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

← All format guides

Tournament variance: why winning players have losing months

Updated Nov 13, 2026

Variance is the gap between what your game earns on average and what it pays you this month. In tournaments the average is built out of rare deep runs and long deserts between them — only about 15% of the field gets paid at all — so the same player, playing the same game the same way, can finish 500 events up 301 buy-ins or down 115. Variance is not bad luck. It is the shape of the payout curve, and it is the reason bankroll rules exist.

Find it live: the variance simulator & bankroll calculator · the bankroll management guide. 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

This guide is the why underneath our variance simulator. The simulator takes your games — up to four at once, each with its buy-in, field size, format (standard, PKO or mystery bounty) and your ROI for it — and plays that stretch out 10,000 times, so you can look at the whole distribution instead of judging yourself on your last fifty results. Everything it shows follows from one fact: in a tournament, most of your expected value sits in finishes that happen a few times a year. Below: the math that makes tournaments this violent, how to read the simulator's three numbers (with its real outputs, worked in full), three decisions every grinder actually faces — a Sunday shot, a 40-buy-in downswing, a move-down — worked end to end, and then the honest part: the two things no simulator can do. It cannot know your real ROI, and it cannot tell a downswing from a leak.

Worked example

A winning micro-stakes regular — $11 buy-ins, 500-runner fields, a genuine 15% ROI — plays 500 tournaments. Our simulator, run 20,000 times on exactly those inputs, has him finishing down about 29% of the time. His typical worst dip along the way is 85 buy-ins, and in one run in twenty the worst dip passes 163 buy-ins — $1,791 at his buy-in, more than the $1,100 roll he started with. He is a winning player the entire time. The full outputs are worked through below.

The math that makes tournaments different

Start with the shape of a single tournament. In a typical online MTT, about the top 15% of the field gets paid, so even a strong player cashes in only 15–20% of the events he enters, and most of those cashes are small. The profit lives in the rare deep runs — a top-three finish might come along in only 2–3% of entries. That is the whole problem in one sentence: your income arrives in a few large lumps, separated by long stretches where you pay full price and get nothing.

Run the desert out. Cashing 15% of the time means failing to cash 85% of the time, so ten straight cash-less tournaments happens about 20% of the time — 0.85 multiplied by itself ten times — and twenty straight happens about 4% of the time. An online regular plays hundreds of events a year, so those stretches are not omens. They are appointments.

Now put numbers on a single event. In our simulator's model, one tournament in a 500-runner field swings a winning player by roughly 5.8 buy-ins of standard deviation, while his edge — at a healthy 15% ROI — adds 0.15 buy-ins per event. The noise on every single result is nearly forty times the signal. Add the signal up over 500 events and it grows to about +75 buy-ins; add the noise up and it grows too, to a spread of about ±128 buy-ins. That is why the average outcome and the likely outcome are such different animals in this game, why the bankroll guide's rules are sized in hundreds of buy-ins, and why "am I even good?" is a question that takes thousands of events to answer — a point the downswing audit below makes painfully concrete.

One more consequence of the shape: the mean is a liar by omission. In the worked example coming next, the average run finishes ahead of the median run by about eleven buy-ins, because a small number of huge finishes drag the average up. When a player says "my ROI is 15%," he is quoting the mean. What he will actually live through, most of the time, is the median — and the left tail is where he will spend his worst months.

How to read the simulator's three numbers

The simulator reports three numbers. Here is each one, on one set of inputs, with its real outputs — 20,000 runs of: $11 buy-in, 500-runner fields, 15% ROI, 500 tournaments, a $1,100 roll. Run those inputs yourself and you will land close to everything quoted here — the percentages within about a point, the drawdown line wobbling more, because the page runs 10,000 simulations where this guide quotes a 20,000-run reference.

The average, for reference: the mean run finishes +77 buy-ins (+$843), the median run +66 buy-ins (+$726). Five percent of runs finish below −115 buy-ins (−$1,268); five percent finish above +301 buy-ins (+$3,313). Same player, same game, same 500 tournaments — a $4,581 spread between the unlucky twentieth and the lucky one.

1. Chance you're down — 29%. After 500 tournaments, nearly three runs in ten still show a loss. This is the number to read before you judge any stretch of your own results: if a genuine 15% player is down after 500 events a quarter of the time, then being down after 500 events is, by itself, almost no evidence about your game at all. It is the base rate, and most players have never seen it written down.

2. The downswing to plan for — about 163 buy-ins. Inside each run, track the worst peak-to-trough dip. The median run's worst dip is 85 buy-ins — that is the normal bad month, arriving on schedule for a winning player. The 95th percentile is 163 buy-ins ($1,791): one run in twenty dips at least that far. Note carefully what a drawdown is: it is measured from a peak, so a 163-buy-in dip does not by itself bust a 100-buy-in roll — the graph usually banks some profit before it falls. What it tells you is the size of the hole you should expect to stand in at some point, feeling exactly as bad as the depth suggests, while playing winning poker.

