Late registration is the most argued-about free decision in tournament poker. One camp says arriving late is free equity, because the field has already shrunk and your buy-in buys a share of a concentrated prize pool. The other says you are skipping the softest tables of the day to buy a short stack. Both are right, and which one applies to you depends almost entirely on how good you are.
We built a multi-table tournament simulator, calibrated it against published field data, and ran 586,000 tournaments. Every one of them seeded sixty identical players who differed in exactly one respect: the level at which they registered. This is the short version of the results. The full study shows the model, the controls, the sensitivity analysis and two bugs we had to fix along the way.
The Short Answer
Register late if you are losing, register early if you are winning. That is the opposite of how the advice is usually framed, and it is the single clearest result in the data.
| If you are a... | Entering at level 1 | Entering at the close | Change | Verdict |
|---|---|---|---|---|
| Recreational player | -31.1% | -21.2% | +9.9 pts | Register late |
| Solid regular | +0.2% | +2.6% | +2.4 pts | Slight edge to late |
| Professional | +53.4% | +36.1% | -17.3 pts | Register early |
A recreational player gained +9.9 ROI points by buying in at the close of registration instead of at level 1. A professional lost 17.3 points. The weaker you are relative to the field, the more late registration is worth, because most of what it buys is a shorter tournament in which to make mistakes.
What Was Already Known
The existing numbers are consistent and all point the same way. GTO Wizard's analysis of real tournament samples puts the ICM boost at 9 to 12% in regular non-bounty events measured at the close of registration, with a satellite in the same sample worth +57% and a progressive knockout worth -14%. Dara O'Kearney, who wrote the book on ICM in tournaments, gets about 10% for a ten-player sit-and-go and around 22% for entering a 33-player event on day two.
A study of real results across 1,004 players and 8,834 entries in Triton events, using each player as their own control, found ROI improvements of +13.2 points for entries made with 60 to 70% of the field remaining. Its cleanest signal was cash rate: players who started at level 1 cashed about 15% of the time, late registrants upwards of 19%.
What none of that tells you is whether the boost survives contact with a real tournament, or who collects it. ICM has no concept of a big blind, no concept of skill, and no way to tell two players holding the same stack apart. That is the gap we went looking to fill.
Why the Boost Exists at All
Every player who busts before you sit down forfeits their prize equity, and that equity is redistributed to everyone still alive. Buy in afterwards and you pay the old price for a share of a pool that has already been concentrated. On a 500-runner tournament paying 15% of the field, ICM values a fresh starting stack at the close of registration at +7.0% over the buy-in.
We tested that in a field where every player was equally skilled, so nothing but the structure could move the number. The table paid +6.5%, with a standard error of 1.4 points. ICM called it.
The critical thing to understand is that this is not free money, it is a transfer. The equity you collect was surrendered by the players who busted before you arrived. In a winner-take-all tournament, where there is no payout ladder to transfer along, the boost is exactly zero at every stage. The flatter the ladder, the bigger the transfer.
How We Know
The simulator is a chip-transfer model rather than a hand-by-hand poker engine, which is the right level of abstraction for a question about entry timing. What it does model is the four things that actually drive the answer: chips are conserved, the blind clock escalates, skill converts to chips faster when stacks are deeper, and elimination is terminal against a non-linear payout ladder.
The free parameters were fitted to published field data before any entry-timing question was asked, then frozen.
| What we tied the model to | Target | What the model produced |
|---|---|---|
| Share of the field still alive when late registration closes | 42 to 48% (real samples) | 47% |
| Average surviving stack, in starting stacks | about 2x | 2.13x |
| Chip leader then, in starting stacks | 8 to 12x | 8.3x |
| Recreational ROI before fees | about -20% | -25.9% |
| Regular ROI before fees | about +4% | +8.6% |
| Professional ROI before fees | about +49% | +54.7% |
| Whole field, before fees (must be exactly zero) | 0.00% | -0.00% |
Each tournament seeds five identical copies of the same player at each of the twelve possible entry levels, sixty test players in a field of 500, all sharing one field and one set of opponents. The difference between two entry levels carries a standard error of roughly 2.0 points, computed across tournaments rather than across entries because clones inside one tournament are not independent observations.
What a Winning Player Gives Up
A strong player collects the same transfer as everyone else. They just pay more for it, and our sensitivity runs split the bill almost exactly in half.
- The early field is softer. The average skill of the players seated at level 1 is -0.506 on our scale against -0.211 at the close, because recreational players register early and then bust. Remove that gradient from the model and half the professional's penalty disappears.
