📬 One free email a week: the weekend's freerolls, ranked by real value
Menu

← All format guides

Using a randomizer: mixed strategy without the guesswork

Updated Nov 13, 2026

A randomizer executes a mixed strategy honestly. You set the frequency your plan calls for — bluff this combo a third of the time, say — roll once, and do what the number says, instead of letting mood, tilt and selective memory pick your frequencies for you mid-session.

Find it live: our free poker randomizer · free tools and training on the Study page. Free to play: open-entry freerolls today · no-deposit poker · all format guides.

By Roger Irwin (Rutland34), FindPokerTournaments — put together with help from other players and the training resources linked on this page.

📚 Keep learning

How it works

Four steps, and the order is the discipline. First, pick the decision before the hand: "if I reach the river with a missed draw, do I bluff?" One decision per roll. Second, set the frequency from your strategy — the percentage is yours, and the tool supplies none of it. Third, roll once. Fourth, obey: at or under your number you do it; over it, you don't. Our free randomizer answers in words — YES or NO — with the decision, the roll and the frequency restated, and keeps your last 20 rolls on screen as a self-audit. If the history doesn't look like the plan, the plan isn't what's being played. And YES/NO is only the simple case: in split mode, one roll picks between up to four named actions — bet big, bet small, check — which is how a lot of real mixes actually look.

Worked example

River bluff at 30%: you reach the river with a missed flush draw and no showdown value. The plan says bluff this combo 30% of the time. You roll a 17 — under 30, so you bet. Same spot two hours later, you roll a 64 — over, so you check and give up. Two identical combos, two different actions, exactly the mix you planned — and nobody at the table can put you on a pattern, because there isn't one beyond the frequency itself.

Why randomize at all

Not to be fancy. Three plain reasons:

None of this requires worshipping solver outputs. Your frequencies do not have to be GTO-perfect to be worth executing. A rough plan executed honestly beats a perfect plan executed by mood, every session.

Two more worked examples

3-bet bluff at 25%. Your plan: 3-bet bluff a suited connector from the small blind a quarter of the time, call the rest. You roll a 25. Still a YES. Rolls 1 through 25 are 25 of the 100 decisive outcomes, so a quarter of the time is exactly 1–25, boundary included. If the boundary feels wrong to you, that feeling is why you let the number decide.

Call-down at 50%. Your plan: call the river with bluff-catchers half the time against this opponent. You roll a 71. Over 50 — fold, even though the hand "feels" like a call and you'd love to see what he has. The roll you obey most reluctantly is usually the one saving you the most money.

One roll, three actions: splits

Binary mixes are the training wheels. Real strategies often split three ways — two sizings and a check, say — and a YES/NO tool can't execute that. The fix is the oldest one in the literature: assign each action a range of the 100 numbers, roll once, and take the action whose range the roll lands in.

The worked version most players learn from is Dara O'Kearney's. Writing for PokerStrategy on how he randomises split decisions, he maps one roll to three actions: bet big on 1–60, bet small on 61–90, check on 91–100 — a 60/30/10 split. Roll a 42 and you're betting big. Roll a 95 and you check, even if checking feels like surrendering the pot. The ranges do the deciding, which is the entire point: left to feel, most players bet big 80% of the time and call it a mix.

That is exactly what our randomizer's split mode does. Name up to four actions, give each its share, and the ranges are worked out in order — the shares must add up to exactly 100%, and the tool refuses to roll until they do, because a split that adds up to 90 is a plan with a hole in it. One roll, one named action, no favouritism and no renegotiating at the boundary: 60 is in the first range because 1–60 is sixty of the hundred outcomes, the same boundary rule as the 25-at-25% case above.

The math most tools get wrong

Plenty of randomizers roll 0 to 100. That dial has 101 outcomes, not 100. Set one to 30% and play "bluff on 30 or under" and you are actually bluffing 31 times out of 101 — about 30.7%, not 30%. One point sounds like nothing until you remember that the entire reason you are rolling is that the edge in these spots lives in one or two points.

Our randomizer handles it directly: a roll of exactly 0 is a dead roll. The tool discards it and rolls again, which leaves exactly 100 live outcomes — and at a 30% setting, rolls 1 through 30 are exactly 30% of them. No conversion, no "close enough." The tool's history shows dead rolls as they happen, because you should be able to watch the honesty working.

