Turn Cards, Measured — Pokalab Lab Notes

The cell every turn taxonomy loses

Double-barrel frequency after a called flop c-bet, single-raised pots BTN vs BB IP

Overcard, flush doesn't completeSame card completes the flush

A queen on 8♠5♠2♥ is two different cards depending on its suit — 14 points apart. Traditional turn maps classify by rank relation first, so the Q♠ gets filed as “overcard” and the flush is lost. The same suppression shows up on non-overcards: a blank runs 49.9%, a flush-completing rag 39.7%.

In one minute

The flop’s rule flips on the turn. On flops, board structure beats the high card. On turns, the first question is the card itself: is it over the board?
One card, two identities. An overcard that also completes the flush barrels 14 points less than a clean one — and traditional maps file both under “overcard”.
Where the pair lands matters. Pairing the middle flop card cuts barreling to 30%. And on an already-paired flop, a turn that puts two pair on the board is the most bet-hostile card in poker: 27%.
Broadway overcards beat ace overcards (65% vs 60%) — the same “your opponent has aces too” effect our flop study found.
The card nobody could name. On T72, an 8 gives J9 and 96 a straight — every traditional map calls it a blank. Straight completions are 27% of all turns.

The experiment

We used the same method as our flop study, Board Textures, Measured: measure what regulars actually do in every possible situation, and let a decision tree draw class boundaries from raw card facts alone — no inherited assumptions, and no boundary survives unless it replicates on independent halves of the player pool. If you haven’t read the flop piece, the one result to know: on flops, board structure (the gap) beats the high card as the primary axis.

The turn spot: the aggressor c-bet the flop and got called, and now decides whether to fire again — 1.05 million double-barrel decisions across 3-bet and single-raised pots — plus 218,000 defender responses to the second barrel.

The axis flips: on the turn, the card is king

The tree’s first split on the flop was board structure. On the turn it is emphatically not: the first cut — worth three times the next one — is simply “is the turn card above the board?” Overcards unlock aggression (59–65% barrel); everything else lives 10–25 points lower. Structure still matters, but it enters as the second layer: what the card does to draws.

That makes poker sense too. The flop sets the terrain; by the turn, ranges have already reacted to it. What moves the needle now is the event — the one card that shifts who just improved.

The map's first decisions, as the tree drew them

Share of turns · typical barrel range (single-raised pots, IP)

Only after separating overcards from everything else does the tree ask what the card did — completed a flush, completed a straight, paired the board — and only then, on the overcard side, which rank it was.

The card nobody could name

Some turns change everything and earn no label at all. An 8 on T72: not an overcard, no pair, no flush card — every traditional map files it as a blank. It is not a blank. J9 and 96 just made a straight. Before the 8, no two cards in anybody’s hand could make a straight on T72; after it, several can.

That is precisely what a flush completing does: the third card of a suit lets a two-card holding make a flush. So this family gets the same name — Straight Completes — and the two axes finally mirror each other, with the rank relation as the child on both sides:

Flush Completes splits into Flush + Overcard (49.0%) and Flush + Undercard (39.7%).
Straight Completes splits three ways: Triple Broadway — any third broadway landing on two (53.0%) · Overcard + Straight (52.9%) · Undercard + Straight (44.2%, check-raised 10.7% — the highest of any turn class).

The family covers 27% of all turns, most of it hiding inside “blank” on traditional maps. And the children justify themselves in the defense columns: a broadway completion favors the raiser’s range, so defenders fold more and raise less; a low completion lands in the middle of the caller’s range, so they fold six points less than on a blank and raise more than on any other card.

Turn pairs: which card pairs is everything

“The board paired” is four different events. Pairing the top flop card (K on K72) leaves barreling at 41.8%; the bottom card, 39.4%; the middle card — where the caller’s peeling range lives — collapses it to 30.2%. And on a flop that was already paired, a turn that pairs a second rank puts two pair on the board: 26.7%, the most bet-hostile turn in poker, because every hand counterfeits into the same two pair. Its mirror sits right beside it: when the turn pairs the already-paired rank and puts trips on the board, barreling jumps back to 46.4% and the check-raise rate falls to 2.9%, the lowest of any turn class — nobody can credibly raise a board everyone shares.

