The bots that actually farmed real-money tables

Before neural networks and solvers, online poker's bot problem was rule-based. Thousands of accounts ran declarative profiles written in PPL — the Poker Programming Language. We interpret sixteen of those era profiles, rule for rule, so you can play the real thing safely.

What a PPL profile looks like

PPL reads like English. A profile is a long list of conditions checked top to bottom; the first match acts:

When raises = 0 and calls = 0 and stilltoact <= 4 and (pairinhand or hand = A suited) raise 2 force

That's real syntax. A commercial profile stacks hundreds of these rules — our collection ranges from a 10-rule miniature to an 813-rule grinder — and plays them with total consistency, hand after hand, forever.

Why they were beatable, and why that matters

Rigidity. A rule-based bot opens the same ranges, sizes the same bets and folds to the same pressure every single time. Regulars who identified a bot could exploit it mercilessly; security teams learned to detect them by the same signatures — identical timing, identical sizings, ranges with razor-sharp edges. Playing against these profiles is the fastest way to develop that pattern-reading instinct.

Total transparency: the rule trace

On our tables, every action a profile takes shows the exact rule that fired:

rule L217: BotIsLastRaiser and StackSize > 20 and bets = 0 and raises = 0 and not (FlushPossible…)

This is the bot's entire mind, on the record. No commercial operator ever showed you this — it is precisely what detection analysts reconstruct from thousands of logged hands, handed to you per action.

What we measured before shipping them

Each profile passed a coverage audit (100% of its rules evaluate against our engine — no silently skipped logic) and a character measurement. The collection spans real variety: the tightest profile opens 8% of hands under the gun and 13% on the button; the loosest opens 16% and 52%. Three table presets package them: a classic tight farm, a loose farm, and a random five-profile table.

The boring legal part

Profiles run server-side only and are never redistributed. We use descriptive profile names, not vendor trademarks, and this site is play-money and educational: the point is learning to recognize and beat rule-based opponents, not helping anyone bot a real-money site — which will get you banned everywhere, and should.