How long should a poker blind level be? The maths behind a tournament that ends on time

Every poker night I run starts with a promise I half make up. We'll be done around eleven. I post it in the group, people plan the evening around it — the lift home, the last bus, whoever's up early in the morning — and then I build the structure by opening the last one I used and changing as little as I can get away with. Twenty‑minute levels. Fine. Probably.

Then one of two things happens.

Either the cards go quick, someone runs the table, and we're shaking hands at half nine — with a couple of people who took the night off feeling a bit short‑changed. Or, more often, it's gone midnight, we're four‑handed, the blinds still aren't big enough to actually end it, and half the people who wanted to see a winner crowned left an hour ago because they had to.

The whole of that comes down to one number: how long each blind level lasts. Get it right and the tournament lands near the time you promised. Get it wrong and there's nothing you can do at 11:40pm except keep going. And the way most of us pick that number is to guess — which is half‑forced, because the night I announce the tournament I don't yet know whether a handful of people will turn up or three times that. The one number that should decide the structure isn't in yet.

This is the stuff I wish someone had laid out for me when I started: why the clock is the wrong thing to stare at, how to actually work out a level length, and the one rule of thumb that turns it from guesswork into something close to a calculation.

The clock isn't what ends a tournament

Here's the reframe that made it click for me. A tournament doesn't end because a certain amount of time has passed. It ends because the blinds have grown big enough, relative to the chips on the table, that stacks get short and someone wins.

So the thing that really sets the length of your night isn't the level length on its own — it's the relationship between two things: how many chips are in play, and how fast the blinds climb to swallow them. Level length is just the dial that controls the speed half of that.

Which means you can't sensibly pick a level length until you know how many chips you're playing for. And that's the bit almost everyone skips.

Working out the chips in play

Start with the easy part. Total chips on the felt is, at minimum:

players × starting stack

Thirty players with a 10,000 starting stack is 300,000 chips. Simple.

Except even this "easy" part is the first thing you can't know. At our club a night might be eight players or it might be thirty‑six, and the structure has to be set when I announce the tournament — before a single person has registered. Eight players at 10,000 is 80,000 chips; thirty‑six is 360,000. That's four and a half times as many chips to blind through on the very same structure — the difference between finishing at half nine and still playing at one in the morning. No single level length is right for both. So if you want the night to end at roughly the same time whatever the turnout, a fixed structure chosen up front is already wrong for most of the outcomes.

The part people underestimate — every single time — is everything that gets added after that. Rebuys, re‑entries and add‑ons don't nudge the chip pool; they can double or triple it. A 30‑runner freezeout and the same thirty runners in a rebuy event are two completely different tournaments, even with identical blinds, because the second one might put 500,000 chips in play instead of 300,000. More chips means more blinds to grind through, which means a longer night — unless you make the levels shorter to compensate.

This is why copying an old structure quietly fails: you carry the level length from a freezeout into a rebuy event, it runs an hour and a half long, and you never quite work out why.

So before anything else: estimate the chips. Starting stacks, plus a realistic guess at rebuys and re‑entries, plus add‑ons — assume most of the field takes one. That total is the thing the rest of the maths hangs off.

The finish line: the 50 big blind rule

Now you need a definition of "finished," and it can't be "one player has all the chips," because no structure is built to play all the way down to that on the clock — the last bit happens fast once stacks get shallow.

The rule of thumb I settled on is this: a tournament is effectively in its endgame once the entire chip pool is worth about 50 big blinds. At that point the remaining players are all short, it's shove‑or‑fold, and a winner turns up soon after. So the finish line isn't a time — it's a blind level. Specifically:

target big blind = total chips ÷ 50

Run the freezeout numbers: 300,000 chips ÷ 50 = a 6,000 big blind. So that tournament is essentially over when the blinds reach roughly 6,000. You look down your structure, find the level whose big blind is closest to 6,000, and that's your end level — the level you're realistically going to finish on.

Do the same for the rebuy version and the finish line moves. 500,000 chips ÷ 50 = a 10,000 big blind — a higher level, deeper into the structure. Same players, same buy‑in, but more chips to blind away, so the natural end point sits later. That single calculation is why the rebuy event needs different levels to land at the same time.

Putting it together: solving for level length

Once you know which level you'll finish on, the rest is arithmetic.

