Against a reg who bets a lot, the first instinct is to check-raise. A flop-only lock agrees. But an aggressive reg is not aggressive only on the flop. We locked two streets of their game in the solver and got a different answer: if the reg barrels the turn often, you do not need to check-raise the flop against them at all, and the money is made on the turn.
Why the pool is usually locked badly
- Only one flop gets locked. In popular solvers a lock is set by hand on a specific node of the tree. The turn has 49 cards, and each has its own node. To lock the reg's barrel in even one line, you have to set the lock 49 times, and there are several lines after the flop. Almost nobody does this: people lock the flop and let the solver play the turn for the reg. But a reg who applies pressure on two streets is a different opponent.
- Frequencies get locked without hand composition. "Bets 70%" and "bets 70% with these hands" are different locks. If the composition is not set, the solver decides on its own which of the opponent's hands go into the extra bets and picks the ones convenient for itself.
- Responses to raises get locked. If you lock how often the pool folds to a raise, the solver starts raising everything: the locked fold does not change however often you raise. A live reg who gets raised every hand quickly stops folding.
Our solver locks frequencies and hand composition by showdowns from the database on every turn card at once and in every line where there are enough hands. Against raises the opponent responds on their own, the way that suits them.
Example: a reg who pushes on the flop and the turn
SRP BTN vs BB, 100 bb, flop K♠7♥4♠. On the BTN — the aggressive third of NL200–2000 regs, 45.9 thousand BTN vs BB hands. The aggressive reg cbets 70% of the time, the solver 52%. After a 33% cbet and a call, the pool barrels the turn almost like the solver — 48% vs 44%. So we built a reg who barrels a lot ourselves: we raised the barrel to 65% and kept the hand composition as the pool's by showdowns.
BB's response to a 33% cbet:
| What is locked | Fold | Call | Check-raise | BB's gain |
|---|---|---|---|---|
| nothing (solver vs solver) | 35% | 53% | 12% | — |
| flop only | 35% | 46% | 20% | 0.4% of pot |
| flop and turn, 65% barrel | 37% | 63% | 0% | 6.6% of pot |
The gain is how much BB earns on top of playing the solver strategy against the same opponent.
A flop-only lock sees only the wide cbet, and the solver check-raises 1.7 times more often. The gain is 0.4% of pot — within the accuracy of the calculation. Once the turn is locked too, the flop check-raise disappears and the gain grows to 6.6% of pot. All the difference comes from the turn lock:
| BB vs the turn barrel after a 33% cbet | Fold | Call | Raise |
|---|---|---|---|
| 75% pot, solver vs solver | 46% | 48% | 6% |
| 75% pot, 65% barrel | 14% | 48% | 39% |
| 135% pot, solver vs solver | 59% | 38% | 3% |
| 135% pot, 65% barrel | 10% | 63% | 27% |
Why. By showdowns, the pool bets the turn with different hands than the solver. In the 75% bet it has more weak hands: 35% vs 29%, and fewer two pair and better: 19% vs 22%. Of its two pair and better, the pool puts 32% into the 135% overbet, the solver 48%. When such a barrel is fired 65% of the time, it holds even more weak hands.
Then the raise moves from the flop to the turn. The same hands that check-raised under the flop-only lock — sets 44 and 77, the open-ender 65s, gutshots 53s and 86s, top pair KQo — call the flop 100% of the time and raise the 75% turn bet. Against the 75% barrel KQo raises 53% of the time, the solver 6%; 86s raises 70% vs 12%. Middle and weak pairs and ace-high stop folding: against the overbet 88 and A4o fold 1% of the time, AJo 13%, while the solver folds them 82%, 77% and 76%.
The logic is simple. A flop check-raise often ends the hand right away, and the reg gets no chance to make a mistake on the turn. A call keeps their weak hands in the pot, and they will bet on their own later.
The same lock can say the opposite
Now let's lock the turn the way NL200–2000 pool actually plays it, in every line. The pool's barrel after a call is almost the solver's. But after a flop check-raise and a call, the pool gives up on the turn: against a BB bet of 75% pot it folds 45% of the time, the solver 34%. This is seen in 97 hands on similar flops and 385 on other boards.
Against such an opponent the flop check-raise grows to 32% (the solver's is 12%), KQo check-raises 91% of the time instead of 27%, and the gain is 3.6% of pot. The flop-only lock gives the same 20% in both cases and is wrong both times. What the reg does on the turn decides the answer on the flop.
The trap: a locked fold to a raise
In a similar calculation for the CO we first locked everything the pool does on the flop and the turn, including folds to raises. The pool folds to a turn raise more often than the solver: 57% vs 48%. The solver's answer is degenerate: flop check-raise 85%, raising the barrel up to 93%, gain 8.4% of pot.
Pretty, but wrong. A locked fold does not change however often you raise into it. It was enough to free one node — the CO's response to the flop check-raise — and the check-raise fell from 85% to 8%. The degenerate answer was held up by one locked fold.
What to take to the tables
- Against a reg who barrels the turn often, do not check-raise the flop. Call with sets and draws and raise the turn. Call with middle and weak pairs and ace-high and catch the barrel.
- Against a reg who gives up on the turn after a check-raise, the flop check-raise pays, and bet the turn after their call.
- Look at the opponent's turn, not only at their cbet: the same cbet frequency gives opposite answers.
- We recommend counting on no more than half the gain: opponents adjust, and the reg's type in a given hand is known only approximately.
Where the numbers come from
The solver is ours, the same one the database review runs on. The pool is NL200–2000 regs: players with 300+ decisions, the aggressive third is the top third by aggression. A reg's type is determined by one half of their hands, and the lock is built from the other. Frequencies and hand composition are taken from similar flops, shrunk toward the other boards; the barrel composition is one for all turn cards.
All BTN bets and BTN responses to bets on the flop and turn are locked, in every line, including after a check-raise. In the "65% barrel" case the barrel frequency after a 33% and a 75% cbet was raised to 70%, with the hand composition kept as the pool's. In the tree the barrel came out at 65%, and that is the number in the text. Against raises and on the river the BTN plays the best response to BB's strategy. If these nodes are frozen too, the check-raise against the 65% barrel is also 0%, but BB stops folding altogether: that is an answer to a fixed strategy, not to a live reg. The accuracy of each calculation is about 0.5% of pot. One flop is an example, not a rule.
The LeakPrice database review is built on the same logic: your game is compared with the solver, so is the pool of your stakes, and the spots where it pays to deviate from the solver are checked by calculation. Subscription and orders are in the account: leakprice.com/cabinet. The express review is free, in the Telegram bot: t.me/leakprice_test_bot.