Am I Running Bad or Playing Bad?

Almost certainly both — and that's not a dodge, it's the mechanism. Nobody runs bad long enough to ask this question without playing worse for it; sustained losing degrades every human's game. But here's the fork that matters: only one of the two can be put on trial tonight. Your results can't be audited on any timescale you live on. Your decisions can — hand by hand, this evening, to a standard outside your mood.

Everything useful about this question flows from that asymmetry, and almost everything painful about it flows from ignoring it. So let's take the two halves in order — the trial you're currently running, why it can't conclude, and the one that can.

The Trial You're Running at 3am

Notice, first, what you're actually doing when you ask this question — because you're not asking it at noon, rested, with a database open. You're asking it in the dark, after the session, in the specific register that only speaks at that hour. And the register is a tell. It isn't grief. Grief has an object and an end. It's prosecution. Not "what a rough stretch" — what is wrong with me, said with a curl of the lip. A downswing doesn't make a player feel gentle sorrow. It makes him his own prosecutor.

And a prosecutor is not conducting a search. He already knows the verdict he wants; he's assembling the case for it. So you marshal evidence. The EV line's above the results line, so it's variance — that holds for a second, then the prosecution stands up: but maybe the line's flat because I've developed a leak I can't see. So you go looking for the leak at three in the morning, with no tool running and no notes open, just an exhausted memory rifling through hands. And you find candidates, because you can always find candidates — a call that looks loose in hindsight, a bluff into the wrong player. Each one is a fresh cut, and each cut is entered into evidence, and the court never adjourns because the court runs on your not-sleeping.

One more feature of this trial: it happens in a sealed room. You've been deflecting — up and down, decent month, you know how it is — keeping the true number from the group chat, the coach, the person in the next room. That's not incidental. Shame conceals; it withdraws you from contact at exactly the moment contact would ground you. Which means the mind conducting the trial has been carefully stripped of every witness who might object, and a sealed room is where a mind convicts itself unopposed.

"Am I running bad or playing bad" is the case name of that trial. Here's why it can never reach a verdict on the evidence it's using.

Your Memory Is a Corrupt Witness

The first instrument you consult is recall — the felt sense of the last three months. And recall, under sustained loss, is systematically falsified in both directions, which is worse than being falsified in one.

On the bad nights, the suckouts you took are vivid, rehearsed, filed in chronological order with timestamps. The times you got there — the rivers you spiked, the coolers that ran your way — are simply gone. Deleted. They don't enter the record, because nothing burns into memory like injustice and nothing evaporates like good fortune. So the run-bad feels impossible, cosmic, personal, when the arithmetic of it may be dead ordinary.

But the same corruption runs the other way on the play itself. Asked "did I play well this month," memory serves up your justifications, not your decisions — the story you told yourself as the chips went in, which was authored by the same tired, tilted narrator now being asked to grade it. The three-barrel you fired because you were card-dead and needed something to happen has been quietly refiled as "an aggressive line." The scared check-back that leaked value for the fourth time this week is remembered as "discipline in a swing." Your memory both exaggerates your bad luck and launders your bad play — which is precisely the configuration that keeps this question circling forever. The evidence for "running bad" is inflated; the evidence for "playing bad" is hidden. No wonder the trial runs all night and lands nowhere.

So strike the witness. What's left is two instruments that don't run on memory: the graph, and the hand histories. They testify to different things, and the whole trick is knowing which.

What the Graph Can and Cannot Testify To

The graph feels like the objective party in the room — numeric, dispassionate, apparently final. Here's what it can actually say under oath.

The standard deviation of 6-max no-limit runs around 90–100bb per 100 hands. A strong win rate is 5bb/100; a solid one is 2–3. The noise is twenty times the signal, and that ratio sets the resolution of the instrument. Using the standard model (normal results per 100 hands, SD 100bb/100 — assumptions stated, limits below), the 95% confidence interval on a measured win rate is about ±9bb/100 after 50,000 hands, and about ±6bb/100 after 100,000. Sit with the second one: after 100,000 hands — months of serious volume — a true 5bb/100 winner's interval still runs from roughly −1 to +11. It contains "small loser." The math cannot, at that sample, distinguish a genuine professional from a break-even player, in either direction.

Run the same model forward and it gets starker. A true 2.5bb/100 winner shows a net loss over 50,000 hands about 29% of the time — over 100,000 hands, about 21% of the time. Even the 5bb/100 crusher is underwater after 100,000 flawless hands about one time in seventeen. Right now, tonight, a meaningful fraction of genuinely winning players are staring at losing graphs and reading them as verdicts. Every one of them is wrong — and so is the run-good break-even player three tables over, reading his rising graph as proof. The instrument is broken in both directions, with the identical face.

