# How to spot errors and inaccuracies in prediction models

#### Previous articles have introduced different types of prediction models as well as methods to measure dispersion. Whatever the model, the prediction is not prophecy, it assigns a likelihood to an event, which can be influenced by error, something bettors should be aware of, and eager to exploit.

As an allegory, consider one of those shape sorting cubes toddlers play with – the kind in which we need to pass the correct shape in a hole. The correct shape represents a correct prediction, but unlike the cubes toddlers have at hand; the number of possibilities are far larger.

### Choosing the wrong shape/model

The first possible error to make is to fit in the wrong shape altogether. A triangle might fit in a square if you really try hard or maybe if it’s much smaller, this still doesn’t make it a good fit.

This is equivalent to using the wrong model for the purpose. For example, while the normal distribution seems to be a good fit to goal differences, it may not be the best predictor for home team goals scored. As demonstrated by the graph below, showing the actual and estimated number of home goals scored using the normal distributions for the 2013/14 English Premiership (using post-ante data).

With a multitude of potential models possible, the perfect fit may not be in use or, worse still, may not be available. The model is a necessary simplification of a real life scenario and is bound to be incorrect. Ways to diminish this is to apply judgement in selecting and interpreting a model as well as fitting the model to old data.

### Choosing the wrong size/parameter

Going back to the shape sorting cube analogy, the correct shape might be picked but of a different size. For example, the wrong size square is being used.

In a model building scenario, this is equivalent to using the wrong parameters. Imagine you are trying to calculate the likely number of goals scored in a certain match. The Poisson Distribution may be the correct model to use but one of the teams has recently had an 8-0 win. This, in turn, distorts the mean number of goals scored rendering this parameter useless.

In this case, a judgement needs to be used and more attention needs to be given to the standard deviation in the parameters used.

### The process error

Finally, the correct shape and size may have been picked in the shape sorting cube example, but each shape may fluctuate in size due to wear-and-tear and slight differences in production.

In a sports prediction environment, not every outcome is replicable. If this year’s Super Bowl final was replayed under the same conditions many times, the result would not always be the New England Patriots winning 28 to 24, due to natural fluctuations.

Albeit, having picked the correct model and parameters, there is always natural volatility in results (that can be measured). The best predictions are available when more relevant data is available – hence why a lot of bettors find it easier to predict an English Premier League soccer match than a World Cup match.

### Conclusion

Betting companies, syndicates and private individuals all have errors in their predictions – the skill is in applying judgement to take advantage of the errors posed by other parties.

In addition to searching for valuable information and advice, bettors should always look for the best odds, which Pinnacle Sports offer – read here about our unbeatable odds – and given we offer the highest limits online, a bet with us will ensure you are best placed to get the best value for your bet.

### MORE: TOP 20 Bookmakers that accept Cryptocurrency >>>

Source: pinnacle.com