Having discussed the possibilities for predicting roulette spins with a real wheel it’s time to think about online roulette which uses a random number generator. It’s worth bearing in mind here that there are a few sites which use a webcam link up of a real wheel. Obviously the methods mentioned in the previous article are relevant there.
It’s worth knowing a little bit about how a random number generator works. Mathematicians actually refer to them as pseudorandom number generators (PRNGs) because they do not generate truly random numbers but instead generate sequences of numbers which (it is hoped) are indistinguishable from random numbers. There are many different approaches but the basic idea is to start with a seed number which is used as an input to the PRNG, then the output is used as a new input. A well designed PRNG will produce outputs which are evenly distributed and unpredictable.
A little thought shows that eventually a PRNG using this method alone will start to repeat (once an output is generated which is the same as the seed). To combat this elements of randomness can be inserted into each input. Common methods include measurements of the time, mouse position, radioactive decay, radio interference or cameras pointed at lava lamps or the sky.
A further problem which any piece of software aiming to track and exploit problems with a casino PRNG will have is that you don’t necessarily get a stream of numbers being generated for a particular table. All the websites I saw selling this type of software seem to be assuming that there’s a separate PRNG running for each table, which generates a new number each time the wheel is spun. In fact it is more usual for them to have the PRNG running all the time, generating many thousands of numbers per second. When a table needs one it is simply given the latest to be generated. This alone provides huge problems for any piece of predictive software, as it is difficult to see how it could possibly infer the link between successive results.
The icing on the cake is that the end user does not have raw access to the PRNG output. The output will typically be a very large number which is then assigned via a hash function to one of the numbers on the wheel. This adds a further level of obfuscation to tracking the PRNG outputs.
In summary I simply cannot see how these pieces of software can possibly work. Even if the PRNG was poorly designed and the numbers were generated on a per wheel basis, the period before repetition would be so immense that you would need to capture a massive amount of data before any prediction could begin.
If I were cynical I would say that these “roulette predictors” were probably nothing more than a PRNG wrapped up in a user interface to trick the user into thinking their input is affecting the output. I’m open to be proved wrong if anyone has one they’d like examined.
In replying to a comment I did some research into systems (computers, software and paper) designed to predict roulette spins, both online and on real wheels. This is a somewhat popular area, probably down to the famous story of the Eudaemons. What I found is that there are a lot of people out there selling these products, and a lot of stories of people getting scammed.
Before we proceed, two notes of caution. Firstly you should avoid any roulette system which is based on betting patterns, progressive bets or anything of the sort. If there is a weak point to roulette then it is in the spin of the wheel. The mathematics of the table bets are simply stacked against you and no amount of fiddling with bet sizes or patterns can change that. Secondly you should treat any “miracle system” with extreme skepticism. Think of the true value of anything which can give you an edge in a casino game. Why is it offered for sale? When a business will usually be valued at several years worth of turnover, why is this (usually low) price being offered?
With those caveats, let’s look at systems for predicting the spin of a real roulette wheel. I’m not vouching for the validity of any particular system or piece of equipment, but I’m fairly certain that this can be done with sophisticated enough equipment and accurate data input. Remember that in American roulette you would only need to be able to predict a number with 1/35 accuracy instead of the normal 1/38 to give you a significant edge. There are three main methods;
[Read more]
Imagine you’re playing a game where you bet on the flip of a coin against a casino, and the casino gives you even money on the bet. They don’t have a house edge. Even without a house edge, if you play for long enough you will lose all your money.
That might not seem to make any sense, but I’ll explain. Let’s say you’re playing the game for £1 per flip, and you have £100 as a bankroll. For the most part you and the casino will be pretty much even - after all the game has a house edge of zero, so in the long run you don’t expect to win or lose anything. The problem with that is that the longer you play, the more likely you are to get a long enough string of losses (it will take about 100) to make you lose all your money. And once your money’s gone you have nothing to win it back with - the casino has won.
It’s known in probability theory as the gambler’s ruin problem, and it highlights the problem with playing a game against an opponent with much more money than you. You don’t stand a chance of making the casino run out of money, but unless you’re careful to always bet small amounts compared to your bankroll the casino can wipe you out with a run of bad luck. The exact chance of this happening is known as the risk of ruin and in general is pretty tough to calculate. Essentially though it depends on two things - what proportion of your bankroll you bet each game, and what the variance of the game is (in other words how risky it is).
Our game above isn’t very volatile, and £100 is quite a large bankroll for a £1 game. The odds of getting 100 losses in a row is 1/1267650600228229401496703205376, which is very small indeed. Imagine we were playing a £10 version of the same game. The odds of getting the 10 losses in a row we’d need to wipe us out is 1/1024. That might seem small but considering we could easily play over 100 games an hour you can see we’d have a very high risk of ruin if we played for long.
Without going into specific figures, it’s crucial to have a large bankroll in proportion to the bet size, and the more volatile the game is (the bigger the up and down swings) the larger bankroll you should have. What sets apart the pros from the amateurs is having a bankroll that can take the downswings which are natural to any gambling activity.
Some people just like gambling, they enjoy the thrill. I can appreciate that. Personally though, I want to be confident that the game I’m playing can be beaten long term. In other words I want a game that doesn’t have a house edge. You still get the excitement because the variance inherent in the game means you get short term swings up and down, and it’s even nicer because you know that as long as you have the nerve to keep playing correctly the game will keep paying you.
To put it another way - if you enjoy gambling why not at least play a game you can win?
I define a beatable game to be one which either has a player advantage with correct strategy, or which can be given a player edge with a little extra work (e.g. card counting). It won’t necessarily be easy to beat the game, but it will be possible for a skilled player.
Beatable games:
- Blackjack (if you’re card counting or playing optimally in a game with very liberal rules)
- Craps (if dice control really works - currently I’m skeptical but open minded about it)
- Poker
- Slots (if you’re playing optimally at certain machines, or if the machine is progressive)
- Sports betting
- Video poker (if you’re playing optimally at certain machines - e.g. full pay deuces wild)
Unbeatable games:
In my opinion if you’re playing an unbeatable game, why not switch to a beatable one? Even if you’re not interested in making money, you can still make your money go a lot further.
Something which follows naturally from talking about the house edge, and which I found of the utmost importance in my work with poker A.I. is the concept of expected value (EV). The EV of an action is the sum of the probability of each possible outcome of the action multiplied by it’s payoff. In simple English it is the average amount you expect to make from taking an action.
A simple example is playing American roulette - the table has 38 equally likely numbers and a winning bet on one of them pays 35-to-1. Thus a $1 bet on a single number expects a profit of
which is -$0.0526 (this also shows the house edge in American roulette is 5.26%). We say that the expected value of the bet is -5.26 cents. When considering gambling strategy be wary of actions with a negative expected value - they are costing you money.
When playing poker we may be forced to choose between calling a bet we shouldn’t (with negative EV) and folding (EV of zero). In this case we should fold. We should always choose the action with the highest expected value.