- Deep stacks are where an edge compounds. Level 1 is 100 big blinds; the close of registration is 13. Remove the depth advantage from the model and the other half disappears.
Neither is decorative. The reason a good player should turn up at level 1 is both that the table is softer and that a hundred big blinds is where their edge actually lives.
The Hourly Rate Points the Other Way
Everything above is return per buy-in. Almost nobody's real constraint is buy-ins; it is hours. Our professional made $35.09 per hour of tournament time entering at level 1 and $61.19 entering at the close, because the second tournament takes 38 minutes instead of 99.
So a winning player registering late is not buying ROI. They are selling it: 17.3 points of return, in exchange for 1.7 times the hourly rate. Whether that is a good trade depends on whether you can actually fill the freed-up hours with more tournaments at the same stake. If your volume is capped by the schedule rather than by your clock, it is a mirage.
There is a trap here for the recreational player too. Their ROI improved, but their hourly loss got worse, from $-23.71 to $-40.66. Registering late does not make a losing player lose more slowly. It makes them lose less per tournament by buying less tournament, and the part they skip is the cheap part.
Format Changes the Answer More Than Anything
Hold the player and the field fixed and change only the payout ladder, and the value of late registration swings further than any other variable we tested.
The two levers are how many places get paid and how flat those payouts are. A structure paying a quarter of the field on a gentle ladder turned a -17.3% penalty into +5.8%. Winner-take-all, the control, came out at -26.9%: with one rung on the ladder there is no transfer to collect, so all that is left is the cost of a tougher field and a shorter stack.
Progressive knockouts move the other way. Half of every buy-in goes into a bounty pool that is being paid out from the first hand, so arriving late means arriving after a chunk of the prize money is gone. Bounties made up 43% of our level 1 entrant's return and only 37% of the late entrant's, and the format cost an extra 2.6 points on top of the standard penalty.
What To Actually Do
| Situation | What to do | Why |
|---|---|---|
| You are losing or breaking even in this game | Register as late as you can | Worth +9.9 points, mostly by shortening the game |
| You are a clear winner and short of buy-ins | Register at the start | You give up 17.3 points by arriving late |
| You are a clear winner and short of hours | Register late, and do the maths | You trade ROI for 1.7x the hourly rate |
| Satellite or any flat payout structure | Register late | Flat ladders make the transfer largest |
| Steep structure or winner-take-all | Decide on field quality alone | Little or no transfer to collect |
| Progressive knockout | Register early | The bounty pool is being spent while you are away |
| Turbo or hyper-turbo | Register at the start | No meaningful window, no stack depth left |
| There is an overlay | Register, whenever you can | The prize pool holds money nobody paid for |
Two things that override all of it. An overlay beats every consideration here: if a guarantee has not been met, the prize pool contains money nobody paid for, and you should take a share of it whenever you can. And in turbos and hyper-turbos, register at the start, because there is no meaningful window and no stack depth left once it closes.
Common Questions
How late is too late? It depends which player you are. On our numbers a recreational player never reaches a point where arriving later is worse in a non-bounty freezeout. A strong player is giving up ground from the first level onward. The usual coaching advice of 40bb or more is too conservative for a weak player and about right for a strong one, which is roughly the opposite of how it is normally presented.
Does late registration raise variance? Essentially not at all, which surprised us. The standard deviation of a single entry's result changed by -8% between entering at level 1 and entering at the close. Cashing more often and winning no more often cancel each other out.
Does it improve my chances of winning? No. Cash rate rises, from 12.6% to 15.9% for a recreational player, because the field you must outlast is smaller. Win rate does not follow it. Late registration buys min-cashes, and the top of the ladder is where the money is.
Should I late register a satellite? Satellites and flat structures are where the transfer is largest, so yes, more than in any standard tournament. Our satellite strategy guide covers why those structures bend every other decision too.
Does this apply to live tournaments? The direction does, the magnitudes will differ. Live events have longer levels, so registering at the close leaves you deeper than the 13 big blinds our online structure implies, which strengthens the case for arriving late.
The One Thing Worth Doing About It
Whatever time you sit down, at some point in every tournament you will be playing ten to twenty big blinds, and late registration simply moves that stretch to the beginning. It is also the most solvable part of tournament poker, which makes it the cheapest edge in this entire study to actually acquire. Getting the push-fold ranges right is worth more than getting the entry timing right, by our own numbers, several times over.
Our short stack strategy guide covers the framework, the ICM article explains why your calling ranges tighten when a covering stack is applying the pressure, and GTO Gecko drills the spots themselves.
If you want the model, the controls, the confidence intervals and the parts where we were wrong, the full study is here.