Execution

From solver frequency to a number: the 0–100 cut table

Last updated: 9 November 2026

Randomizer Widget shows a whole number from 0 to 100, 101 outcomes. To map a frequency onto it, pick one convention before the session and act if the number is under your frequency. Read it before you decide. “Under 30” lands 29.70% of the time; on a 1–100 tool, “30 or under” is exactly 30%.

Which frequency you are executing: the node total or your hand’s mix

A solver node shows two kinds of number. The node total is the share of the whole range that takes each action, for example “bet 38%.” The hand mix is what one hand does: some hands bet every time, some never bet, and a few split. The total averages all those hands together, so it isn’t the number to randomize on any single hand. Randomize every hand at 38% and you hit 38% overall while betting the wrong hands.

Map the frequency for the hand in front of you. Players do this at two levels: per combo, from the solver’s own mix for that hand, or per hand class, with one simplified frequency from their study for a group of similar hands. The cut is the same either way. Where those frequencies come from in the first place is the subject of the GTO frequencies guide. Every frequency on this page is an illustration; the numbers you use come from your own study.

Rounding solver output before you take it to the table

Solver output runs to the decimal; your memory at the table doesn’t. Round each frequency you plan to use to a step you can recall, 5% or 10%: bet 63 / check 37 becomes 65 / 35 or 60 / 40. Do the rounding in study, write the rounded number in your notes, and use that number every time, so every decision in a session comes from one plan.

Very small mixes are a separate choice. A hand that bets 95% and checks 5% is close to a pure bet, and many players simplify it to one: always bet. Whether a small branch is worth keeping is a strategy question. Settle it in study, not at the table.

The step you can hit also depends on the source. A 0–100 number resolves any whole percentage; a single card’s suit only reaches 25% steps, and a clock’s second hand needs a percentage-to-seconds conversion in your head. The methods guide compares them side by side.

The cut table: act if the number is under your frequency

One rule covers every two-way split: act if the number is under your frequency. For a 30% action, 0–29 means act and 30–100 means take the other action. Randomizer Widget’s readout runs from 0 to 100 inclusive, 101 equally likely values, so a cut at f lands f/101 of the time. That is slightly under f, always by less than one percentage point.

Optional refinement. If you want the rate closer to your frequency, cut at round(1.01 × f) instead. Below 50 it gives the same cut as the main rule; above 50 it moves the cut up by one and keeps every rate within about half a point: 60 becomes under 61 (60.40%), 70 becomes under 71 (70.30%). At exactly 50, under 50 and under 51 are both half a point off; pick one. It is a refinement, not a second rule: if you use it, use it every time; otherwise the main rule stands.

Bands for three- and four-way splits

When a hand splits three or four ways, lay the actions end to end in a fixed order and cut at the running totals. Each band gets its share in values, and the last band also takes the top value, 100. For raise 25 / call 55 / fold 20 the running totals are 25 and 80:

A four-way split shows up when you plan a whole street out of position in one go. Say your study gives a hand bet 10 / check-raise 15 / check-call 45 / check-fold 30. The running totals are 10, 25 and 70:

The four numbers must be the hand’s shares of the whole street and add up to 100. A solver shows the check-raise, check-call and check-fold split at the node where you face a bet, as shares of your checks: multiply each by your check share first (check 90%, then raise 1 in 6 facing a bet, is 15% of the street), or read a fresh number at each decision instead.

A whole-street plan takes one read: note the number before your first action and use it for the rest of the street. If the opponent checks behind, the three check bands all simply mean you checked. The same idea with three parts is bet / check-call / check-fold. If your notes are per decision instead (bet or check first, then raise, call or fold facing a bet), read a fresh number at each decision and use the cut table and the three-way bands.

Low or high numbers for the aggressive action

Two conventions are common. Low: the aggressive action takes the low numbers (“under 30, raise”), which is the convention this page’s examples use. High: the aggressive action takes the high numbers. For 30% the cut becomes “over 70,” which is 71–100: the same 30 values and the same 29.70%. In general the high mirror of “under f” is “over 100 − f.” With bands, list the actions from the passive end instead; whichever band sits at the top takes 100.

Neither is better, because every value is equally likely. What matters is that you pick one before the session and keep it. A player who decides after seeing a 72 whether high or low applies this time has chosen the action, not randomized it.

Worked spots: three frequencies from study, mapped

The frequencies below are illustrations, not solver output for any real spot. Use the numbers from your own study.

River: a polarized bluff

Say your study has a missed draw on the river bluffing 33% of the time and checking the rest. Why a polarized range bluffs at a set ratio is covered in the GTO frequencies guide. The cut is under 33: a 21 means bet, a 58 means check. Over 101 outcomes it lands 32.67%.

Flop: a check-raise mix

Say your study has a hand that, facing a flop continuation bet after checking, raises 25% of the time and calls the rest. The cut is under 25: a 7 is a raise, a 40 is a call. It lands 24.75%.

Big blind: a three-way defense

Say your study has a hand in the big blind facing a button open: 3-bet 15, call 50, fold 35. The running totals are 15 and 65, so 0–14 is a 3-bet (14.85%), 15–64 is a call (49.50%) and 65–100 is a fold (35.64%). A 64 is a call; a 65 is a fold. Write the band limits in your notes so a boundary value is never a judgment call.

Why your results can’t show whether you followed the number

It is tempting to check yourself by counting: did I really bluff 30% of the time? Over the samples one spot gives you, the count is too noisy to show anything but a large drift. At 30%, one standard deviation is about 6.5 percentage points over 50 decisions, 4.6 over 100 and 2.0 over 500 (it is √(0.3 × 0.7 / n)). Over 100 decisions, a player who overrides the number now and then can still count 30%, and one who followed it every time can count 25%.

So the check is the process, not the tally. Read the number before you form an opinion, take the action it lands on, and never override it. That is the third step of the how-to guide, and it is most of what execution means.

Using the cut table on several tables at once

The cut table doesn’t change with the number of tables. What changes is that decisions on different tables can be live at the same moment, and each needs its own number: one shared reading would tie them together. Randomizer Widget’s Pro tier puts a readout on each detected table of a supported client, each drawing its own number; the multi-table page shows how it follows them.

Frequently asked