◎The Round In Progress round state Open the partner account
paid
Affiliate disclosure. The partner link in the masthead and in the band beside the copy on this page is a sponsored link to a partner operator, and this site may be paid if you open an account through it, at no extra cost to you. It carries rel="sponsored noopener" and opens in a new tab. A desk about what happens to a round when the connection breaks should not leave its own funding unsaid: one link funds the site, no operator is named, rated or recommended anywhere on it, and it describes no real game, provider or account.
The Round In Progress / myths
Six beliefs about a dropped round

Six beliefs, checked

Interruption questions attract a small set of confident beliefs, most of them reasonable and most of them wrong in the same direction. This page checks six of them against the samples: what is true, what is partly true, and what the arithmetic rules out.

Desk spec
beliefs checked
6
false
3
partly true
2
true
1
rounds changed by a break
0
runtime javascript
none
The round-state railsample B / the four stages of a game round and what a break costs at each
stage 01
placed / running

The stake is committed and the round is running. A break here, before the round is decided, is the only break that can void the stake - and only if the rules have no default action.

stage 02
decision / interrupted

The decision window is open. A dropped connection or a timed-out decision is decided here: by a default action, by a void, or by the window expiring.

stage 03
decided / settled

The outcome is fixed on the server. A break now is a display problem only: the round settles as decided and the result is shown on reconnect.

stage 04
void / returned

The round is voided and the stake returns to the balance. This is a rule outcome, not a loss: the money is where it started.

Of 24,000 sampled rounds, 2,880 met an interruption (12.0%). On 72.9% of the dropped connections the round had already been decided at stage 03, so the only thing lost was the display; the risk sits at stages 01 and 02, and only when the rules have no default action.
Direct answer

The six beliefs split three false, two partly true and one true. The common error is treating an interruption as a result: a break is not a void, not a loss and not a restart, and it changes the round only where a decision window is open and a default action exists.

The six, in full

falseA dropped connection voids the round.

Most drops do not void anything. Of 1,536 drops, 72.9% landed after the outcome was fixed and settled as decided; a void needs an outcome that was never fixed, which is a minority case.

falseIf the screen freezes, the spin is lost.

If the outcome was already decided, the round is complete on the server and the result is shown on reconnect. In sample I a frozen client still settled for 2.40 and lost 0.00 to the drop.

falseReconnecting restarts the round from the beginning.

A reconnect asks the server for the round state; it does not open a new round. Of the 1,728 resumed rounds, the outcome matched the server on all of them and none was restarted.

partly trueA break always returns the stake.

It returns the stake only when the round is voided, which was 40.0% of interrupted rounds. The other 60.0% settled as placed, for better or worse - a break is not a refund.

partly trueA timeout means the round is lost.

A timeout applies a default action or voids. It resolved to a void with the stake returned on 43.3% of sampled timeouts; on the rest it applied the passive action, which can win or lose.

trueThe account history is the version to trust.

The round is written to the server before the client is told, so the record holds the outcome even when the screen does not. The history matched the server on every sampled interruption.

The one direction of the error

Five of the six beliefs overstate what a break can take. The reason is the same each time: they read the round as living on the screen, when it lives on the server. Once the round is placed in the record, the connection is a viewer of it, and a viewer that stops does not change what it was viewing.

The screen is a reader. The record is the text. A break can stop the reading; it cannot edit the text.

The one belief that is straightforwardly true - trust the history - is also the practical fix for the other five: whatever the screen did, the record says what happened.