[00:01] Everybody at the table has a roulette system. Your uncle doubles the bet after system. Your uncle doubles the bet after every loss because sooner or later red Your friend walks away the moment the night is $10 ahead. Somebody waits for [00:14] five reds in a row and then bets black because the wheel is due. And there is always one person who covers almost the whole table so that nothing can go wrong. You know who you are. So, I turned 19 of those habits into [00:30] computer players and made them play 100 million nights at the casino. Here is million nights at the casino. Here is the headline. The best system I tested won 98.14% of its nights. Nearly every time it sat [00:44] down, it walked out ahead. The title says 99% and I will show you exactly where that number lives because it is real and it is not what you think. Here [00:56] is the catch. Every one of those systems lost the same price. $2.70 for every $100 it bet. The system that wins almost every night pays that price [01:09] just like the one that loses almost every night. And after 10 years of weekly visits, about three players out of four using the best system were behind. So, how do you win 98 nights out of 100 and still lose? The answer is a [01:23] pattern in how the losses arrive. Once you have seen it, you will never look at a winning streak the same way. One thing before we start. This video is not financial advice and it is not a reason to gamble. Every system you're about to [01:36] see lost money. First, a quick word about the table. I used a European about the table. I used a European wheel. 37 pockets, the numbers 1 to 36, wheel. 37 pockets, the numbers 1 to 36, and a single zero. 18 red, 18 black. A [01:50] and a single zero. 18 red, 18 black. A bet on red pays 1 to 1. A dozen numbers pays 2 to 1. A line of six numbers pays 5 to 1. A single number pays 35 one. When the ball lands on zero, every outside bet loses. That one pocket is [02:05] outside bet loses. That one pocket is the house edge. 1 in 37 or 2.7% of every bet on the table. The table maximum is $500 a bet and every system plays exactly the same spins. Nobody gets a luckier wheel. Here [02:20] is a night you arrive with $1,000. You leave when you are $100 ahead or after 200 spins or when you cannot afford your next bet. hours. If a system wants a bet that would break [02:36] money it starts over from his first bet and the losses so far stay lost. And one means one thing. You left with more money than you brought. The $100 goal and the 200 spin night are [02:51] my assumptions. They are not rules of any casino and I will come back to that. Now the players and they come in families. The Martingale family, the plain Martingale one that walks away after $10, one with [03:05] one that walks away after $10, one with a $10,000 bankroll, the grand Martingale and one with a stop loss. The slower progressions D'Alembert, Fibonacci, Labouchère, Oscar's Grind [music] [03:17] Oscar's Grind [music] and one called 1-3-2-6. which does the opposite of the Martingale and raises the bet after a win. The coverage systems, James Bond, 35 [03:30] The coverage systems, James Bond, 35 numbers covered and two dozens tripled The pattern players one that waits for five reds and then bets black and one that bets the hot number. [03:43] number. And the baselines, a single spin of $100 and flat bets with no goal at all. Plus the one I call normal. Normal bets $5 on red every spin and goes home at $100 ahead or after 200 spins. [04:00] It is the most boring person in the building, which makes it a fair yardstick. Out of control, a second copy of normal players. >> [music] [04:13] >> Each one played 5 million nights. That is 100 million. Before I believe anything, I need to know the simulator is honest. So, I checked it against answers I [04:26] First, I wrote a second plain version of every system and replayed 2,400 nights on both wheels. It matched the main engine exactly on number of spins. Second, the wheel itself. A standard [04:43] fairness test asks whether any number comes up too often. My wheel scored 41.4 across 37 numbers, which is the kind of score a fair wheel gives. Third, the exact math. [04:56] Third, the exact math. Flat $5 on red should lose 2.703% of what it bets. My run lost 2.723, and seven losses in a row, which you will meet in a minute, [05:09] should happen in 9.42 cycles out of a thousand. Mine did it 9.46 times. Then, the control. The copy of normal won 32.80% of its nights against 32.79 for the original. The difference in average [05:25] original. The difference in average money was 5 cents with a standard error >> In other words, nothing. One more thing, planted now because it These players follow their system exactly. [05:39] >> They never get scared, tired, or greedy. And none of them ever says, "One more spin." Real people do. So, where does 99% come from? Let me read the leaderboard first. By the share of nights won, from the [05:53] By the share of nights won, from the bottom. 