The Hold'em Simulator allows you to simulate all-in situations between two or more hold'em hands. It allows for simple simulations, where you know exactly what every player has, as well as complex simulations, where some players could have any one of a range of hands.

The race results page tells you the pot-equity percentage for each hand, as well as the number of wins and ties for each hand.

This represents the average amount of money a particular hand wins from the pot, after taking ties into account.

For a race to be valid, it must have two or more hands. Also, make sure there are enough cards in the deck to satisfy your request. For example, in hold'em, you can't race aces against ace-king if there are two aces on the board.

"Unroll" allows you to see the pre-flop, flop, turn, and river odds for a hand on one page. Simply enter all of the hands and the final board, and click the button.

You'll want to look at the Equity distribution graph instructions.

Good question. This only becomes a problem if only one card has a suit specified. In that case, for suited you put the 's' after the second card. For spades, put the 's' after the first card. Here are some examples:

• AsKs - ace of spades, king of spades
• AKs - ace king suited
• AsK - ace of spades and any king
• KsA - king of spades and any ace

We support two kinds of ranges. In an "anchored" range, the largest card listed on both sides must be the same, such as AK-AT. In a "distance" range, the distances on both sides must be the same (which is why AK-J9 is not valid, but QTs-86s is). In both kinds of ranges, the suits must be identical on both sides.

Not at this time. If you specify any board cards, they must all be completely specified. For example, AsKdQh is ok, while AKQ is not.

An exhaustive simulation is one where every possible combination of hands and boards is dealt, and the resulting answers are exact. A random simulation is an approximation. In a typical random simulation, ProPokerTools performs 600,000 trials, which for most folks is more hands than they will see in years. With 600,000 trials, pot equity percentages will rarely be off by more than a few tenths (or hundredths) of a percentage point. In some cases, even 600,000 trials may take too long, in which case a smaller number of trials are run.

We used a computational method inspired by an academic paper. The algorithm gradually refines a set of "good" hands, and in the process, an ordering is created. For more details on the algorithm, take a look at this blog entry.

Yes, you can use this form.

Sure thing - here they are in plain-old text:

• Hold'em hand ordering
• Omaha-hi hand ordering
• Omaha-hi/lo hand ordering

When you request the top N% of all hands, we generate as close to N% hands without going over, just like in "The Price is Right" for you gameshow rerun addicts. Furthermore, we ensure that all equivalent hands are either all in or all out. (For example, in hold'em, there are 1326 possible hands. The top 1% of hands should be 13 hands using our price is right system. The top 13 hands would be six pairs of aces, six pairs of kings, and one pair of queens, but we remove the one pair of queens to conform to our "equivalent hands are either all in or all out" rule. Our result? 1% in hold'em means aces or kings).

If our simulator dealt through each and every hand and board combination, then yes, it would require some massive computer power to compute so many results so quickly. However, there are many symmetries in a typical hold'em simulation, and a clever implementation can use these to drastically cut down the actual number of boards that need to be generated. In addition, our server may use caching to avoid duplicating simulations.

Even with our highly optimized simulator, some problems are just too big to solve in any reasonable amount of time. In general, the simulator will perform an exhaustive search if it thinks it can complete it in a matter of seconds. Anything bigger than that, and it switches to random mode.

Currently, ProPokerTools uses a queue to ensure that only one simulation is every being computed at a time. This is to prevent our server from being overloaded by multiple requests. If your request is "being computed" and is still taking forever, please email us with the details and we'll check it out.