For 14 years (starting in 2009) Advanced Poker Training has been abuzz, with users playing a staggering 400 million hands against our AI virtual opponents. Our players hail from 175 different countries, and all walks of life. The achievement of 400 million hands played stands as a testament to the universal appeal and enduring fascination with this game of poker that we all love.
How It All Began
I first started working on the APT bots back in 2007, long before the era of AI had gained widespread recognition. The first version of our site played exclusively in the Adobe Flash Player (remember that? I still miss it.)
Since its humble beginnings, our system has grown, incorporating over 100,000 lines of programming code that control the intricate behaviors of the AI players. This evolution has culminated in the recent introduction of our revolutionary GTO-Style bots.
Lets Talk About those 400 Million Hands
I wish I could tell you that the 400,000,000th hand itself was an exciting one, but it wasn’t. It was played by user kaimadden, who was dealt A3o from early position. He folded of course, and that was the end of it.
This got me thinking, however. How many of those 400,000,000 hands were pre-flop folds? One thing led to another, and I queried the following statistics for your enjoyment.
What was the final street the APT user saw in each hand?
Hands Played by Game Type:
Ok, now let’s look specifically at some 9-max stats. I’m doing this because some game types above (e.g. Beat the Pro Challenge, Combat Trainer) would skew the data since those are pre-determined hands.
This is the number of random hands dealt to users of our 9-max simulator: 129,411,900
Note that this value is less than the 143M shown in the table above because I also eliminated hands where the user told our software to deal them a specific hand.
Now, let’s check to make sure our random number generator is truly accurate. The table below shows the actual number of pocket pairs, suited hands, and offsuit hands (e.g. A4o) dealt, and the theoretically correct percent that should be dealt in the long run.
For example, the actual percentage of pocket pairs dealt is 7,617,973 ÷ 129,411,900 = 5.887%.
Since there are 1,326 starting hands, and 78 pocket pairs, the theoretical value is 78 ÷ 1326 = 5.882%
Looks pretty good to me. You can see the Actual and Theoretical are extremely close in all cases.
Hands at Showdown
Total 9-max hands that went all the way to showdown: 6,266,596
Of those, at showdown the user had:
Now that’s an interesting result! Obviously, if you randomly dealt hands to showdown (and never folded), straights would be much more common than flushes, and flushes much more common than full houses. But here, you can see they are all very close – in fact there were slightly more flushes than straights. So is our software broken? No, of course not. The data is being influenced by the fact that people play a lot more suited hands (and pocket pairs) than off-suit hands, leading to the increase in flushes (and full houses) at showdown.
That’s about all for now. I’m so proud of this monumental milestone of 400 million hands, because to me it stands as a testament to APT’s enduring legacy and ongoing innovation. I hope to have a special celebration when we hit 500,000,000!