3. Chance of going broke — about 16%. This one counts only the paths whose cumulative graph touched −100 buy-ins — the starting roll, gone — at any point in the 500 events: roughly one run in six. Two honest caveats. First, the simulated player is stubborn: he never moves down, never takes a week off, never mixes in a cheaper game, so his ruin number is what the 70-buy-in move-down rule exists to cut. Second, 16% is not a rounding error. At 100 buy-ins, a winning player who changes nothing really can go broke inside 500 big-field events. The roll buys you time, not immunity.

Using it for real decisions

Numbers are only worth computing when they change a decision. Here are the three decisions this tool gets pointed at most, worked with its actual outputs.

1. The Sunday shot decision

The situation. An $11 regular with a $1,100 roll — properly bankrolled for his game, 100 buy-ins — keeps eyeing the $55 Sunday major. One tournament. How much damage can it do?

What the calculator says. At $55 his roll is 20 buy-ins. The bankroll rule for shots is 2% of the roll: his biggest sane shot is $22, and the $55 is 5% — two and a half times the line. The $55 game becomes a legitimate 2% shot at a roll of about $2,750, and a fully bankrolled level (100 buy-ins) at $5,500.

What the simulator adds. A single shot is one draw from a very wide distribution — the simulator prices the habit instead. Run "every Sunday for two years" — 100 events, 1,000-runner fields, the same 15% ROI, $1,100 roll — and the average run makes +$852, but the median run finishes down (−$350), the chance of being down after the 100 is 55%, and the $1,100 roll hits zero in 64% of runs. The 95th-percentile worst dip at that level is 64 buy-ins — $3,497, more than three times the roll he started with. Same player, same 100 events at his regular $11 game: the roll was threatened in zero of 20,000 runs. (Being down after just 100 events is a coin flip at both levels — 55% — short stretches are noise for everyone. The difference is not profit. It is survival.)

The decision the numbers support. A single $55 shot, budgeted and rare, is a $55 story and the bankroll guide allows it. A $55 habit on a $1,100 roll is not a shot — it is an under-rolled level with a 64% ruin rate. Satellite in for $5.50, or let the roll reach ~$2,750 and take the shot by the rule.

2. The downswing audit

The situation. A player is 40 buy-ins down over his last 300 events at $11 in 500-runner fields, and the question eating him is the only one that matters: am I unlucky, or am I leaking?

What the simulator says. Run his exact stretch — 300 events — under three different assumptions about who he is:

Where the tool's answer runs out — and this is the most important paragraph in this guide. All three worlds produce his exact graph with real probability, so the simulator cannot settle the question, and neither can anything else at this sample size. The arithmetic: one tournament carries about 5.8 buy-ins of standard deviation, so the error bar on an ROI measured over 300 events is roughly ±34 points — his measured −13% ROI is consistent with everything from a losing player to a strong winner. After 2,000 events the bar narrows to about ±13 points. It is still wide. This is why tracked samples are quoted in the thousands and why PrimeDope's own calculator page warns that at 500 tournaments the data cannot distinguish a strong winner from a losing player.

The decision the numbers support. The tool acquits the downswing — it is fully consistent with winning poker — and refuses to acquit the player. So: if the roll is sized for the game, keep playing the plan and stop grading yourself in 300-event windows; spend the worry on the inputs you control (game selection, study, reviewing the biggest pots), and let the tracked sample — thousands of events, filtered by buy-in and format — deliver the verdict when it is actually capable of one.

3. The move-down moment

The situation. The downswing from scenario 2 kept going. The roll that was $1,100 is now $550 — and he is still loading $11 tournaments, because $5.50 "isn't his game anymore."

What the calculator says. At $11 on $550 he holds 50 buy-ins, and his move-down line at the $11 level — 70 buy-ins — was $770. He crossed it weeks ago; by the rule he wrote when he was calm, the decision was already made. At $5.50 the same $550 is 100 buy-ins again: covered for standard MTTs, move-down line $385, move-back-up line $605, biggest sane shot $11.

What the simulator says. Next 500 events, 15% ROI, 500-runner fields, $550 roll. Staying at $11: the roll hits zero in about 44% of runs. Moving to $5.50: about 16%. And here is the detail that reframes the whole argument: the chance of a losing 500-event stretch is identical, 29%, at both levels — because in buy-in units it is the same graph. Moving down does not make you luckier and does not make the downswings rarer. It makes the same downswing cost $896 instead of $1,791, and turns a 44% chance of an empty roll into a 16% one.

The decision the numbers support. Pride frames moving down as quitting. The numbers frame staying up as paying a 28-point ruin premium for the same distribution. Move down, play the level the roll actually covers, and move back up at $605 — a line, not a feeling.

The ROI problem: the input that moves everything

Every output in this guide is hostage to one input: ROI, your long-run profit per dollar of buy-ins. Fifteen percent at an $11 buy-in means $1.65 of profit per tournament, averaged over more events than you have probably played. It is the most important number in the model — PrimeDope says the same about theirs, in almost the same words — and it is the one players are worst at knowing.

