A table with an empty seat at eight o'clock is not bad luck, it is a process failure. Almost every recurring game bleeds a seat or two a month to the same handful of causes: a soft deadline nobody enforced, an invite sent too late for people to plan around, and no agreed answer for what happens when someone simply does not show. None of that is random. It is fixable with a few decisions made before the next invite ever goes out.
Decide how many seats the table actually has and treat that number as real, not aspirational. A game that "usually" plays eight but sometimes crams in ten is a game where nobody actually knows they are in until they arrive and see a chair. A fixed seat count turns "are we full yet" from a guess into a fact anyone can check for themselves.
Send the invite far enough ahead that people can genuinely plan around it, and put a real deadline on the reply — not "let me know when you can," but a specific day and time after which the seat moves on to somebody else. A deadline nobody enforces is not a deadline, it is a suggestion, and a table run on suggestions fills late, unevenly, and always at the host's expense.
The fix for a seat that might free up later is not squeezing in an extra chair — it is a visible, ordered alternates list. Whoever replies after the seats are already full goes to the bottom of that list, in order, and everyone can see exactly where they stand on it. That is what turns a dropout from an awkward scramble of group-chat messages into one obvious next step: the next name on the list gets the seat, no debate required.
A short confirmation message the day before catches most of your no-shows before they ever happen — it is far easier to admit "actually I can't make it tonight" to a text message than to simply not turn up and leave a seat empty. It also gives the first alternate real notice, instead of a same-day scramble to find a replacement an hour before cards go in the air.
The worst possible time to invent a no-show policy is the night someone no-shows, with the table watching and an opinion already forming about whether the host is being fair to that player. Write the rules down in advance, before there is a specific person and a specific excuse attached to them, and share them with the whole group the same way you would share the stakes for the night.
Every game carries a seat count and an alternates list, so "are we full" and "who is next" stay visible answers instead of group-chat archaeology days later. When a seat frees up, OnTheButton promotes the first alternate automatically, in the same instant the seat opens, and emails them right away; everyone below them on the list re-ranks without anyone having to manage it by hand. A cancellation inside 24 hours and a no-show the host marks on the night both record a strike against that player, and whether other players can even see that history is a setting the host controls, off by default. Three active strikes at a venue suspends a player from that venue's games automatically, and the host can override a strike whenever the circumstances genuinely call for it.
Get the seat count, the deadline, and the policy right once, and filling the table stops being a monthly negotiation with the same three people. For the rest of what keeps a recurring game holding together across a full season, see how to run a poker league — or take a look at what a game set up this way looks like from the front.