Setting Walk-Away Rules

A walk-away rule tells the bots when to leave the table: up a certain amount, down a certain amount, or after giving back part of a big win. Every rule needs two things from you: the dollar figure, and what happens to the bets still in play when it triggers.

Examples

The words to know

There are three walk-away rules. You can use any of them, all of them, or none.

• Win goal. Walk away when the bots are up this much from where they started.
• Stop loss. Walk away when the bots are down this much from where they started.
• Giveback from the peak. A trailing stop. It activates once profit reaches one figure, then triggers when profit drops by a set amount from its high point.

Always say which direction you mean. "Set a walk away limit of $400" could be a win goal or a stop loss, and the AI has to guess. "Walk away up $400" and "walk away down $400" can't be misread.

A win goal is profit, not a bankroll total. On a $500 buy-in, "walk away up $300" means leave with $800. If you think in totals, say it that way: "walk away when the bankroll reaches $800."

A system with no walk-away rules is normal

If you don't ask for a win goal, the bots play the full session. If you don't ask for a stop loss, the bots ride out the full session. The AI won't invent one for you, and the ticket will say so.

The bankroll is the only hard wall

A bot that brings $300 can never lose more than $300. The table rejects any bet the rack can't cover. So "I don't want to lose more than $300" can mean two different things:

• Bring $300. That's the buy-in. It is a hard limit.
• Bring more, but leave when down $300. That's a stop loss. It is a soft limit.

Say which one you mean.

A hard limit of $300:

A $300 stop loss on a bigger buy-in:

When you set a buy-in, you can also say what happens when the bots can't afford a bet:

• Skip unaffordable bets and keep playing (default). If the bots can't afford an intended bet, they skip it and continue placing the system's other bets. The session keeps going.
• End the session the moment a bet fails. "Play until the money runs out" means this.
• Wait for active bets to resolve before ending. The bots resume play if their bankroll climbs back above a recovery amount you name. Useful when the bots have bets in play that could still win.

Play until the money runs out:

Wait for the bets in play, with a recovery amount:

When the rule triggers: three timings

A walk-away rule rarely triggers at a clean moment. There are usually bets still in play. You choose what happens to them.

Finish the current shooter (craps only, and the craps default). If the trigger fires mid-hand, the bots keep playing through the rest of the current shooter, then leave.

Let active bets resolve (the roulette and baccarat default). The bots stop placing new bets and let the ones already in play resolve, then leave.

Stop immediately. The instant the trigger fires, removable bets are pulled down and contract bets are forfeited. The bots leave. Uncommon.

The same system with each timing:

If you don't name a timing, you get your game's default, and the ticket lists the other options so you can switch with one reply.

A stop loss is a soft limit

With "finish the current shooter" or "let active bets resolve", the actual loss can run past the threshold while play winds down.

It can also go the other way: if losses recover above the threshold before the wind-down completes, the stop is cancelled and the bots keep playing.

The same is true of a win goal. If profit drops back below the goal before the shooter sevens out or the active bets resolve, the stop is cancelled and the bots keep playing.

A worked example. The stop loss is $500 and the timing is "finish the current shooter."

1. Mid-shooter, losses reach $520. The stop triggers and the bots keep playing this shooter.
2. One way it can go: the shooter sevens out, and the bots leave down $640.
3. The other way: the shooter hits two numbers and losses come back to $450. The stop is cancelled and the bots keep playing.

Ask for "stop immediately" if you need a hard limit.

Does money on the table count?

By default, no. Profit and loss are measured on the bots' bankroll, the money in the rack. Bets in play are wagers that haven't yet resolved, so money on the felt counts as spent until it comes home.

What this means for each rule:

• Stop loss. Putting bets on the table looks like a loss. A $300 stop loss on a system that puts $400 in play triggers the moment the bets go up, before anything has lost.
• Win goal. Money on the table isn't counted toward the goal, so the goal triggers later than you might expect. A bot with $200 more in the rack and $300 of bets on the table has not reached a $250 win goal.
• Giveback from the peak. Putting bets on the table looks like a giveback. See the next section.

The rule of thumb: if your system keeps more money in play than the stop amount, either raise the amount or ask for the other measure: "count bankroll plus bets in play."

A system with a lot of money in play, asking for the other measure:

Giveback from the peak: a worked example

You say: "Once profit reaches $500, walk away after giving back $200 from the peak."

1. Profit reaches $500. The stop activates and starts tracking the peak.
2. Profit climbs to $1,000.
3. Profit slides back to $800. That's $200 below the peak, so the stop fires.
4. The bots leave with about $800 in profit, or less if bets are still in play.

Below the activation amount, nothing happens. A plain "walk away up $500" is a fixed target, not a giveback. If you want the giveback, ask for it by name and give both numbers.

The giveback is a trigger, not a guarantee

The bots can give back more than the amount you set. The stop is checked after each roll, spin, or hand, and one roll can move more than the giveback. A seven-out with $500 in play takes $500 at once, so the bots leave at the first check at or past the line. With "finish the current shooter" or "let active bets resolve", the bets still in play can lose too.

Size the giveback to your bets

Set the giveback larger than the money your system normally keeps in play. Profit is measured on the rack, so putting $500 of bets on the table looks like a $500 drop. A $200 giveback would fire the first time the bets go back up after it activates, before the bots have lost anything. For a tight giveback, ask for "count bankroll plus bets in play."

The whole thing as one prompt:

Session length is a separate rule

A win goal is not a session length. One is a dollar amount, the other is how long the bots play. A prompt that mixes the two leaves the AI guessing.

Wrong. Don't copy this one:

Is that 60 spins, a win goal, or both? It doesn't say how much the win goal is.

Right. Say each one on its own:

You can cap a craps session by shooters or by rolls, a roulette session by spins, and a baccarat session by shoes or hands.

The longest session you can run is 50 shooters or 1,000 rolls in craps, 1,000 spins in roulette, and 3 shoes or 400 hands in baccarat. If you ask for more, the ticket will tell you it is running the maximum.

Casinos don't run on a clock in the simulation. "Play for one hour" gets converted, so give the number of shooters, spins, or hands yourself.

Streaks

"Leave after 11 losses in a row" is a walk-away rule too. It's a counter, not a dollar amount. Say "in a row", and say what counts as a loss for your system: a losing spin, a seven-out, a lost pass line bet.

What you can't do yet

• Carry a bankroll from one session to the next. Each bot plays one session. A plan like "ten sessions, raise the bet after each winning session" runs as stages inside a single session, and the report scores the whole run.
• Leave when the point resolves. The three timings above are the choices. "Let active bets resolve" is the closest.

Check the ticket

Before you build, find the walk-away lines on the ticket. Each one should name the amount, the direction, and the timing, like this:

If a line isn't what you meant, quote it and say what it should be.

Compare two walk-away rules

Not sure whether to leave up $300 or up $500? Run both as variants. Every bot plays the same dice in each one, so the difference in the results comes from the rule alone. See Compare two systems.