Count the levels from level one to your end level — say your 6,000 big blind lands at level 14. That's 14 levels to play. Now take the time you've actually got:

level length = available playing minutes ÷ number of levels

Available playing minutes is your start‑to‑finish window minus the breaks, because breaks don't blind anyone down — they just pass time. A 7pm‑to‑11pm night is 240 minutes; knock off two 15‑minute breaks and you've got 210 minutes of actual play. Across 14 levels that's exactly 15 minutes a level.

For the rebuy version, with the finish line out at level 17, the same 210 minutes divided across 17 levels gives you about 12 minutes a level. Shorter levels, same finish time — exactly as you'd expect now that you can see the chips behind it.

One guardrail worth keeping: floor the level length somewhere sane. If the maths spits out four‑minute levels, you haven't designed a tournament, you've designed a turbo by accident — better to accept a later finish, or cut a few levels out of the structure, than to play unreadably fast. As a rough floor I don't let levels drop below about eight minutes for a normal evening.

That's the whole method. Chips in play, divide by 50 for the finish line, find that level, divide your playing time across the levels up to it. It genuinely works — and the first time you do it deliberately instead of by feel, it's a little annoying how close it lands.

Why it's still hard — and impossible to keep right by hand

Here's the catch, and it's a big one: almost every input to that calculation is either a guess or a moving target.

You don't know how many people will turn up until they're through the door — and as we just saw, the gap between a light night and a busy one is enough to move the finish by hours. You don't know how many rebuys the night will generate until it's over. Late re‑entries keep topping the chip pool up through the first couple of levels, every one of them quietly pushing the finish line further out. None of the numbers the calculation needs are actually settled at the point you have to commit to a structure.

To stay accurate you'd have to redo the entire calculation every time the field changed — recount the chips, re‑find the end level, re‑divide the time, and adjust the clock mid‑tournament. In theory you could. In practice nobody is standing there at level five rebuilding their blind maths while also running the night. So the structure you started with is the structure you're stuck with, errors and all.

And breaks make it fiddlier still. They eat into playing time, so they have to be in the calculation, not bolted on after. You also want them in the right places — ideally lined up with the natural moments to colour up and remove small chips from play, and with a proper break at the end of the rebuy period so you can count the field and lock the prize pool. More things to get right by hand, on a night when your hands are already full.

How pkrclub handles it

This is the part of running a club I most wanted to stop doing in my head, so it's one of the first things I built properly.

In pkrclub you don't set a level length at all if you don't want to. You set the two things you actually care about — when the tournament starts and when you want it to end — and it solves for the level length that lands there. Which means the thing you couldn't possibly know when you announced the night — whether eight people or thirty‑six would come — is no longer something you have to know up front. You commit to a finish time, not a player count.

Under the hood it's doing exactly what's above: modelling the chips in play, applying the 50‑big‑blind finish line to find the end level, subtracting the breaks, and dividing the time across the levels. (If you'd rather lock the level length yourself, you can — and it'll tell you what time you'll finish instead. The maths runs both directions.)

The difference is what happens when reality moves. Because it re‑solves from the actual registered field, a quiet night of ten that swells to thirty after the latecomers arrive just shortens its own levels to still hit your end time — no mid‑tournament mental arithmetic. If your structure turns out too short for a deep, rebuy‑heavy night, it extends it for the calculation with sensibly rounded extra levels rather than just giving up, and flags that your structure is on the short side so you can add levels properly. And it builds the break schedule for you — spaced roughly every couple of hours, nudged to land on the colour‑up moments, with a dinner break on the long events and a dedicated break at the close of the rebuy period.

The chip estimate is the one part that's only as good as its assumptions, and it's where this gets genuinely interesting over time. Today it works from sensible baseline numbers for how many rebuys and re‑entries a field will generate. Where it's built to go is to learn from history instead — how events of this type and this structure have actually played out at your club, and eventually down to how individual players tend to rebuy — so the chip estimate, and therefore the whole structure, gets sharper the more nights you run. It doesn't do the per‑player part yet; that's the direction, not a promise about today.

The takeaway

The level length was never really meant to be a guess. It's a calculation with two awkward unknowns — how many chips end up in play, and how the field grows once people start arriving — and those unknowns are exactly the kind of thing a computer is happy to keep chasing in the background while you get on with running the night. The 50‑big‑blind rule gets you a good structure on paper. Keeping that structure honest once the door opens is the part worth handing off.

pkrclub is poker tournament software for clubs — a live clock, seating, payouts and accounting in one place, with blind structures that work out their own level lengths and breaks. It's in active development, with early access opening to clubs and players soon.