bots that read the board weight the reply risk one and a half times when the board says defend, and half when it says attack. The points and hazard terms stay as they are. The board-position page owns the par holes and the streets.

What changes in the endgame?

On the last deal the defaults flip, and the endgame page owns that ground. Colvert's defender leads below a 5, preferably from a pair, dumps a lone jack and counts losers. His attacker keeps pairs and pieces of 4-5-6 for last and counts winners, holding a small pair to trap past 21 when six or more are needed. The endgame page takes the last deal from there.

By the Nobs Cribbage team Last reviewed 2026-09-03