We tested the build the boring way: 100,000 rolls at a 30% setting came back YES 29.96% of the time. That is random noise around 30%, not a tool with opinions.

And don't take our word for it that the 101-outcome dial matters — take the market leader's. Randomizer Widget, the paid Windows incumbent, tells its own users in its FAQ that because the range has 101 outcomes, '"under 30" comes up 29.7% of the time rather than 30%' — and calls the gap 'negligible in practice.' Maybe, at some stakes. Our view is narrower: a tool whose entire job is a frequency should deal the frequency it was set, and the fix costs nothing — so we fixed it.

Profiles and stats: your rolls vs. your plan

A mix you run every session shouldn't be rebuilt every session. The tool lets you save a setup — the named decision plus its frequency, or a whole split — as a profile, and load it back in one tap. "River bluffs 30%," "Small-blind 3-bet split," whatever your recurring spots are: set them once, honestly, away from the table, and session-you just loads and obeys. Everything is stored on your own device; nothing is sent anywhere.

The profiles pair with the part no other randomizer we know of keeps: stats — your rolls vs. your plan. Every decision roll is counted against the setup it was made under, so over time you can read lines like "Bluff — 212 rolls, YES 29.7% vs 30% target," or, for a split, each action's realised share next to its range share. Judge them over hundreds of rolls, never over a session: 22% over 40 rolls is noise, 22% over 400 is information.

One honesty note, because it matters: the stats count the numbers the tool rolled. The tool can't see your tables and does not know whether you actually bet when it said bet. If the rolls match the plan but the actions didn't, the leak is between the tool and the table — and no software can close that one for you. What the stats do catch is the quieter failure: a plan you keep quietly re-setting mid-session, or a frequency you swear you play that your rolls say you don't.

Two workflow extras round it out. Ambient mode deals a fresh number every few seconds for players who keep a number live beside their tables and glance at it — clearly labelled, and ambient numbers never enter your history or stats, so a glance can never contaminate the record. And on Chrome or Edge, the pop-out floats the number and verdict in a small always-on-top window next to your tables, with a working ROLL button inside it.

Where it helps, and where it's a leak

The tool earns its keep in a narrow job: executing a mix you have already decided is right, mostly heads-up, against opponents who are paying attention. Outside that job it does damage:

Is it allowed where you play?

Rooms are starting to answer this in writing, and the first big answer went the standalone randomizer's way. In October 2026, GGPoker clarified its third-party software policy ahead of enforcement on December 1, 2026: HUDs, trackers and table managers that run alongside the client are banned, with detection and warnings from that date — but in GGPoker's own words, 'standalone RNG generators remain fully permitted,' classed with general window managers and productivity utilities rather than poker software (statement issued October 18, 2026; reported by PokerExclusive). That is exactly the class our randomizer sits in: it reads no cards, no tables and no game state — it deals a number you asked for, and nothing else.

Two honest limits. GGPoker's policy is GGPoker's — it says nothing about your room, and rooms differ: CoinPoker, for one, requires study tools to stay closed while its client is open. And policies change. So check your room's current terms before you roll beside its tables — but where a room has drawn the line in writing, the standalone randomizer is on the permitted side of it.

The live version: clocks and suits

You cannot tap a phone mid-hand at Canterbury or Running Aces, and you shouldn't be reaching for one at any live table. The live game runs on two older randomizers that have always been legal because they're already in the room: the clock and your cards.

The clock. Upswing's mixed-strategy teaching offers the clock as the standard fallback when you have no RNG handy, and the math is simple: a minute has 60 seconds, so your frequency is your window. Need 50%? If the second hand is in the first 30 seconds of the minute, do it. Need 25%? The first 15 seconds. Need 30%? The first 18. Decide the window before you glance, glance once, and obey. If you look twice, you're not randomizing — you're shopping.

Suits. Your own cards work at coarser steps. Decide on one card in advance — the bottom hole card is the classic. Red suit means do it: that's a clean 50%. One named suit, say spades: a clean 25%. Between those two there is no suit trick, which is the honest limitation of the method: live, you mostly get 50/25 granularity, and most live mixes that matter live near one of those anyway.

The suit-draw button on the tool page is the online version of the same trick — one tap, one suit, for the decisions you run at a 25% or 50% frequency.

