BOOK 2 · PART I · PLAY · TEACHER CHAPTER
Running a Crawl Session
One class period, a room full of students, nine robots each, and a stack of hand journals. How to run it, what to say while you walk the room, how to judge a decision without looking at the chips, and what to keep for later.
What a Crawl session is for
The class game in "The Class Plays the AI" was loud: thirty people arguing over one seat. A Crawl session is the quiet version. Each student plays the robots alone, at their own pace, and keeps a hand journal. Your job isn't to teach poker. It's to keep asking one question until students start asking it themselves: what were you thinking?
Students don't have to win, and they don't have to finish. One hand is enough to learn something; a hundred is the most, not the goal.
Before class
- Print the journal. What Were You Thinking? is three pages. Print pages 1 and 3 for everyone, and a few spare copies of page 2.
- Pick the table. The Advanced Simulator seats nine robots and the student. Use the standard nine-robot lineup from "The Standard Table" so everyone faces the same opponents, or leave the Sushi Menu blank so every robot plays every hand. Blank is the easiest start: the robots never fold, so students see plenty of flops.
- Check the logins. Every student needs to sign in and later open Game Review. Try it yourself once, start to finish, before class.
- Have the refresher handy for anyone unsure what a check or a raise is.
A 50-minute period
| MINUTES | WHAT HAPPENS |
|---|---|
| 0–5 | Hand out journals. Students fill in the top line and write down their game number as soon as the first hand is dealt. |
| 5–30 | Play. Easy folds get a tally mark; hands worth a thought get written up. You walk the room. |
| 30–40 | Game Review. Students open their game, find each hand they wrote up, and fill in the "At Review" strip. |
| 40–50 | Share-out (questions below). Students fill in page 3. Collect page 3. |
Students who quit early go straight to Game Review. That isn't giving up; it's the same as a player leaving the table and studying the hands.
What to say while you walk the room
- Ask about the why, never the score. "What were you thinking on that one?" works. "Are you winning?" teaches the wrong lesson.
- Treat "Not sure" as a good answer. It's honest, and it's exactly what Game Review is for.
- Ask what the gut noticed. When a student fills "Gut" or "Something else," ask what they noticed. Sometimes it's a real read; write it down if so.
- Watch for tilt. A student who loses a big hand and starts playing every hand is showing the lesson of "What Were You Thinking?" live. Ask quietly whether their last few reasons look like their first few.
- Don't let them write everything. If a journal is filling up with seven-two folds, point to the tally line. If a journal is empty after twenty minutes, ask about the last hand that was close.
Judging a decision without the chips
The "At Review" strip asks: was it a good decision, whatever the result? Students find this hard at first. One test helps: does the reason match the facts? If the reason was "strong starting hand," check the cards' Sklansky group. If it was "the price was right," check the bet against the pot. If it was a read on the player, check what that player did earlier.
Try it yourself first
Five journal entries. Good decision, bad decision, or not sure?
- Pocket kings, all-in before the flop heads-up. Lost to aces. Reason: Strong starting hand.
- Seven-high, called a huge river bet, won. Reason: Gut.
- Checked a flush on the turn against a robot that never folds and rarely bets. It checked behind; won a small pot. Reason: Setting a trap.
- Folded on the flop when nobody had bet. Reason: Weak cards, let it go.
- Called a small bet on the turn with a flush draw; the pot was large. Missed, folded the river to a big bet. Reasons: The price was right, then They bet big, scared me.
ANSWER KEY
- Good. The reason matches the facts: kings are Sklansky group 1. The result was the cards.
- Bad, probably. A seven-high call wins only against a bluff with even less. "Gut" is honest; ask what it noticed.
- Bad read. A trap needs an opponent who bets. This robot plays like a Calling station: it never folds and rarely bets, so it checked behind and the flush won a small pot. Bet it instead. See "Read the Player, Set the Trap."
- Bad. When nobody has bet, checking is free. Never fold when you can check.
- Good, both times. A cheap call with a draw into a large pot is a fair price. Missing and folding to a big river bet is sensible too. The result doesn't change either decision.
Share-out questions
- Which reason did you use most? Was it the one you expected?
- Find a hand you lost but played well. Find one you won but played badly.
- Which robot at your table was easiest to read? What gave it away?
- Read out one line from your "What no checkbox could say" box.
What to keep
Collect page 3 from every journal. The tallies show the class's habits. The "What no checkbox could say" lines are the most valuable thing the session produces: in Part II, students try to turn those reasons into a robot's rules and find which ones won't go; in Part III, the same lines become instructions for an AI.
Written by Kevin Swinson with Claudette, the pen name for Claude, the AI from Anthropic that helped build HoldemRobots.AI. The 614 flags and their five patterns are from Backroom Dialog 10. The period plan is a suggestion; one hand, or a hundred, both count.