What the defense hears

As on the flop, defender folding is nearly deaf to texture — most classes sit within a few points of each other. The sharper defense signal is again the check-raise, and it peaks on exactly the classes where barreling collapses:

Check-raise vs the second barrel, by turn class

BB defending vs BTN's turn barrel, single-raised pots · 218k decisions

OvercardsDraws completeTurn pairsBlanks

Straight completions and middle-card pairs draw the raises; the boards nobody can represent — trips on board, the fourth flush card — draw almost none.

The map

Seventeen turn classes, named like a poker player

Barrel% (single-raised BTN vs BB IP) · fold% and check-raise% (BB vs that barrel)

Class	Definition	Example	Share	Barrel	Fold	X/R
Overcard — clean: no flush, no straight arrives
Broadway Overcard	K, Q or J above the board	Q on 8♦5♠2♥	6.4%	65.4	53.6	7.7
Ace Overcard	an ace arrives	A on T74	3.1%	59.5	52.5	7.5
Mid Overcard	7–T, still above the board	9 on 764	1.9%	59.2	53.1	7.7
Blank
Blank	changes nothing at all	4 on K92	26.7%	49.9	50.3	6.4
Double Flush Draw
Double Flush Draw	a second flush draw appears	7♣ on K♠9♠2♣	4.9%	53.5	47.3	6.5
Flush Completes
Flush + Overcard	overcard that completes a flush	Q♠ on 8♠5♠2♥	2.4%	49.0	51.8	9.0
Flush + Undercard	card under the highest board card that completes a flush	7♠ on K♠9♠2♦	8.7%	39.7	45.9	9.3
Straight Completes
Triple Broadway	third broadway card lands (any of them)	J on KT2	6.3%	53.0	49.2	8.1
Overcard + Straight	overcard that completes a straight	8 on 652	3.2%	52.9	47.2	10.3
Undercard + Straight	card under the highest board card that completes a straight	8 on T72 (J9, 96)	17.9%	44.2	43.2	10.7
Turn Pairs
Top Card Pairs	pairs the highest flop card	K on K72	4.4%	41.8	45.2	7.0
Middle Card Pairs	pairs the middle flop card	7 on K72	4.2%	30.2	44.2	10.5
Bottom Card Pairs	pairs the lowest flop card	2 on K72	4.5%	39.4	46.4	7.6
Two Pair on Board	pairs a second rank — the board is now double-paired	7 on KK7	1.0%	26.7	43.5	7.6
Trips on Board	pairs the already-paired rank	K on KK7	0.7%	46.4	47.6	2.9
Four Card+
4-Straight	puts four to a straight on the board — one card completes	6 on 754	2.7%	38.2	48.1	9.8
4-Flush	fourth card of a monotone flop's suit	♥ on ♥♥♥	0.9%	33.2	49.8	3.2

Definitions are pure functions of the flop and the turn card, and every card gets exactly one class — the event beats the rank relation, and the rank relation becomes the child. Resolution order: 4-Flush, 4-Straight, the flush completions, the Straight Completes family, then clean overcards and pairs. “Broadway” means K/Q/J — the ace gets its own class. Triple Broadway catches every third broadway landing on two, whether or not it is also an overcard.

The scorecard

On the same taxonomy-agnostic grid of turn situations, a traditional turn map (overcard / pairs / flush hits / blank, with undercard and over-middle kept apart) explains 0.31 of class-explainable barrel variation — split-half validated at 0.25 — and almost none of the defense (fold 0.09, check-raise 0.04). The 17-class map above scores 0.51 on the barrel (split-half 0.40), doubles the fold signal (0.18) and more than quadruples the check-raise signal (0.17) — while keeping every name a player already knows.