The tool now draws the clock for you. Its live/clock mode shows a real-time second dial with your action windows already on it: set a 30% frequency and the first 18 seconds of the minute are coloured as "do it"; set a split and each action gets its own stretch of the dial, the way O'Kearney maps seconds ranges to three actions live. It generates nothing and records nothing — it's a clock face with your plan drawn on it, which is the one thing a wall clock can't do. Use it to learn the windows until they're automatic, or keep it open beside a home game.

And know the clock's documented flaw: it's random for the first decision and less so for the one right after. Decide two spots a few seconds apart and both glances land on nearly the same clock position, so your actions cluster — the second "random" choice was mostly made by the first. The defence is the same discipline as everything else on this page: re-decide the window each hand, glance only when the decision arrives, and never pre-read the clock while the action is still coming to you.

Strategy tips

  1. Set the frequency before the hand, never after you see the action — a roll taken after the bet comes in answers a different question than the one you planned.
  2. Name the decision out loud, or type it into the tool. A roll for "the river thing" invites renegotiation; a roll for "bluff the missed flush draw" doesn't.
  3. Let the history audit you: after a session, your realized frequencies should look like your plan. If your 30% bluffs show up nine times in ten, you weren't rolling — you were shopping.
  4. Keep frequencies round while you learn this: 25, 33, 50. Precision you can't remember is worse than a rough mix you can execute.
  5. Judge your stats over hundreds of rolls, not sessions. A 30% plan landing at 22% across 40 rolls is noise; across 400 rolls it is information — about your execution, not your luck.
  6. Obey reluctantly on purpose. The folds the roll forces when you wanted to call are the entire point — note them, review them later, but take them.

Common mistakes

🎯 Sharpen your game

Hand-picked places to get better at this format. Free stuff first. Community marks working players, coaches and smaller creators worth supporting. Links checked Nov 5, 2026.

📖 Free articles & videos

  • Mixed Strategy: Why Top Poker Players Make Decisions At Random ↗ Article
    Upswing Poker · The concept's source, in plain English: why strong players randomize, how to execute it with an RNG, and the clock fallback for live play. Upswing is the training site Doug Polk founded, and its teaching is where this habit went mainstream.
  • How to randomise split decisions at the poker table ↗ Article
    Dara O'Kearney · PokerStrategy · A working pro's method, and the source of the split in this guide: one roll mapped to three actions by ranges — bet big, bet small, check — plus his seconds-windows version for live play.
  • Free Poker RNG ↗ Free tool
    Upswing Poker · The original: a plain 1-to-100 roller with a pop-out window you park next to your tables. Ours adds the verdict, the named decision and the history — the right tool is whichever one you'll actually obey.
  • GTO Wizard: Theory (playlist) ↗ Video
    GTO Wizard · Why solvers mix in the first place — indifference, equity and frequency — explained in depth without needing a subscription.

Find these tournaments

The randomizer tool → All tournaments → Sunday majors →

FAQ

What is a randomizer in poker?

A randomizer executes a mixed strategy honestly. You set the frequency your plan calls for — bluff this combo a third of the time, say — roll once, and do what the number says, instead of letting mood, tilt and selective memory pick your frequencies for you mid-session.

Do I need a solver to use a randomizer?

No. A solver is one way to find frequencies; your own reads and population tendencies are another. The randomizer only executes the frequency — where the number comes from is your strategy's business, and a rough plan executed honestly beats a perfect plan executed by mood.

Why not just roll 0 to 100 and eyeball it?

Because 0 to 100 is 101 outcomes. "30 or under" is 31 of 101 — about 30.7%, not 30%. Our tool treats 0 as a dead roll and re-rolls, leaving exactly 100 decisive outcomes, so the frequency you set is the frequency you get.

Can I use a randomizer when I play live?

Not on a phone mid-hand — and you don't need one. Use the clock (your frequency is your window of seconds in the minute) or the classic suit trick: red means do it for 50%, one named suit for 25%. Decide the window or the card before you look, look once, and obey.

Related guides

ICM for micro-stakes players · Bankroll management · Freezeout · Deepstack

Play responsibly. Tournaments are entertainment, not income. Set a budget and time limit, and never chase losses. 18+/21+ depending on site and state. Help: 1-800-GAMBLER.