If you’re a poker fan, you’re probably most likely to recognize Kenny Hallaert’s name as a player, specifically a player who made the November Nine in the 2016 World Series of Poker Main Event. But although he’s still active in that regard – he’s cashed three bracelet events and two daily Deepstacks at this summer’s WSOP – he’s spent the last few years building a career on the industry side of things as a tournament director for Unibet.
Hallaert tweeted on Monday about some research he’d conducted in that capacity, as to whether lengthy late registration periods are harmful to the game. In his opinion, in can be, and many tournaments these days, both online and off, allow late registration far too deep into their structures.
Why I think we should limit late registration and why having long late registration is bad for poker.
An analyse made by me: https://t.co/g5LXZaCeqD— Kenny Hallaert (@SpaceyFCB) June 23, 2019
The obvious upsides
Even those who, like Hallaert, are opposed to long late registration windows are likely to agree that some sort of late registration is a necessary evil. Except for shootouts, where late registration would run counter to the design of the event itself, virtually every tournament has a late registration period.
It’s good for players, in that no one wants to miss an event because they got stuck in traffic or there was a bigger line at the cage than they expected. It’s also good for tournament organizers, because an event running an overlay will attract more players in late registration due to the added value, which in turn mitigates the losses for the organizer.
A not-so-obvious downside
Hallaert, however, argues that late registration can be abused and that it benefits late entrants at the expense of the rest of the field, particularly when the window to do so is long. At some point, he says, the harmful effects of late registration outweigh its benefits.
On the surface, this claim might seem paradoxical. After all, if you start an event in Level 1, you’ll have 75-100 big blinds or more, depending on the event, while if you register very late, you might only have 30, 20 or even fewer. Meanwhile, pros are always demanding slower structures, and recreational players gravitate towards events with more starting chips. If everyone wants deeper stacks, why would anyone intentionally register late in order to play a shallower one?
The thing is, though, that when you buy in late, you’re not only shallower stacked: You’re also closer to the bubble. Some percentage of players will always bust out in the early levels. When players sit down for the first level, between one-eighth and one-tenth of the field will cash. When you late register, the percentage of players who will reach the money is much higher. Yes, you’ll be at a chip disadvantage against most of them, but tournament math says that you’re still more likely to reach the money when you have the same stack in a later level than in an earlier one.
Tournament math
If you’ve done any reading about tournament strategy, you’ve probably got at least a passing familiarity with the math in question, known as ICM, short for “Independent Chip Model.” In a nutshell, the principle behind ICM is that excluding skill edge, chips behave like tickets in a raffle where the prizes consist of every payout in the tournament, but where each ticket-holder can only win a single prize. You can’t finish both 2nd and 4th, so if you’ve won the 2nd place prize, your “tickets” are removed from consideration for all lower prizes. This produces the effect that chips diminish in value, so each chip in a shorter stack is worth more than each chip in a larger one. That effect becomes stronger the closer the tournament is to a pay jump, and also the bigger the variation in stack sizes is across the remaining field.
Thus, the same number of chips is worth more as a short stack close to the bubble than it is as a starting stack in Level 1, and players therefore get an equity boost by registering late. More importantly, that additional equity doesn’t come from nowhere, nor does it come from the organizer’s pockets, because the tournament fees collected per player are fixed. Rather, it logically must come from the other players in the tournament, and that’s the problem; late registration benefits the player doing it, and it benefits the organizer by increasing entries, but it does so at the expense of everyone who entered earlier, albeit in a small, near-invisible way.
A question of magnitude
Just how big of an effect is it? That’s the question Hallaert set out to answer, and he found that it can actually be a pretty big effect, perhaps bigger than you might have expected.
ICM calculators are readily available, but they’re mostly designed for final table calculations. The trouble with the ICM formula is that it’s easy enough for a computer to solve for small fields, but the computational complexity grows exponentially with the number of stacks involved. Hallaert therefore used code written by a contact of his, using “R,” a programming language expressly designed for doing statistical calculations.
The person in question asked Hallaert not to share his code, but explained that in a nutshell, it plays out the tournament randomly a large number of times and averages the results, a technique known as “Monte Carlo” in the field of game AI. Hallaert tested and confirmed that with a large enough number of trials, the results the script produces match those produced by the ICM formula very closely, so the results of the simulation should be good enough for all practical purposes in this discussion.
Bigger than the rake
Hallaert first used the script to analyze a $530 tournament on PokerStars with 165 entries. In that tournament, by the time late registration closed, the field was down to 52 players remaining, with 23 places paid. Three entries – two re-entries plus one first entry – all came in right at the deadline, so Hallaert looked at what sort of value those players were getting for their money.
With $30 per entry being taken by the site, the expected equity for a starting stack should be $500. And yet, Hallaert’s simulation found that entering at this stage, the late entrants were getting $581 in equity. Not only is that 16% more than the average buy-in is worth, it’s considerably more than the buy-in plus rake.
It also means that those three late entries alone cost each of the other 49 players an average of $4.96 in equity apiece. That’s maybe not a huge number in the context of a $530 event, but it’s not nothing, and it doesn’t include all the other late entries that would have come somewhat earlier. Players who registered for Level 1 and were still on their first bullet would have been bled for considerably more than that over the course of the event.
