![]() This delivers a numerical assessment of the significance of each variable to the final prediction. The calculations also provide a second piece of information, called a T value. The final predicted return for the future race is an aggregate of all the 50,000 influences. Others are either somewhere in between or a mixture (all over the place). If it large (far away) it is given a correspondingly weak influence. If the difference is small (close) the historical result is given a strong mathematical influence on the predicted return. A numerical value is calculated that defines the difference between the two tracks. Once the track of the future race is in place, the track of each of the 50,000 historical races is individually compared to the target track. Its a kind of barcode for the future race we are attempting to assess. The single beads mark out a distinctive track across the variables. Each wire represents a variable, and is scaled accordingly (number, distance, odds etc). Imagine an abacus with only one bead on each wire (above). This is the result and it should become clear why I have called it the abacus bar code model. Choosing a model to illustrate the principle without using too much maths has been an interesting exercise. For this reason the only way it can be done is by a computer doing the tedious bits, and delivering outcomes. Multiple regression analysis is not so much mathematically complex as mathematically tedious and time consuming. That method is multiple regression analysis. What we need is method that doesn’t do this. What makes this method useless is the increase in power with the addition of the each new variable (10, then 10 squared, then cubed and so on). You might think that we should give up at this point, but not so. Accuracy is now completely out the window and we still haven’t got to £1. Second we now have 10 quadrupled (10,000) locations, with only 5 results per location. First we can no longer visualise the result, and have to rely on mathematical expression. The margion of error now is ☑4% and therefore we will have much less confidence in the results.Īdding a fourth variable has two dreadful effects. Even worse we now have 10 cubed (1,000) locations for our 50,000 historical results which is only 50 per location. ![]() Once again we have made a little progress towards £1, but still not enough. ![]() It can be thought of as the distorted layers of an onion where each layer represents a specific return on investment. This time we can only visualise the three way relationship with a three dimensional object. So now we need to add a third variable, say the odds of the favourite. A good and simple start is to look at just one variable lets say the number of horses. Ten years worth would give us data on 50,000 races. In order to study these variables we would need to have at our back a large number of historical racing results that include information about the variables. The variables in horse racing include the number of horses, the distance, the odds, the historical performance of both horse and rider, the weather, the condition of the ground and many more. Variables interact and the more variables there are the more complex those interactions. Horse racing is like the weather, in that it has a large number of variables, making it difficult to make predictions. But how do you try to make that distinction? It’s a bit like picking the best stocks and shares to buy. The only way to get around this is to avoid betting on races that are likely to give you the least return and concentrate on those that are likely to give you the best return. This means that if you put £1 on 5,000 races in a year, you would spend £5,000 and get about £4,250 back, hardly satisfactory. Bookmakers set the odds for races so that they can top-slice a profit, usually around 15%. The first thing you notice is that you don’t make a profit, and the reason for this is straightforward. ![]() Imagine too that you want to gamble on the horses, and to keep things simple you will just put a small bet, say £1 to win on the favourite of every flat race in the UK (that’s about 5,000 races a year). Only imagine mind, in the interests of this thought experiment.
0 Comments
Leave a Reply. |