Why Am I Losing at Poker? 9 Causes, Ranked by Cost

You've checked the graph three times tonight, which is two more times than the graph has changed. It's the same shape it was on Sunday: a long flat stretch, then that stairwell in the middle of June, then flat again but lower. And the worst part isn't the money. It's that you genuinely don't know whether you're a bad player having a normal month or a decent player having a bad one, and nobody can tell you.

You're losing at poker for one of three reasons: variance you haven't outrun yet, a repeated decision that costs money every time the spot appears, or a stake where the rake is bigger than your edge. Most losing players pick the wrong one. They blame variance when their sample is far too small to blame anything, or they tear their strategy apart when the real problem is that they're playing NL5 on a site that takes more from the pot than they take from the table.

Short answer: If you're losing at poker, the cause is variance, a leak, or a stake you can't beat after rake, and you cannot tell which from your results until roughly 50,000 hands. Diagnose from your decisions instead: audit the three spots where you have no plan, and check what the rake costs you at your stake.

Key takeaways

• At a standard deviation of 100 bb/100, a 25,000-hand sample carries a 95% error bar of about plus or minus 12 bb/100, which is wider than the best achievable win rate at NL10.
• Rake is the cause nobody ranks. Players report paying around 7 bb/100 at NL10, and elite NL10 win rates before rakeback run about 8 bb/100.
• Variance is the most-blamed cause and the cheapest one. Over a large sample it costs zero. It just costs everything in the month you're living through.
• The diagnostic order is stake, then tilt, then technical leak. Fixing a technical leak while playing an unbeatable game changes nothing.
• Results answer "did I win." Decisions answer "am I good." Only one of those two is knowable at your sample size.

Is it variance, or am I actually bad at poker?

At fewer than 50,000 hands, your results cannot distinguish between the two, and no amount of staring at the graph will change that. This is arithmetic, not consolation. Variance is the statistical spread of poker outcomes around your true expected result, and it is enormous relative to any edge a human can hold.

GTO Wizard's article on variance and bankroll management puts it in plain figures. A player whose true win rate is 2.5 bb/100, after 100,000 hands, will land somewhere between minus 0.66 and plus 5.66 bb/100 about 70% of the time, and between minus 3.82 and plus 8.82 bb/100 about 95% of the time. That's after a hundred thousand hands. A winning player can genuinely show a losing graph at that sample and still be a winning player.

Here's the same math at the sample sizes people actually have. Standard error in bb/100 is the standard deviation divided by the square root of the number of 100-hand blocks, so at a standard deviation of 100 bb/100 (the figure implied by GTO Wizard's own worked example, and inside the 75 to 120 bb/100 range Primedope lists for 6-max no-limit hold'em) the 95% interval works out like this:

Now put that next to what a good win rate even looks like. BlackRain79's win rate benchmarks for small stakes, updated for 2026, put an elite NL10 result at 8 bb/100 across one to eight tables, and an elite NL50 result at 5 bb/100. At 25,000 hands your error bar is ±12.4 bb/100, which is wider than the entire win rate an elite NL10 regular achieves. Your graph is not hiding the answer from you. It doesn't have the answer.

So stop asking the graph. It's the poker equivalent of weighing yourself four times a day and drawing conclusions about your metabolism.

What are the 9 real causes of losing at poker?

The nine causes below are ordered by expected damage over a year of play, not by how loudly they hurt. That ordering is deliberately different from how most players rank them, and the biggest disagreement is at the bottom.

Variance sits at number nine because over any meaningful horizon it costs you exactly nothing, and it is still the first answer most players reach for. Phil Hellmuth has built a thirty-year career on that instinct and sixteen bracelets on the part of his game that isn't that instinct. The reason variance feels like the main character is that it's the only cause with a highlight reel.

How do I tell whether it's variance, a leak, or the stakes?

Run the causes in this order. Each step is answerable in under an hour, and the order matters because fixing a technical leak in an unbeatable game changes nothing.

1. Price the rake first. Find your rake paid per 100 hands in your tracker, or estimate it from your site's published rake table. Players report paying around 7 bb/100 at NL10, and rake at the micros regularly runs into the same range as a strong player's win rate. If your rake per 100 is close to or above what an elite player earns at your stake, the game itself is the problem and no study plan fixes it. Move sites, claim rakeback, or move up.
2. Check whether your worst sessions are clustered. Sort your sessions by result. If your five worst sessions are wildly worse than everything else and each one followed a specific hand you can still remember, that's tilt, and tilt is a scheduling problem before it's a psychology problem.
3. Check your sample against the table above. Under 50,000 hands, take variance off the suspect list entirely. Not because it isn't happening, but because it's unfalsifiable at that sample and it will absorb every hour of thought you give it.
4. Now, and only now, look for a leak. Write down the three spots where you genuinely don't know what you're supposed to do. Not the spots where you lose the most money. The spots where you have no plan. Mark every hand you play in those three spots for one week.
5. Count, then compare the pattern rather than the hands. Forty occurrences in a week is a leak. Two is an anecdote. You're looking for a direction, too tight or too loose, too passive or too aggressive, not a verdict on any individual hand. Our ranked list of 21 common poker leaks gives you the names and the tiers, and the plain-English guide to what GTO actually is covers what you're comparing against.
6. Fix one. The five-step framework for fixing poker leaks runs the measure, compare, drill, re-measure loop on the single leak you picked.

