Spin a coin a million times and it is almost certain that you will have virtually 50% heads and 50% tails. Spin a coin twice and there's only a 50% chance that you will have this ratio. There is a 25% chance that you will have 2 heads and a 25% chance that you will have 2 tails - both of these outcomes are wildly out of context compared with the long-term average.

If FootySimulator was used to predict the outcome of a coin-toss, firstly it would gather evidence of all the coin tosses that had taken place over the years and then a simulation would be produced that announced that over the long run, say 1,000 tosses, that there would be around 500 heads and 500 tails. It could not predict with confidence the outcome of the next coin-toss to take place.

FootySimulator does not claim to know the outcome of any given football match in advance. To do that would require a time machine. FootySimulator works by absorbing information on over 100 years of competitive international football matches and then making risk assessments in terms of what would happen if that fixture was played x number of times. This is essentially no different from how an insurer operates. If an insurer knew with certainty who was going to have an accident and who wasn't, they would only insure the people they knew would not be putting a claim in. The insurer assesses risks based on input according to the policyholder's age, environment and lifestyle. They can then come up with a percentage chance of a claim being required and price up the policy accordingly.

There are two types of gamblers. Those who bet on an event because they think it will occur - and those who are prepared to place a bet that has a higher chance of occurring than the price would suggest. Those in the first category will lose over the long term because they will back teams at 1/10 when they should be priced 1/6. Those teams will win far more often than they lose. Yet they will lose more often than the investor can afford to make it a profitable exercise. The second type of gambler is seeking value. These people will be prepared to take a punt on a 7/1 shot that should really be 4/1. Although they would lose more times than they would win, the few wins would more than pay off the losses.

FootySimulator is not a device for predicting the outcome of an individual game. There are too many unknown factors - the bounce of the ball, an odd decision by the officials, the length of the grass, the metatarsals of the star players, to make certain the prediction of a one-off game. FootySimulator is a tool that can help the value gambler by providing an assessment of an event should it take place a large number of times. It is a probability generator - not a fortune-teller.

Returning to my initial coin-toss analogy, if you spun your coin twice more, then you would most likely get a different result to the first two spins. Yet if you span the coin one million times more, you would most likely see the same, or virtually the same overall result. The point being that an accurate assessment requires a large amount of simulations.

In a General Election in the United Kingdom, approximately 20 million people will vote. Suppose you were charged with producing an opinion poll which could accurately predict the result. At one extreme, you could ask every person in the country how, or whether, they intended to vote. But this would be a massive task. At the other extreme you could ask ten people in your local pub. This would be quick but would probably be wildly inaccurate. The trick is to find a number in between which, should another sample of the same number of people be taken subsequently, would yield the same result. In practice pollsters usually interview 1000-2000 people and these polls are accurate to around 2%. With FootySimulator, the data becomes more reliable the more simulations you run. I recommend that you should not be prepared to bet on any outcome for a simulation that has been run less than 499 times.