Running Below EV: How Long Can It Last?

Longer than any forum thread will tell you — and the honest answer has two halves. Computed: over 100,000 hands, ordinary luck runs a player 20+ buy-ins below true expectation about one time in four. Uncomfortable: the EV line only corrects all-in luck, so your real deviation from deserved results is bigger than the gap on your screen — and the two lines are under no obligation to converge on any sample you'll live through.

You're not here casually. You're here because there are two lines in your tracker, the green one and the orange one, and for weeks or months the green one — the money you actually have — has been sagging below the orange one — the money the software says you should have. The gap has become the first thing you check and the last thing you see. Somewhere in there, without deciding to, you started treating the orange line as the true you: the player you'd be if the deck were fair. And the question you're actually asking isn't statistical. It's how long can the graph keep lying about me?

Before answering "how long," we have to be precise about what that orange line is — because it is both less objective and less complete than nearly everyone staring at it believes.

What the EV Line Actually Measures

The all-in EV line does exactly one thing. When you get all-in before the board is complete and the hands are on their backs, the software freezes the moment, computes your equity, and credits you that share of the pot instead of the actual result. Get $400 in as a 49.6% favorite in an $403 pot and win: your results line records +$203, your EV line records your equity share minus your contribution — about minus 13 cents. That's it. That is the entire adjustment. Hands that never see an all-in pass through the EV line completely untouched, at face value, luck and all.

Two more things about the instrument, verified against how the trackers themselves describe it, that almost nobody staring at the gap knows. First: it isn't EV. Expected value is forward-looking; this stat is a results adjustment based on the actual hands shown, which is why PokerTracker formally renamed it "All-In Equity" — the vendor's own literature concedes the popular name overclaims. Second: it's biased. In multi-way pots the software would need the folded and mucked holecards to compute your true equity at the all-in moment, and it can't have them — so every tracker either counts those hands with incomplete information or skips them, and either choice skews the line. The direction documented in the trackers' own analysis is not comforting: the common handling systematically overestimates what you were owed, meaning users as a group tend to look slightly unluckier than they were. Some fraction of the gap you're grieving may be an accounting artifact. Not all of it. But you should know your oracle has a bookkeeping problem before you build a theology on it.

Most of Your Luck Never Touches That Line

Now the bigger omission — the one that inverts how you read your own graph.

Think about everything luck does to a poker result that never involves an open-handed all-in. It deals you kings the same orbit someone else holds aces, or doesn't. It decides whether the fish's flush gets there on the river he called off two streets on — no all-in, no adjustment. It decides whether your flopped set faces a one-street pot or a cooler stack-off, whether your bluffs meet the top or bottom of a caller's range, whether the deck pairs the board against your straight in the 60bb pots that never trigger the stat. Card-distribution luck, cooler luck, positional and texture luck, the luck of which hands went to showdown at all — every bit of it flows straight into your results line and is invisible to the EV line.

The all-in adjustment covers one narrow, showy slice of variance: the slice with the cards face up and the money already in. Everything else — plausibly most of what "running bad" actually is — happens off-screen. Which means the gap between your lines is a floor on your luck story, not the story. You can run brutally, historically bad while your EV line hugs your results line like nothing happened, because the beating came through coolers and card-death rather than flipped-up all-ins. The players who say "my EV line looks fine, so I must be playing terribly" are trusting an instrument that was never pointed at most of the sky. (If that's the trial you're running, it has its own essay: am I running bad or playing bad?)

So let's answer the real question with the real quantity: not the on-screen gap, but total luck — the distance between your results and your true expectation. That we can compute.

The Computed Answer: What an Ordinary Gap Looks Like

Model, stated plainly: results per 100 hands drawn from a normal distribution around your true win rate, standard deviation 100bb/100 — the defensible figure for 6-max no-limit (real values run ~90–100; at 90 these numbers shrink by about a tenth, e.g. the 100k-hand ≥20-buy-in figure drops from 26% to 24%). Limits: real results have fatter tails than normal, and the model assumes your win rate holds constant, which no human's does. Both simplifications make these numbers conservative — read them as the minimum weirdness you signed up for. The luck gap below is results minus true expectation; the deficits are in 100bb buy-ins.

Sample	SD of the luck gap	Typical gap (median, either direction)	10+ buy-ins below	20+ buy-ins below	30+ buy-ins below
25,000 hands	15.8 buy-ins	10.7 buy-ins	26%	10%	3%
50,000 hands	22.4 buy-ins	15.1 buy-ins	33%	19%	9%
100,000 hands	31.6 buy-ins	21.3 buy-ins	38%	26%	17%

Read the middle column first. At 50,000 hands — a serious quarter for an online regular — the typical player sits 15 buy-ins away from his true expectation, one direction or the other, as the median case. Not the unlucky player. The typical one. A full fifth of players are 20 or more buy-ins below deserved at that sample; roughly one in eleven is 30 below. At 100,000 hands, one in four is 20+ buy-ins under, and one in six is 30+ under — while playing, by construction, perfectly.

