How to Play a Small Pocket Pair When You Miss a Set

In the three solves below, a small pocket pair that misses its set is checked more often as more of the board ranks above it. With 55 on the button, after the big blind calls preflop and checks the flop, the solver checks 17.5% of the time on 9s 3d 2c, where one card is higher than the fives, 77.6% on Jh 8s 3d, where two are, and 97.6% on Ks Qd 9c, where all three are.

The small bet, a third of the pot, falls in the same order, and the bigger bet only gets real weight on 9s 3d 2c. The rest of this page shows those splits, what changes between the boards, and how far the results can be trusted.

The spot

Every number here comes from one setup: 6-max cash, 100 bb stacks, the button opens to 2.5 bb, and the big blind calls. The flop pot is 5.5 bb with 97.5 bb behind. The big blind checks, and the button acts.

On all three boards the big blind checks nearly its whole range first, so the button facing a check is the usual way this flop starts in these solves.

The button can bet 33% or 75% of the pot. The solver's screens and CSV files show those sizes as 2 bb and 4 bb, so this page uses those labels.

These are averages for the 55 hand class. None of these boards has a five on it, so all 6 combinations of pocket fives are possible, and the numbers cover all six together. All three boards are rainbow, with three different suits, so 55 has no flush draw and no backdoor flush draw on any of them. The averages do not describe a particular pair of fives, and this page has no data on how individual combinations differ.

What 55 does on each board

The 97 bb bet stays in the tree. Its 55 frequency is zero in the CSV on all three boards, which is not a claim that the action is impossible.

9s 3d 2c: one higher card, and mostly bets

Only the nine ranks above the fives; the three and the two are below them. This is also the only one of the three boards where 55 has a draw, and only a backdoor straight draw: with the 5, 3 and 2 it still needs a four plus an ace or a six, one on the turn and one on the river. In this solve 55 is checked only 17.5% of the time. It bets 2 bb 53.6% of the time and 4 bb 28.8%, so it bets most of the time, and the small bet is its most common action. Of the three boards, this is the only one where the 4 bb bet gets real weight.

Jh 8s 3d: two higher cards, and mostly checks

The jack and the eight rank above the fives, and only the three is below. 55 has no draw here: with a five and a three it would still need three more cards to make a straight. It checks 77.6% of the time, far more than on 9s 3d 2c. The 2 bb bet drops to 22.4%, and the 4 bb bet is almost never used, at 0.0%.

Ks Qd 9c: three higher cards, and almost always checks

Every card on this board ranks above the fives, and 55 has no draw. It checks 97.6% of the time. The 2 bb bet falls again, to 2.3%, and the 4 bb bet is almost never used, at 0.1%.

Does the whole button range do the same?

These solves don't show that it does. The button's whole range changes its overall check rate by only a few percentage points between these three boards, far less than 55 does, and not by enough for this page to rank the boards. The pattern above is about 55, not about everything the button holds.

That statement, and the earlier one about the big blind checking nearly its whole range, are calculated from the CSV files in this study, not read from the solver's screen. Each hand class's frequency is multiplied by the number of combinations still possible on that board, once without preflop weights and once with them. The percentages shown on the app's own overall display do not include preflop weights, so they are not the same measure.

Why you can't name one cause

From one board to the next, several things change at once: how many cards rank above the fives, how high the top card is, how connected the board is, and how each player's range connects with it. 55's backdoor straight draw on 9s 3d 2c is one more difference. So "more checks as more cards rank above the pair" is an observation from three solves, not proof that the number of higher cards is the reason.

Treat the three boards as representative examples, not as an experiment that isolates one variable. A new board is a new question: if you hold a small pair on a board that isn't listed here, this page does not give you its frequencies.

How to use this in your study

Use it to review hands after a session. It is study material, not assistance for decisions while you are playing.

1. Confirm the spot first. These numbers need the same positions, preflop action, stacks, ranges, and bet sizes.
2. Don't play every missed set the same way. On these three boards, 55 goes from betting most of the time to checking almost every time. If you play a small pair the same way whenever it misses its set, these solves are evidence that the habit is worth reviewing.
3. Look at your sizes, not just whether you bet. On 9s 3d 2c, 55 uses both sizes; on the other two boards the 4 bb bet almost disappears.
4. Compare a new board to these three, then solve it if it matters. The comparison gives you a question to investigate, not a frequency to copy.

If mixed frequencies are new to you, start with What Is GTO in Poker?. For the other side of this spot, when a pocket pair does flop a set, see How to Get Maximum Value When You Flop a Set. The glossary defines terms such as range and EV.