Hallaert also looked at a live event, this one a $5000 buy-in, with $4650 going to the prize pool. It had late entry available until the start of Day 2, which makes it convenient to analyze because many players registered between the end of Day 1 and the start of Day 2, all therefore starting simultaneously.
In that event, 170 of 499 players were remaining at the end of Day 1, and there were 19 of these overnight entries, bringing the total to 189 of 518 to start Day 2, and 78 places paid. Notice that in both this and the PokerStars event, the ratio of players remaining to seats paid was between 2-1 and 2.5-1 when late registration closed.
For the live event, the simulation gave the result that each of the late entries was receiving $5117 in equity for their $4650 contribution to the prize pool, a bonus of 10% which once again necessarily involves a corresponding reduction in the equities of the players already in.
The damage done
It’s considerably harder to assess the total damage to the equities of players who register early, because it depends on exactly how many late entrants there are, and just how late they register. It also depends on whether you’re calculating the loss at the time registration closes and applying it only to those still in, or whether you see it as a statistical loss of equity for early entrants based on the number of late entries you expect later on.
In extremely general terms, though, if about 10% of the field is entering considerably later than the other 90%, and benefitting by about 10% of a buy-in, then the earlier 90% is, on average, getting hit for an invisible loss of 1% of buy-in worth of equity. That guess could easily be off by a factor of two, but probably not much more than that. And considering the fuss that players can make when the rake is raised by 1%, that’s not an effect we should be discounting.
How much is too much?
Hallaert stops short of prescribing a solution in his analysis. Rather, he ends by musing what maximum benefit we should be willing to tolerate for late entrants, and proposes 5% as a starting point for the discussion, presumably because that’s a round figure which is about a third to a half as large as the values for the tournaments he looked at.
Personally, I think most tournaments could afford to go lower than 5% simply because ICM effects aren’t linear; they grow exponentially as the bubble approaches. To halve the effect, you shouldn’t have to halve the late registration period. Keep in mind that in the examples Hallaert considered, about 30% to 35% of the field remained when late registration closed. If it were instead 50%, I think that would be enough to reduce the effect substantially.
The live/online difference
I also think that the ideal solution changes considerably depending on whether you’re talking about live tournaments or online. The reason is twofold: that skill edges run larger in live events, and that busting out of a live event has a much greater opportunity cost than an online one. Both of these facts give live players a compensating incentive to register earlier, rather than take the direct equity bonus of registering at the last minute.
Online, events are starting up all the time, and professional players can play numerous events at a time, even on multiple sites at once. There’s almost no cost beyond the buy-in to busting out of one event, as it means being able to close that table and open a new one. ROI is therefore of lesser concern than hourly rate, and shortening one’s statistical play time by buying in late with a short-stack may actually be beneficial in that regard even disregarding the equity benefit.
Furthermore, playing online, there are far fewer of the very weakest players to target, and less opportunity to get inside opponents’ heads. The very best live tournament players generate a ton of their profits by reading a table and playing exploitatively, which is easier to do with lots of chips, while online play is more mathematical in nature, and shorter stacks make that easy.
The opportunity cost argument also applies to the intended upside of late registration, namely making sure that everyone who wants to play has a chance to do so. Missing the end of registration for a live event might leave a player without a good alternative, but that’s rarely the case online; at worst, it might mean the player has to wait 30 minutes or an hour for some other event in their price range to start up.
All-in-all, then, I see a big danger in players exploiting extended late registration periods in online events, and not much upside for the site or the average, naïve player in leaving them open for long. Live, even the most calculating of players may have reasons to prefer entering early, unless the equity advantage of late registration is extremely large, and there are plenty of good reasons – traffic, lines, etc. and the lack of alternative events – to have a generous late registration period. Thus, the current situation for live events might not even need fixing, but I think a dramatic reduction in late registration for online events is in order.
Will we see changes?
The next Unibet Open event will be held in Malta in mid-September, so it will be interesting to see whether Hallaert acts on his feelings and moves to reduce late registration. In the past, he’s shown a willingness to tinker with the formula to produce what he feels is a better event for all players. Under his watch, for instance, the Unibet Open was quick to follow the WPT’s lead in adopting a big blind ante last year.
Hallaert also lets player opinion guide him, however. For instance, although he’s generally opposed to re-entry, he introduced a single re-entry for Unibet Open Main Events last year because players said it’s what they wanted. Shortening late entry would be a move that would have an obvious downside for players who make frequent use of it, but the upsides are less obvious, so it might prove to be an unpopular decision.
Hallaert is right that long late entry periods hurt a majority of players who don’t take advantage, but how many of those players actually see the harm? Even when a change is for the greater good, it can be hard to make it if the benefits are widely distributed and hard to see, while the cost is more obvious and borne by a minority that’s likely to be vocal in their displeasure.
Alex Weldon (@benefactumgames) is a freelance writer, game designer and semipro poker player from Dartmouth, Nova Scotia, Canada.
