Poker
Strategy models, hand-history analysis, simulators and bots. Model the decision using the cards and information actually available, then account for rake and the opponents in the game.
A hand-state example
Synthetic example with assumed probabilities and costs.
There are 100 units in the pot, including the opponent’s bet. Calling costs 40 units. Assume no further betting, no ties and a 30% chance of winning.
| No rake | +2.00 units |
|---|---|
| 5 units of rake from completed pot | +0.50 units |
No rake: 0.30 × 100 − 0.70 × 40 = +2.00
With rake: 0.30 × 95 − 0.70 × 40 = +0.50
After rake, a win returns 135 units from the completed 140-unit pot. After the 40-unit call, the net gain at this decision is 95.
Break-even equity rises from 40 ÷ 140 = 28.57% to 40 ÷ 135 = 29.63%. The calculation covers one terminal decision under the assumed rake schedule.
The model’s inputs and output
A decision record includes format, position, effective stacks, pot size, action order, bet sizes, board cards and the information available when an action was chosen. Unshown opponent cards remain unknown.
- Hand analysis
- A comparison of candidate actions under explicit range, rake and future-play assumptions.
- Strategy research
- A simulator or evaluation tool for a specified game format and opponent population.
- Automation
- A strategy connected to an authorised environment, with integration and stopping checks.
Game and format coverage is scoped for the project. In tournaments, chip and prize expected value can differ; payouts and future-play assumptions belong in the comparison.
Win rate and opponent mix
Test against players, sessions or later periods that match the intended use. Show game format, stakes, opponent selection, sample, rake, baseline and uncertainty.
Showdown hands are a selected sample. Missing sessions, repeated opponents and changes in the games played can distort a win-rate estimate. More hands alone cannot remove those biases.
Synthetic population example
A fixed strategy has an assumed pre-rake expectation of +8 big blinds per 100 hands against population A and −2 against B. The percentages below are shares of hands played against each population.
| 80% A, 20% B | +6 bb/100 |
|---|---|
| 20% A, 80% B | 0 bb/100 |
Subtract an illustrative average cost of 2 bb/100 and the net rates become +4 and −2 respectively.
The same strategy produces different win rates against a different opponent mix.
Where the bot acts
A bot needs a named game environment, connection method, authorised users, platform permissions and stopping conditions. Live automation needs permission and performance tests for that environment. A hand-review tool or simulator does not imply live-game access.
Start with the format, decision and intended use. Hand histories and project details can be discussed after the disclosure and access terms are set.
Discuss a poker project