What this doesn't cover

These solves only cover the button's first decision on the flop, in position, after a single raised pot and a big blind check. They don't cover what to do after the big blind raises, the turn or the river, other small pairs, or playing a small pair out of position. They also don't cover 3-bet pots, other stack depths, or multiway pots. Each of those is a different model, so don't carry these frequencies over to them. Nor do they show how a small pocket pair that misses its set should be played in general, whether that means giving up or always betting: they only show how often 55 is checked in these three solves.

Inputs and results

Downloads

• 9s 3d 2c: node CSV (10,036 bytes) and setup record (5,644 bytes).
• Jh 8s 3d: node CSV (10,028 bytes) and setup record (5,644 bytes).
• Ks Qd 9c: node CSV (10,001 bytes) and setup record (5,644 bytes).
• Download MelaGTO Direct v1.1 (10).

Each CSV has 92 BTN hand classes for the node after the big blind checks, with frequencies as 0–1 fractions and action EVs in bb. It is not a complete tree or a per-combination result.

The setup records were not written by the app. Before these solves we read both ranges back again from Direct's Bulk edit fields and checked the tree settings by eye in its setup screens; they matched the setup used for the AK, flopped-set and overpair articles. Each record is the AK article's setup record with only the board changed, in the app's setup format. The thread count, print interval, and isomorphism switch are not shown in the interface, so they carry the source code's default values.

Convergence is unknown. The 0.5% target is a setting, not a result. The capture exposes neither a final iteration count nor measured exploitability, so we don't claim the target was reached or that these are complete equilibria. Observed completion took at most 193 seconds on 9s 3d 2c, 184 seconds on Jh 8s 3d, and 180 seconds on Ks Qd 9c. Those are upper bounds from when completion was first seen, not exact solve times.

Ranges used

These are the 6max_RC_100bb_200NL ranges, artifact version 2026.08.12-v1, with their mixed weights, as read back from Bulk edit. A class with no number has weight 1; unlisted classes have weight 0.

BB, out of position (90 classes)

ATs:0.949,A9s,A8s,A7s,A6s,A5s:0.03,A4s:0.366,A3s:0.917,A2s,KJs:0.699,KTs:0.521,K9s,K8s,K7s:0.829,K6s:0.801,K5s:0.978,K4s:0.967,K3s:0.675,K2s:0.896,AQo:0.238,KQo:0.639,QJs,QTs:0.68,Q9s:0.771,Q8s:0.885,Q7s:0.952,Q6s:0.922,Q5s:0.767,Q4s:0.903,Q3s:0.752,Q2s:0.673,AJo:0.765,KJo:0.584,QJo:0.75,JTs:0.567,J9s:0.169,J8s:0.922,J7s:0.414,J6s:0.702,J5s:0.498,J4s:0.605,J3s:0.577,J2s:0.259,ATo:0.833,KTo:0.643,QTo:0.709,JTo:0.758,T9s:0.116,T8s:0.599,T7s:0.79,T6s:0.479,T5s:0.275,A9o:0.844,K9o:0.711,Q9o:0.889,J9o:0.845,T9o:0.866,99:0.588,98s:0.742,97s:0.925,96s:0.9,A8o:0.935,K8o:0.341,T8o:0.301,98o:0.082,88:0.814,87s:0.693,86s:0.824,85s:0.808,A7o:0.723,87o:0.019,77:0.887,76s:0.63,75s:0.969,74s,A6o:0.375,66:0.902,65s:0.58,64s,63s:0.688,A5o:0.765,55:0.909,54s:0.504,53s,52s:0.592,A4o:0.575,44,43s,33,22

BTN, in position (92 classes)

AA,AKs,AQs,AJs,ATs,A9s,A8s,A7s,A6s,A5s,A4s,A3s,A2s,AKo,KK,KQs,KJs,KTs,K9s,K8s,K7s,K6s,K5s,K4s,K3s,K2s,AQo,KQo,QQ,QJs,QTs,Q9s,Q8s,Q7s,Q6s,Q5s,Q4s,Q3s,Q2s:0.629,AJo,KJo,QJo,JJ,JTs,J9s,J8s,J7s,J6s,J5s,J4s:0.974,ATo,KTo,QTo,JTo,TT,T9s,T8s,T7s,T6s,A9o,K9o,Q9o,J9o,T9o,99,98s,97s,96s,A8o,K8o:0.923,J8o:0.526,T8o:0.929,98o:0.32,88,87s,86s,A7o,K7o:0.247,77,76s,75s,A6o,66,65s,A5o,55,54s,A4o,44,A3o:0.84,33,22:0.692

To rerun a board in Direct, create a Custom spot with the settings above, paste both ranges with Bulk edit, choose CPU solving, and compare the node after the big blind checks with the downloaded CSV. Version, hardware, and stopping conditions can all affect an exact rerun.

Explore preflop ranges