And those are endpoint numbers, which understate the lived experience, because you don't experience your sample at its endpoint — you experience the worst moment on the way. Simulating the full path: within any 100,000-hand stretch, the chance of at some point sitting 20+ buy-ins below your true pace is 52%; the chance of touching 30+ below is about one in three. The median player's worst dip below his own deserved line is 21 buy-ins. And here's the strange, load-bearing fact: those path numbers don't depend on your win rate at all. Luck is the same size for the crusher and the grinder; a better win rate climbs out of the hole faster, but it falls in just as often and just as deep. (What the hole does to your bankroll and your year — depth and duration by win rate — is computed separately in how long do poker downswings last?)

So: how long can it last? The dry answer is that "below expectation" is not a spell that breaks — at any moment, over any sample, roughly half of all flawless players are below it. What you mean by "it" is the conspicuous gap, and the table says a conspicuous gap is ordinary weather for six-figure samples. Tens of thousands of hands of visible, painful, screen-lit injustice is not the universe malfunctioning. It's the machine running to spec.

The Lines Are Not Required to Converge

One more piece of tracker folklore to retire: "the lines have to meet eventually." They don't — not in the way you're hoping, and the hope is doing damage.

What converges is the rate: your luck measured in bb/100 shrinks toward zero as the sample grows, like an error bar tightening. But the absolute gap — the buy-in distance between the lines, the number your eye actually reads on the graph — has a standard deviation that grows with the square root of hands played. Look at the table: 16 buy-ins of ordinary spread at 25k hands, 22 at 50k, 32 at 100k. More volume makes the on-screen gap bigger in expectation, not smaller, forever. The lines drift like two ships on the same heading, closer in bearing every year and further apart in miles. And the multi-way accounting bias adds its own permanent drift on top. If you are waiting for the day the green line climbs back up and touches the orange one — the day the ledger clears and you get paid what you're owed — you are waiting for an event the mathematics does not schedule. There is no repayment coming. Variance doesn't keep accounts. It only keeps dealing.

The Second God

Which brings us to what the EV line is really doing in your life at 2am, because it isn't statistics anymore.

The results graph used to be the thing that told you who you were — and when it turned red, it became unbearable, so you did something that felt like sophistication: you switched your faith to the EV line. The orange line became the true scoreboard, the one that sees through the noise, the receipt for the player you really are. And now you refresh the gap the way you used to refresh the results, feeling the distance between the lines as a debt — the money the universe owes you, the acknowledgment being withheld. I deserve the green line.

Look at the structure of that sentence. It's the same altar. You haven't stopped asking a graph to rule on your worth; you've just appealed to a higher court in the same broken courthouse — an instrument that, as above, sees one slice of your luck, carries a documented bookkeeping bias, and diverges from your results in absolute terms by design. The EV line can't acquit you any more than the results line could convict you, and the demand that it do so is the actual source of the pain you're calling variance. The gap doesn't hurt. The entitlement to the gap's closure hurts — the nightly audit of what you're owed, the suffering of a man checking a mailbox the mathematics never delivers to. Notice, too, the asymmetry of your faith: the months you ran above EV, you never once discounted your winnings as unearned. The oracle was only ever cross-examined when it said something you couldn't bear.

There's a longer argument here — that the verdict you want from the orange line does not exist at any sample you'll reach, and that the need for it, not the running-bad, is the thing eating you — and it's the spine of The Downswing, which takes the whole question apart floor by floor.

What to Do With the Gap

Practically, tonight, three moves.

Size your own gap honestly. Take your hands played and your buy-ins below expectation and place them on the actual distribution — The Verdict computes exactly this, with your numbers instead of the benchmarks above. Most below-EV stretches land between the 1-in-3 and 1-in-10 lines: painful, ordinary, and radically unlike how they feel from inside. If yours is genuinely extreme, that's worth knowing too — not as a conviction, but as a reason to audit the play rather than the luck.

Demote the orange line to what it is. A partial luck-correction on all-in pots, biased in known ways, blind to most variance. Useful as one coarse signal among several. Unfit to be a god. If checking the gap has become a compulsion, the fix isn't a better line — it's moving your self-evaluation to the one scoreboard that isn't noise: your decisions, graded cold, before the river's opinion arrives.

Stop waiting for repayment. There is no ledger, and nothing is owed. What the sample will actually deliver, slowly, in hands, is your edge expressing itself — never the acknowledgment, never the apology, never the meeting of two lines on a screen.

The graph can lie about what you deserve for a hundred thousand hands, in the strict sense that the gap between deserved and received can stay huge for longer than your patience or your bankroll. What it can't do — up or down, green or orange — is tell you who you are. Unhooking the second question from the first is a book-length job. The book is The Downswing. The whole thing is on the shelf.