Self-estimates run high for structural reasons, not vanity. Memory keeps the Sunday score and files the 200 buy-ins before it under "that rough patch." Samples get chosen kindly: the good site, the good game type, the run since the big win. And small samples flatter in both directions — the ±34-point error bar from the audit above means a 300-event heater can print a 40% ROI on a break-even game with nobody lying about anything.

Sanity-check yours the only way that works: tracked results. Your full history on SharkScope or in your own log, thousands of events, filtered to the buy-in and format you are actually sizing — not your lifetime number across three sites and a different game. Then, in the simulator, run the preset below the one you believe. If the plan only survives at the ROI you hope for, it is not a plan. If it survives at the one below, it might be.

That is also why the presets are bands — break-even, winning micro reg, solid reg, strong — instead of a blank box demanding a number you cannot know. And one baseline to keep in your pocket: a player whose finishes are perfectly average returns the pool minus the fee, which in this model is −10% ROI. Many players who call themselves break-even are standing below that line and don't know it, because nobody ever showed them where zero actually is.

What the simulation can't tell you

A simulator is a model wearing your inputs. Ours assumes:

And the line under all of it: a simulation is a distribution of what could happen to a player with those inputs, not a prediction of what will happen to you. Your actual next 500 events are one draw from the distribution — including the parts of it you would rather not draw. The tool's job is to make sure that when the bad draw comes, you recognize it as a draw you were shown in advance, at a size your roll was built to survive.

Strategy tips

  1. Judge your game over thousands of events, not hundreds. After 300 tournaments your measured ROI carries an error bar of roughly ±34 points — wide enough to hide almost any truth.
  2. Plan for the 95th-percentile downswing, not the average one. The average run never tests anyone.
  3. Run the simulator at one ROI preset below what you believe. A plan that only works at your hoped-for ROI is not a plan.
  4. When the roll crosses your move-down line, the decision is already made. The simulator just prices ignoring it: about 44% ruin versus 16% over 500 events in the worked example above.
  5. A shot is a budgeted exception: 2% of the roll at most, a fixed number per month, no re-firing. Priced as a weekly habit on too small a roll, it is the most expensive line in this guide.
  6. Track everything. The simulator is only as honest as the ROI you feed it, and tracked results are the only ROI that counts.

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

  • Tournament Variance Calculator ↗ Free tool
    PrimeDope · The incumbent, and a good one: the variance tool tournament players have used for years, with real payout structures and mixed tournament types. Ours is built to run your numbers in the same breath as the bankroll rules — and its own page makes this guide's ROI point: at 500 tournaments the confidence interval is too wide to tell a strong winner from a losing player.

📚 Books & audiobooks

  • The Mathematics of Poker ↗
    Bill Chen & Jerrod Ankenman · 2006 · The rigorous version of this page: distributions, risk of ruin and why the long run is longer than you think, from the source the subject's literature builds on.
  • The Mental Game of Poker ↗ 🎧
    Jared Tendler & Barry Carter · 2011 · The math says the downswing is normal. Living through month three of it is the other problem, and this is still the best book on that one.

🔎 Tools

  • SharkScope ↗ (affiliate link)
    Your ROI input lives or dies here: pull your real ROI and ITM from tracked results instead of memory — filtered by buy-in and format, and over thousands of events, not hundreds. A 300-event sample cannot grade your game; a 5,000-event one starts to. How to use it

Find these tournaments

Run the variance simulator → All tournaments → Sunday majors →

FAQ

What is variance in tournament poker?

Variance is the gap between what your game earns on average and what it pays you this month. In tournaments the average is built out of rare deep runs and long deserts between them — only about 15% of the field gets paid at all — so the same player, playing the same game the same way, can finish 500 events up 301 buy-ins or down 115. Variance is not bad luck. It is the shape of the payout curve, and it is the reason bankroll rules exist.

How many tournaments before my results mean something?

More than feels fair. With swings around 5.8 buy-ins per event, the error bar on your ROI after 300 events is roughly ±34 points, and after 2,000 events it is still about ±13. Thousands of tracked events, filtered by buy-in and format, is where a win rate starts becoming a fact instead of a hope.

Is a 40-buy-in downswing normal?

For a winning player in mid-size fields, yes. In the simulator, a 15% ROI player finishes a 300-event stretch at least 40 buy-ins down about one time in five, and the 95th-percentile worst dip at that ROI is around 122 buy-ins. Normal is not the same as painless — and normal is not proof you are winning. That is the whole audit, above.

Why does the simulator show me going broke at 100 buy-ins?

Because its player never moves down, never mixes in a cheaper game and never takes a week off — he fires the same buy-in into the worst part of the graph. At 100 buy-ins, 15% ROI in 500-runner fields, that stubborn player still hits zero in about 16% of 500-event runs. Your move-down rule is the difference between the simulation and your actual risk.

Related guides

Bankroll management · ICM for micro-stakes players · Using a randomizer · Guaranteed (GTD) and overlay · Freezeout

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.