Why am I losing at poker even when I play well?

Rake. It's almost always rake, and it's the cause the entire "why you lose at poker" genre skips, because it's the one that doesn't flatter the reader with a character flaw to work on.

Rake is the percentage the house takes from each pot, capped at a maximum per hand. At micro stakes that percentage stays roughly constant while the pots get smaller, which means rake as a share of your win rate gets brutal as you go down. Players in public discussions report paying around 7 bb/100 at NL10. Set that against BlackRain79's elite NL10 figure of 8 bb/100 and the shape of the problem is obvious: the house is extracting something close to a world-class win rate from your table before anyone gets paid.

This is why "just study harder" is bad advice for a losing NL2 player. The edge required to beat a heavily raked micro game exceeds the edge required to beat a mid-stakes game, which is the least intuitive fact in low-stakes poker and the one that sends the most people home. If you're beating your opponents and losing to the operator, the fix is rakeback, a different site, or a higher stake, in that order.

The second answer is game selection, which is rake's quieter cousin. You can be a solid winner and still lose money by consistently sitting in the toughest available game. Red Chip Poker's breakdown of why players lose names insufficient edge over the field as one of the core causes, and it's the one that's fixed by standing up rather than by studying.

How long can a poker losing streak last?

Long enough to make a winning player quit. Primedope's variance calculator documents a simulation in which the player finished more than 25,000 big blinds up over 2.5 million hands, and still endured a downswing of nearly 10,000 big blinds between hand 1.2 million and hand 2 million. That's 100 buy-ins of drawdown, inside a career that was profitable by a wide margin.

Their downswing tables put a 1,000 big blind downswing at roughly a 32% likelihood for a typical winning profile, and estimate that around 16% of your playing time will be spent inside a downswing that lasts 50,000 hands or more. One session in six, roughly, belongs to a stretch that feels like your game has stopped working.

Two honest things about that. First, none of it is a reason to keep playing through a stretch you can't afford, financially or otherwise. A stop-loss and a night off cost you nothing you can measure, and if poker has stopped being something you choose to do, the correct move is to stop, not to grind through it. Second, and this is the part nobody wants: an unusually long downswing is also exactly what a small, real, negative win rate looks like from the inside. Variance and being a losing player produce the same graph in the short run. That is precisely why the diagnostic runs on decisions rather than on results.

How to actually drill this

Knowing you have a leak in one specific spot doesn't fix that spot. You already knew you fold too much in the big blind, probably for months, and you folded there again on Saturday anyway. Leaks are motor habits. They respond to repetition, not to insight, which is annoying if you were hoping the reading would be the work.

1. Pick one spot from step 4 of the diagnostic, with all its parameters named. Not "big blind defense." Big blind versus a button open, 100 big blinds effective, 6-max. The narrower the spot, the faster it fixes.
2. Run 20 reps a day, five days a week, for two weeks. Twenty reps takes about five minutes. Commute, queue, the ten minutes before you sit down to play. Two hundred reps total is roughly what it takes for a spot to stop being a decision.
3. You're done when you act without the pause. That's the observable change, and it's binary. The hand comes up live and your read is instant. If you're still tanking at rep 200, you're drilling a slightly different spot from the one that's costing you money. Narrow it and go again.
4. Then re-run the diagnostic from step 1. Same order, same questions. Rake first, always.

This is the whole reason LeakSeek exists: real spots, solver-backed feedback on the decision you just made, five minutes at a time, with a leak report that ranks your mistakes by EV loss rather than by how bad they felt. It's phone-only, and after the 7-day trial you get one free session a day unless you pay, so if you want to build your own sims you'll need something on the desktop too. Our roundup of poker training apps covers what those look like, and this worked example shows what an EV-ranked breakdown of a single hand actually says.

You're going to open the graph again tomorrow. It'll be the same shape, because graphs are lagging indicators of decisions you made weeks ago and there's nothing in yesterday's line that you can act on today. The number that moves first is the one nobody plots: how many spots you can play without thinking.