Poker Equity Calculator

TODO:

• CLI
• Further optimization?
• Benchmarks
• Usage instructions / examples (aka finish this README)

Motivation

My goal was to write a pure Go implementation of a poker equity calculator. I want to do so without using any external libraries, and without using any external data sources. Furthermore, I didn't want to use a Monte Carlo simulation to calculate the equity.

Usage

CLI WORK IN PROGRESS

For now, update main.go with your own hands and run:

go run main.go

Benchmarks

To check out the benchmarks, run ./bench.sh. If you can't run the script, try running chmod +x bench.sh first. You can also just run go test ./benchmarks -bench=. -v in the root directory.

This will run two tests, one for hand evaluation, and the other for heads-up preflop equity calculation.

Hand Evaluation

My approach is based on Cactus Kev's Poker Hand Evaluator.

If you don't want to click on any links, here's a summary:

After enumerating and collapsing all the 2.6 million unique five-card poker hands, we wind up with just 7462 distinct poker hand values.

Hand Value	Unique	Distinct
Straight Flush	40	10
4 of a Kind	624	156
Full House	3744	156
Flush	5108	1277
Straight	10200	10
Three of a Kind	54912	858
Two Pair	123552	858
One Pair	1098240	2860
High Card	1302540	1277
TOTAL	2598960	7462

We can turn that into a lookup table. By representing each card with a prime number, we can get a unique product for any 5 card hand.

Cards can be represented by 32-bit integers as follows:

+--------+--------+--------+--------+
|xxxbbbbb|bbbbbbbb|cdhsrrrr|xxpppppp|
+--------+--------+--------+--------+

p = prime number of rank (deuce=2,trey=3,four=5,...,ace=41)
r = rank of card (deuce=0,trey=1,four=2,five=3,...,ace=12)
cdhs = suit of card (bit turned on based on suit of card)
b = bit turned on depending on rank of card

Equity Calculation

Simple — we just enumerate through all possible run-outs and count the number of times each player wins/loses/ties!