So here is the graph's honest testimony, in full: over your last few months I can rule out almost nothing — not that you're running terribly, not that you're playing terribly, not both at once. That's the whole statement. The graph is made entirely of outcomes, and outcomes on a monthly timescale are nearly all noise. What it cannot testify to — cannot, not "won't yet" — is the quality of a single decision you made inside it. (How long the noise can hold your results under water is its own computed answer: how long do poker downswings last?)

Which sounds like despair until you notice what it frees you to do. If the results-court cannot reach a verdict on any timescale you can survive, you're excused from attending it. There's another court in the building, and it's open tonight.

The Fork That Works: Your Decisions Are Auditable Tonight

Two events happen every time you act at a table, and they only look like one. The first is the decision — the line you chose, given your cards, your position, your read, and the price. It's finished the instant the chips leave your hand, and it is entirely yours. The second is what the deck does next. That's not yours at all. The graph records only the second event. The audit examines only the first — and the first is the only one that predicts anything about tomorrow.

So run the audit, and run it like an audit, not like a feeling. Concretely:

Pull twenty pots from the last month — selected by size or by spot-type, never by how they felt. Feelings select for injustice; you'll pull twenty coolers and acquit yourself. Take the twenty biggest pots, or every 3-bet pot over 40 big blinds, mechanically.

Cover the outcome. Read each hand only to your final decision point and stop. Grade the choice — against a solver baseline, a study group, or your own cold, written reasoning — before you look at what the river did. The single question: was this the best available line given what I could actually know when I had to act? Not "did it win." Winning is the deck's opinion, and the deck was shuffled before it knew your name.

Write the grade down, both directions. "Ran bad, played well" and "ran good, played badly" are both real rows in the ledger, and the second one matters more, because a green session with three bad calls that happened to hit is the worse night — in the only ledger that predicts tomorrow.

Look for the four fingerprints. Sustained run-bad converts into real play-bad along four known routes: revenge (barrels fired to make something happen), desperation (the shot-take, the "get it back tonight" buy-in), scared money (thin value unbet, rivers checked back, "discipline" that's actually flinching), and over-correction (tearing open a healthy game mid-swing, hunting a leak that variance never left there). Audit specifically for these. They're what "both" looks like in the file.

Unseal the room. One honest peer, one real number, one set of hands reviewed by eyes that aren't yours. From inside the fog, a run-bad winner and a genuinely declined player feel identical — you are the one person structurally unable to tell them apart, and the people you've gone quiet on are the only external instrument that can. This isn't about sympathy. It's a second opinion on the evidence, from a witness the prosecutor doesn't own.

This audit can come back bad — that's what makes it real. It's not a machine for exoneration; it's the one instrument honest enough to convict you when the conviction is true, and to acquit you when it's false. Either finding is actionable, which is more than the graph will ever offer: a leak you can locate is a leak you can fix, and a clean game is a game you can keep playing without shame while the sample does its slow work.

And notice what the audit quietly does to the original question. "Am I running bad or playing bad" dissolves into "which of my last hundred decisions were wrong, and why" — an answerable question, graded tonight, no verdict pending. The trial you couldn't stop running gets adjourned not because you won it, but because the charge turns out to be unreviewable in that court. You stop the trial by changing courts.

The Results-Side Tiebreaker, and Its Honest Limits

There's still a legitimate results-side question, and it deserves a real instrument rather than a mood: is my actual stretch inside what variance hands to a player like me, or outside it? That's computable. The Verdict takes your win rate, your tracker's standard deviation, your hands played, and your buy-ins lost, and places your stretch on the actual distribution — a 1-in-4 run, a 1-in-20 run, or something the model itself calls strange.

Use it with its limits in plain view. It cannot tell you your true win rate — that's the whole problem, the input is itself an estimate wrapped in noise. What it can do is check consistency: whether the stretch you're suffering is ordinary given the player you believe you are. If the answer is "ordinary," the results-court has nothing for you, and every remaining question belongs to the decision audit. If the answer is "genuinely extreme," that's not a conviction either — extreme runs happen to someone every week — but it's a fair reason to weight the audit harder and get a second pair of eyes on your game. A tiebreaker, not a judge. (Related, if your tracker is the thing torturing you: running below EV — how long can it last?)

And if you've read this far with a heavier question underneath — if the real query wasn't "which is it" but "how much longer can I live like this" — that one deserves better than a strategy essay, and it has its own: should I quit poker?

The last word belongs to the fork. You will never fully separate the run from the play at the sample sizes a life allows; the math refuses, and it refuses everyone equally. What you can do — tonight, with your own files — is grade the one thing that was always yours. The card tells you what happened. It has never once told you whether you played it well. That was always yours to know, and there's an entire book about learning to stand there — the trial, the sealed room, the scoreboard that can't convict you and the one that can. It's called The Downswing. The whole thing is on the shelf.