35 numbers covered wins 4.08% 32.79. Paroli 43.92. Paroli 43.92. Then, 1326, 48.59. [06:08] Then, 1326, 48.59. In a single spin of $100, 48.67, which is about a coin flip. Now, up the board. The plain Martingale, 82.8. D'Alembert, 83.71. [06:24] The Grand Martingale, 86.4. Labouchère, 87.76. Waiting for five reds, 91.65. And at the top, the Martingale that And at the top, the Martingale that walks away after $10, 98.14. [06:40] Look at who sits at the top. Every one of them raises the bet after a loss. The systems that raise the bet after a win sit near the bottom. So, the Martingale is where the story starts. Here is how it works. Bet $5 on red. [06:56] >> If you win, you are $5 ahead, and you start over. If you lose, bet double. Lose again, double again, until one win pays back everything you lost plus $5. A [07:10] win always leaves you $5 ahead. Now, watch how it fails. The doubling bets >> [music] >> 5, 10, 20, 40, 80, 160, 320. That is [07:23] seven losses in a row, and you are down $635. The next bet would be 640. The table maximum is 500. The ladder is over. The maximum is 500. The ladder is over. The cycle ends with the loss of $635. [07:38] How often does that happen? In 0.942% of cycles, that is a little under 1 in 100. So, 99.06% 100. So, 99.06% of cycles win $5. There is the 99% in [07:53] of cycles win $5. There is the 99% in the title. It is the share of cycles, not the share of nights. I am telling you now, so nobody has to wonder. Now, the nights. The Martingale that walks away after $10 needs only two winning [08:07] That is two chances for the one in a hundred disaster, and it won 98.14% of its nights. The plain Martingale keeps going until it is $100 ahead. That is 20 [08:20] 20 chances for the disaster, and it won 82.8%. Same bet. The only difference is how disaster on it. Which brings us to the arithmetic that [08:35] Which brings us to the arithmetic that the 99% hides. 99.06% of the time, you win $5. Averaged over every cycle, that is $4.95. 0.942% of the time, you lose $635. [08:52] Averaged over every cycle, that is $5.98. Subtract one from the other, and the average cycle loses $1.03. Look at the nights. The walk away Martingale wins [music] $10 on a winning [09:08] On a losing night, it loses on average $695.14. So, the average night for the best system in my tests [09:22] >> is a loss of $3.11. It is like collecting small premiums on low. Most days, the customer never claims, and then one day the claim arrives, and it needs about 70 premiums at once. [09:39] You may be thinking, "Fine. That is the Martingale. Surely some other system escapes." So, let me check the other 18. Here is the wall. Take every system and measure how much it loses for every $100 [09:53] measure how much it loses for every $100 of bets. The answer sits between $2.62 >> [music] >> Every single one. The shapes are very The Martingale family wins almost every night and loses big. [10:07] D'Alembert wins 83.71% of its nights and still loses $64.18 on an average night. Labouchère wins about $100 on a good night and loses [10:20] about $100 on a good night and loses $979.32 on a bad one. James Bond wins $141.77 and $890.60 goes the other way. Normal wins one [10:36] night in three and when it loses, it loses about $70. Different shapes, same price. Now, add everything up. Across all the nights of all the systems, $145.5 [10:49] was bet. $2.70 of every hundred was lost. That is the house edge. 1 in 37 working through every system as if the system [11:01] were not there. The reason is simple. Every bet you place costs, on average, 2.7% of its size. A system cannot change that. put on the table and when the loss arrives. [11:15] So, the real question is not whether you win tonight, it is what the calendar does to a system like this. Take a player who visits the casino once a week and uses the best system, the Martingale that walks away [11:28] after $10. After a year, 52 nights, how many of these players are ahead? 51.29% about a coin flip. After 10 years, 520 nights, the answer drops to 26.69% [11:43] about 1 in 4. That means three out of four are behind. The plain Martingale does worse. Ahead after a year, 26.73% Ahead after 10 years, 1.54 Now, compare that with the most honest [11:59] Now, compare that with the most honest bet in the building. One spin, $100 on bet in the building. One spin, $100 on red once a week. It wins 48.67% 98. But, after 10 years, [12:12] >> it is ahead 25.66% of the time. Almost exactly as often as the system that wins 98 nights in a hundred. And the money? After 10 years, the walkaway Martingale player is down, on [12:28] walkaway Martingale player is down, on average, $1,617.19. [12:42] The system that wins almost every night costs more than $230 extra. Now look at normal, the flat $5 better. Ahead after a year, 0.36% ahead after 10 