When using or building a method/system, Likelihood Ratios are brilliant at identifying areas of weakness and strength within your plan.

In order to use Likelihood Ratios to best effect, you'll need a set of results around the 500 mark, any less than this and results of filtering are likely to be based on too small a sample.

With 500 results to work on, basic filtering can be carried out, for instance, say you have a system and you would like to research the performance of runners in Handicaps against the performance of runners in Non Handicaps?

This is a basic filter and should provide a decent sample from both Handicaps and Non Handicaps.

If you conduct a filtering process and results from that filtering provides a sample less than 100 it would probably be unsafe to place too much importance from the filtering process and more results would be needed for greater confidence.

Systems are ideal for using Likelihood Ratios but you can also Filter your own results, this is where keeping a record of all your bets and information about those bets is invaluable.

Recording as much information about each bet you make will pay good dividends in the long run.

You want to compare the performance of your method in Handicaps verses Non Handicaps, you have at least 500 results to work with, and the filtering process is performed this way......

(Filter winners x Total Losers) divided by (Filter Losers x Total winners)

A method has given 790 selections of which 258 were winners = 32.6%

258 winners, 532 losers, 258/532 = **0.4849**

Selections in Handicaps gave.....

87 Winners and 233 losers (remember in Filtering we take Total Losers NOT Total Bets)

We then calculate (Filter Wnrs 87 x 532 Total Losers) divided by (Total Winners 258 x 233 Filter Losers)

Written thus (FW 87x532TL)/(TW258x233FL) = **0.77**

We take the earlier figure of 0.4849 and multiply by 0.77 = 0.3733

To convert back to probability we do the following....

Add one to the figure 0.3733 = 1.373 so 0.3733/1.3733 = 0.2718 (27%)

Handicap Selections have an LR of 0.77 and this converts to a probability of 27%

Likelihood Ratios come into there own when combining pieces are information to calculate an accurate percentage for different types of races your method or system has selected.

We can find the percentage chance of a selection running in a Handicap with 15 or more runners in the race, categories for runners could be 2 to 9 runners, 10 to 14 runners and races of 15 runners or more. Calculate the ratios in the same way to find the LR for the different groups and combine the resultant LR with the Handicap LR of 0.77 and convert back to percentage chance.

For example, for the runner groups listed above you have LR's of 2 to 9 runners 1.40, 10 to 14 runners 1.10 and 15 runners or more an LR of 0.65.

You have a selection in a 16 (LR 0.65) runner Handicap (LR0.77)

The calculation is 0.65x0.77x0.4849 = 0.2426 (remember the 0.4849 is your overall winner and loser record for your method)

To convert back to probability add 1 to 0.2426 =1.2426 to get 0.2426/1.2426 = 0.1952 (20%)

On average, your selections win 33% of the time but using the filtering process and applying the LR’s you can identify areas where your method/system has (in this case) a lower strike rate and adjust your betting accordingly if you so wish.

With a little practice, calculating Likelihood Ratios will become second nature as is sure to benefit your betting.

We can see that generally Handicaps are more competitive than Non-Handicaps and that more runners in the race equals more competition for our selection but LR’s can give a more precise picture on which to frame the likely chance of a selection, in this case we can see that the chance has decreased from the normal strike rate of around 33% down to 20% but remember, the bigger the sample size the more accurate the picture becomes.

Return from Likelihood Ratios to Bayes Theorem