ORLOV GAMES

Blog · Video Poker ·

Why practising video poker offline teaches more

Two numbers explain why free, offline practice is the better way to learn video poker: the tiny cost of the game per hand, and the enormous swing that hides it.

Most people learn a skill by doing it and watching what happens. Video poker is a poor fit for that method, because what happens on any one hand, any one session or even any one month is mostly luck. The maths below, for the standard 9/6 Jacks or Better paytable under correct play, shows why your results cannot tell you whether you are playing well. Something else has to.

Two numbers that describe the game

9/6 Jacks or Better (a full house pays 9, a flush pays 6) returns 99.54% when every hand is played correctly. For each bet, 0.9954 comes back on average, so the shortfall is 0.46% of a bet. Call that the edge.

The second number is the standard deviation, a measure of how far a single hand's result typically lands from that average. For this table it is 4.42 bets per hand. Most of that comes from the royal flush, which pays 800 for a five-coin bet and arrives about once in 40,400 hands. The full working is in What variance actually means at a video poker machine.

Put side by side, the edge is 0.0046 of a bet and the typical swing is 4.42 bets. The swing is roughly 960 times larger.

How each number grows with hands played

The two numbers do not grow at the same rate, and that is the whole story. After n hands of one bet each:

Both are in bets, so the table holds whatever the bet. About two results in three land within one standard deviation of the expected figure, and about nineteen in twenty within two.

9/6 Jacks or Better, correct play, in units of one bet per hand. The 400-hand row is one hour at 400 hands an hour.
HandsExpected lossOne standard deviation
1000.46 bets44 bets
4001.84 bets88 bets
1,0004.6 bets140 bets
10,00046 bets442 bets
100,000460 bets1,398 bets

In an hour at 400 hands, the game's built-in cost is under two bets, while the ordinary swing is 88. A good hour and a bad hour are both well inside the noise. The ratio of swing to expected loss is 960.9 ÷ √n, and it only reaches 1 when n is about 923,000 hands. Before that, luck is the larger of the two.

One standard deviation of the result for every 1 bet of expected loss, by hands played, 9/6 Jacks or Better. At 400 hands, one hour at 400 an hour, the swing is 48 times the loss; it takes about a million hands to bring the two level.

Why that makes results a poor teacher

Suppose you play with a small leak: say a habit that costs one percentage point of return, 0.01 of a bet per hand. That is more than twice the game's own edge. How many hands before the numbers show it?

To tell the leak apart from luck at two standard deviations, the leak has to reach twice the swing: 0.01 × n = 2 × 4.42 × √n. Solving gives √n = 884, so n is about 781,000 hands. At 400 hands an hour that is nearly 2,000 hours.

So the feedback loop most skills rely on does not exist here. A session in which you broke a rule and won tells you nothing, and neither does a session in which you played perfectly and lost. Both happen routinely. Strategy has to be learned from the strategy itself: which hold has the highest expected value for this deal. Jacks or Better strategy, ranked sets out the order of holds, and the most useful practice is the kind that checks your choice against that order on every deal.

What free practice gives you

Because results cannot tell you how you are doing, the things that do tell you matter more. These features come from the app's own description.

The rebuy changes how you can learn. A bad run costs nothing, so you can sit through a few thousand hands and see what a 4.42 swing feels like: stretches where nothing pays, then a cluster of wins, none of which says anything about your play. The bankroll and session length post has the arithmetic behind how long a given run can last.

Over any sensible number of hands, luck outweighs the game's built-in cost by dozens of times. Whether you are playing well shows up in your decisions, not your results, so practise where you can check the decision.

A practice routine that uses the numbers

  1. Learn one game first. Jacks or Better has the shortest hold order.
  2. Use hints until the holds are automatic, then switch them off and see how often your choice matches.
  3. Ignore your net result. Count hold mistakes instead.
  4. Move to a higher-variance game later. Full-pay Deuces Wild has a standard deviation of about 5.1 against 4.42 here, so the swings feel bigger.

The hours then go into the one thing you can improve, the choice on each deal, rather than the outcome you cannot control.

Video Poker: Jacks or Better is a free, offline practice game with no ads. The chips have no cash value and the game does not offer real money or prizes.

Get it on Google Play