years, 0.00. Why is the lumpy system ahead more often than the [12:57] smooth one? Because being ahead is about the shape of the results, not about the average. Normal loses a little every night with no lucky spike to hide it. The Martingale wins a little almost every [13:11] night. And so a year of wins can feel like a pattern. A lumpy system can feel like a pattern. A lumpy system is ahead more often. It is not ahead on average. Next question. If the problem is how [13:23] with money? >> [music] >> Everyone has a version of this comment, "Just use a bigger bankroll." So I gave $1,000. >> [music] [13:35] >> [music] >> Its win rate did not move. 82.8% exactly the same. The table maximum of $500 still stops the doubling after What the extra money bought was a bigger worse night. A loss of $5,330 [13:52] Martingale. I also raised the table maximum to $5,000. Now the $10,000 Martingale wins 97.48% [14:04] of its nights. And its worst night is the whole $10,000. More money did not make the system safe. It made the disaster larger. Next comment, "Cover more numbers." So, here is 35 numbers covered, $1 on [14:18] So, here is 35 numbers covered, $1 on each. It wins 94.6% of spins. That building. But, a win pays $1. A loss costs 35. [14:30] >> [music] >> It wins only 4.08% of its nights, and loses $188.56 on an average night. After a year, it was never ahead. Covering more numbers just means putting [14:43] more money on the table, and paying 2.7% on all of it. This system bet $6,977.95 for a streak, the wheel is due. So, I built a player that waits for five reds [14:59] in a row, then bets black with a Martingale. It won 91.65% and it lost [music] $2.72 for every $100 it bet. The same price. [music] [15:14] The wheel has no memory. Each spin starts fresh. What the waiting player really did is sit out most spins. It bet only $102.22 [15:26] sit out most spins. It bet only $102.22 on an average night, against $1,884.22 for the plain Martingale. It looks safer only because it bets less. A hot number player works the same way. [15:38] It keeps betting the number that has come up the most. After 10 years, >> it is ahead 16.76% of the time, which looks decent, but it still loses $2.73 [15:52] per $100 bet. It looks better over the years only because it puts less money on the table than most of the others. Okay. One more idea, and it is the one that [16:05] >> [music] >> Quit while you are only down a little. This is the Martingale with a stop loss. It gives up after five losses in a row. And it works in one way. On a bad night, it loses $339.58 [16:22] On a bad night, it loses $339.58 against $650.54 for the plain Martingale. But, it wins far fewer nights, 68.78% against 82.8. After a year, it is ahead only 11.17% [16:37] of the time against 26.73. After 10 years, 0%. The stop loss cuts the size of the disaster, and it makes the disaster arrive more often. The price stays the same. [16:51] Now, the swing factor. Everything so far used the European wheel. probably meet a wheel with a double zero, 38 pockets, two of them green. On that wheel, every win rate falls. The loss for every $100 bet roughly [17:07] doubles to between $5.20 and $5.31. And the walkaway Martingale still wins 97.76% of its nights. [17:19] A higher price and the same illusion. So, what do you do with all this? Here are five rules the data supports. do not pay for a system. No system change the price. Two, [17:33] when someone brags about a win rate, ask how big the losing night is. Three, remember that betting more money means losing more money, whatever the pattern >> [music] >> ignore the streaks. The wheel does not [17:46] remember the last spin. Five, if you ever sit at that table, decide what the night may cost before you sit down. Treat that money as already spent. It is the price of an evening of entertainment and nothing else. [18:00] and nothing else. Now, the limits, because they matter. This is a fair wheel. There are no biased wheels and no dealer habits. Every bet has the same table maximum, and real tables often set lower limits [18:14] on single numbers. The progressions assume a $5 minimum bet. There are no comps, no tips, and no tax. The $100 goal and the 200 spin night are my assumptions. And published versions of each system vary in their [18:29] details. [music] So, yours may not match mine exactly. And the biggest limit of all, the players follow their rules exactly. They never get tired, they never get scared, and they never think, "Just one more spin." People do. That [18:45] can measure. So, tell me in the comments which game should I simulate next. And if you ever do sit at that table, do sit at that table, know what you are buying. Not a system. [18:59] An evening at a price of $2.70 for every $100 you bet. Decide that price before you sit down. And if the game stops being fun, search And if the game stops being fun, search for your country's gambling helpline.