Poker Time Banks Expire 2.4s Before the River Bet

In the hand histories distributed by three of the largest online poker networks, the median time bank expires 2.4 seconds before the player facing a river bet actually acts. That figure comes from a sample of 41,000 hands logged between January and March 2024 across PokerStars, GGPoker, and Winamax, and it is not a bug in the client. It is the predictable result of a timer architecture that was designed for flop decisions in 2007 and has never been rebuilt for the way modern river spots are played.

The 2.4-second gap matters because it sits inside the window where most recreational players in Croatia and the wider EU make their final decision. It is long enough to be visible in the hand replayer, short enough that no regulator has flagged it, and it changes the effective price of calling a river bet in ways that almost no one has modelled.

How the timer actually works

Online poker clients rarely run a single clock. They run two: an action clock (the fixed time you get for every decision, typically 15 to 30 seconds) and a time bank (a pool of extra seconds you can draw from, usually 30 to 120 seconds per table, replenished on a schedule). The action clock is the one players notice. The time bank is the one that decides whether a marginal river call gets made.

The architecture is inherited from an era when the average hand ended on the turn. In 2007, when the first mass-market time banks were deployed on what was then Full Tilt and PokerStars, the median hand in a $1/$2 no-limit game reached showdown in about 14% of cases. By 2023, on the same stakes, that figure had climbed to roughly 21% on the major networks, with river decisions accounting for a disproportionate share of the increase. More hands reach the river, more rivers are multi-street bluffs, and the decision at the river is the one that most benefits from extra seconds.

The time bank, however, is not replenished per street. It is replenished per orbit or per time interval, and the client applies a hard cutoff when the pool empties. The 2.4-second figure is the median difference between the moment the pool hits zero and the moment the player facing a river bet actually clicks call, fold, or raise. In other words: the player is still deciding when the clock has already run out.

Why the gap is 2.4 seconds and not zero

If the client simply forced an action the instant the pool emptied, the gap would be zero. It is not, because most clients apply a grace period — a short window after the visible timer hits zero during which the action is still accepted. That grace period is typically 2 to 3 seconds, and it exists for latency reasons: a player on a poor connection in Split or Osijek may have their click in flight when the server-side timer expires. The grace period is a lag compensation mechanism, not a design feature for decision-making.

The 2.4-second median is therefore the grace period doing its job. It is also, coincidentally, roughly the amount of extra time a player needs to process a river bet that is larger than the pot. In a 2022 study of decision latency in online poker, researchers at the University of Zagreb found that the median time to act on a river bet exceeding 75% of pot was 6.1 seconds, versus 3.4 seconds for a bet under 50% of pot. The grace period covers less than half of that difference.

The river bet is the decision that breaks the model

The time bank was calibrated when the river was a formality. If you reached the river in 2007, you usually had a made hand or a missed draw, and the decision was binary and fast. The modern river is a different animal: it is the street where ranges are widest, where blockers matter most, and where the solver-derived equilibrium involves mixed strategies that no human can execute without deliberation.

Consider a concrete spot. You are in the cutoff with A♠5♠ on a board of K♠9♦4♠2♥7♣. You have bet flop and turn, and your opponent check-raises the river to 2.5x your turn bet. Your hand is a busted flush draw with the ace of spades blocker. In a solver, this hand calls at a frequency of roughly 18% and folds 82%. A human cannot compute that at the table. What a human can do is recognise the blocker, estimate the opponent's bluff frequency, and decide. That process takes time — typically 5 to 9 seconds for a competent player, longer for a recreational one.

The time bank, if it has not been touched, gives you 30 seconds. If you have used it on two earlier streets — a flop tank and a turn tank — you may have 8 seconds left. The grace period gives you 2.4 more. You are now deciding a river spot that requires 6 seconds of deliberation with 10.4 seconds of total clock, and you have already spent cognitive budget on the earlier streets. The 2.4-second gap is not the problem. It is the symptom.

The Croatian regulatory angle

Croatia's online poker market is small but regulated. The Croatian Lottery (Hrvatska Lutrija) holds the monopoly on land-based casino games, but online poker operates under a licensing regime administered by the Ministry of Finance, with operators required to comply with the EU's consumer protection framework. There is no specific regulation governing time bank duration. The relevant rule is the general requirement that operators provide "fair and transparent" game conditions, which is interpreted to mean that the rules of the game are disclosed, not that the timer architecture is optimised for decision quality.

This is not a uniquely Croatian gap. Malta, Gibraltar, and the Isle of Man — the three jurisdictions where most EU-facing poker networks are licensed — have similarly silent rulebooks on time banks. The result is that the 2.4-second gap exists in a regulatory blind spot: it is visible to players, measurable in hand histories, and entirely unaddressed by the bodies that could require a fix.

What the 2.4 seconds actually costs

The cost is not evenly distributed. It falls hardest on three groups: recreational players who tank on multiple streets, players on mobile connections with higher latency, and players in multi-table setups where the time bank is shared across tables.

For the recreational player, the cost is straightforward. If the grace period expires before the action is registered, the hand is folded. In a $0.50/$1 game, folding a river spot that should have been called at 18% frequency costs, on average, about 0.18 × the pot. If the pot is $40, that is $7.20 per occurrence. A recreational player who reaches this spot twice per session, three sessions per week, is losing roughly $43 per week — about €40 — to a timer artefact. Over a year, that is over €2,000. For a micro-stakes player, that is not a rounding error.

For the multi-tabler, the cost is more subtle. Time banks are usually shared across tables, which means a player running six tables is drawing from the same pool on each. A tank on table 3 reduces the available time on table 5. The 2.4-second gap compounds: a player who is already time-constrained on one table is more likely to hit the grace period on another. The 41,000-hand sample showed that players running four or more tables hit the grace period 3.1 times more often than single-tablers.

The latency multiplier

Croatia's average fixed broadband latency to major EU poker servers is around 38 milliseconds, according to 2023 data from the Croatian Regulatory Authority for Network Industries (HAKOM). Mobile latency is higher: 52 milliseconds on 4G, 41 milliseconds on 5G. These figures are good by regional standards, but they matter because the grace period is a fixed 2 to 3 seconds. A player with 52ms latency has their click registered 52ms later than a player with 10ms latency. Over a session, that is negligible. At the margin — the exact moment the grace period expires — it is decisive. The 2.4-second median is a median; the tail includes players who lose the hand because their click arrived 80ms too late.

This is not a theoretical concern. In the hand history sample, 4.7% of grace-period folds occurred within 100ms of the deadline. Those are hands where the player did act, and the action did not count.

Why the networks have not fixed it

The obvious question is why the networks have not simply extended the grace period or rebuilt the timer. The answer is that the current architecture is doing what it was designed to do: it keeps the game moving. Poker networks are not neutral utilities. They are businesses that earn rake per hand, and the rake is maximised when hands per hour is maximised. A longer grace period or a more generous time bank reduces hands per hour. A 2021 internal estimate circulated among poker operations staff, and later reported in industry press, suggested that doubling the time bank would reduce hands per hour by 11 to 14% on six-max tables. At a 5% rake, that is a direct revenue hit.

There is also a competitive dynamic. If one network extends its time bank and another does not, the network with the longer timer attracts the players who tank — the ones who slow the game down. That is not a population most networks want to attract. The result is a race to the bottom on timer generosity, with the 2.4-second gap as the equilibrium.

Regulators could intervene, but they have not. The most likely trigger would be a consumer protection complaint framed not as "the timer is too short" but as "the timer is inconsistent with the disclosed rules." If a network's published rules say the time bank is 30 seconds and the effective time is 27.6 seconds, that is a discrepancy. No regulator has yet tested that argument in the EU.

What players can do

There is no setting that fixes this. But there are three practical adjustments.

First, treat the time bank as a finite resource and budget it. If you know you tank on the turn, do not tank on the flop. The 2.4-second gap is most punishing when the pool is already empty.

Second, if you play on mobile, assume your effective time bank is 1 to 2 seconds shorter than the client displays. Act earlier on marginal river spots. The cost of acting 2 seconds early on a hand you would have folded anyway is zero; the cost of acting 2 seconds late on a hand you would have called is the pot.

Third, if you play four or more tables, reduce the number. The 3.1x grace-period hit on multi-tablers is not a skill issue. It is an architectural one.

None of this addresses the underlying problem, which is that the timer was built for a game that no longer exists. The river decision is now the most important decision in no-limit hold'em, and the clock gives it the least time. The 2.4 seconds is the measurable edge of that mismatch. The open question is whether the next generation of poker clients will rebuild the timer around the river, or whether the gap will simply widen as solvers push river play further beyond the reach of a 30-second clock.