This selection technique is the result of an interest in horses and playing with squared paper. During my misspent student days I came across "minimal pairs" in Linguistics. There the device was used to compare similar language features in pairs, to identify distinctive features.

It gave me the idea of comparing and contrasting each horse in a race with EVERY other horse in that race on EXACTLY THE SAME criteria, to identify significant differences. Much later I discovered this process imitates neural nets.


The minimal pairs method is not exactly quick, but it is systematic, consistent and intensely detailed. Furthermore, pairing horses to compare on the same variables immediately helps identify superior horses.

Minimal pairs can be done by computer, or manually with pencil and paper - even on the back of a matchbox!

Computer aficionados will figure out their own database needs and other adaptations.

In this article I am interested only in outlining a manual method. Minimal pairs uses a lot of subjective judgement, but then really, which process DOESN'T in selecting horses?

Either way, this process produces some surprisingly well-priced winners. What's more you don't need a matchbox; you can do a minimal pairs workout on the back of an old betting ticket!

To do this you'll need comprehensive form data or a formguide with a fair bit of detail, like the Sportsman or similar.

The method is worked on a matrix. Each horse is compared head-to-head with each other horse in the race, as if they are the only pair in the race. The better horse out of each pair, on each criterion, is recorded. The sum of all the rankings is used to decide who should win.

Here's how: Along the top and down the left vertical side of some squared paper write the number of each horse you consider a genuine winning contender within a race. If you have a big field of 13 to 16 I suggest you reduce it judiciously to perhaps the best 10, or the formwork will become laborious. Except for when you pre-reduce the field, in this method, you consider each contender against every other on THREE criteria: -

C - Class
DW - Distance and Weight
F Form - trending/ fitness

If you don't like my criteria you can choose others. For example, if you, have strong leanings towards pace or speed as significant indicators, change one of the three criteria to pace/speed. If you feel strongly about "jockey" include jockey. There was a time when I would have baulked at making such suggestions but I have come to learn that no single variable in racing is so essential that its inclusion is absolute. However, let me explain the choice of the three criteria above. They were not made lightly.

When splitting horses on Class, Distance and weight and form trending/fitness, it is important to put on mind-blinkers and regard each horse as if it is match-racing only the other horse in the pair being compared, and ONLY on that criterion. A decision MUST be made, of which one of EACH pair is the better horse on the attribute being compared.

Class (C).
There is no doubt Class in racing is an elusive thing. Yet it is powerful; if we get it right it tells us about the innate ability of the horse. Class MUST be displayed in a performance or two at some stage of a horse's life. There is no need to be totally scientific in determining comparative class: this is not an exact science. On-balance gut instinct is near enough. If you absolutely can't decide which is the classier beast, split the point half to each, but only as a last resort. At worst you can only be WRONG on this variable, or just be wrong on the day. Remember, no one knows the right answer to which horse is the higher class - that's what they race to decide!

The class of a horse may be determined by using the CLASS OF RACE the two horses have recently been engaged in, the subsequent performances of other horses each has raced against (a good indicator actually), average prize money per start, or other considerations. Don't be tempted to use only CLASS OF RACE though to decide class. Class horses invariably have elements of consistency in their career record to suggest superior ability, BUT ALSO an occasional extraordinary performance. Ultimately, one horse must be placed above the other one being compared on this attribute.

Distance and weight (DW).
For this attribute, use indicators like weight carried and distance(s) won over compared to today's, whether a horse has won at the course and distance, etc., for each horse. If the two compared horses are equal on one element, such as distance, go to weights previously carried to separate them. The reason I have linked weight AND distance is i that some years ago I investigated the relationship between weightcarrymg ability and distances won at, and found a strong connection between the highest carried weight at a specific distance and the ability to perform at the distance again with the same or lower weight, irrespective of quite a few other factors that might influence a race result.

Form – trending/fitness (F).
This attribute is a mixed bag and hard to be exact about. Scour the form closely. In comparing two horses on this it is important to consider, then synthesise, multiple factors. A top trackwork effort, a tr~nding set of placings or small losing margins or lowish average losing margins for two or three runs are legitimate indicators to use. So are number of runs in from a spell, days since last run, whether there is a first-up win record, etc. These should all be part of the thinking mix. Ultimately, one horse must be ranked above the other on this attribute.

OK, so we apply the process using these three variables, to each pair, using a detailed formguide and subjective judgement and we rate one higher than the other. What then? In the head-to-head graphic alongside, we see the result of a seven horse match-up. The horse considered better on the variable has its number entered in the box where the two horse numbers intersect.

When Horses 3 and 5 were compared, for example, Horse 5 was deemed to have better Form-trending/ fitness and distance-getting and weight-carrying ability for the upcoming race, so was indicated for both F and DW ("5f; 5dw"). Horse 3 on the other hand, was ranked higher for Class ("3c") - possibly a classier horse not yet showing full return to form.

Deciding which horse is 'better' requires a lot of close scrutiny of form detail. There will be much turning of pages, checking and rechecking specifics like distance-get, ability in the wet, or whether the barrier draw suits one horse's running style more compared to its rival. It may take an hour or so to analyse a single race sometimes - or as little as 15 minutes. The depth and intensity of the form study is up to you. My results have been best in races that had more time put in. For illustration's sake let's say after doing the work we came up with this:

20051209

The number/letter in each box indicate the horse deemed better, and the variable.

What is it telling us anyway?

Horse 1 ranks higher than an opponent 12 times
Horse 2 ranks higher 11 times
Horse 3 ranks higher 8 times
Horse 4 ranks higher 7 times
Horse 5 ranks higher 11 times
Horse 6 ranks higher 7 times
Horse 7 ranks higher 7 times

This suggests Horses 1, 2 and 5 are the best contenders. However, there is a small fault of the minimal pairs approach that requires a final tweak. Because in each pair one or the other horse has to be picked as the better of that pair, we do end up with some false stars - inflated horses that rack up points because, well, someone had to get the point out of each pair...

We get around this problem by looking more closely at the highest scoring horses. So, in this example, Horses 1, 2, 5 are the most significant contenders because they all scored very similarly - 12, 11 and 11 respective hits, and clearly ahead of the next band of contenders. If we compare how ONLY these three fared against each other now, we see that:

Horse 1 v Horse 2 resulted in 1dw 2c, 2f
Horse 1 v Horse 5 resulted in lc, 5f, 5dw
Horse 5 v Horse 2 resulted in 5f, 2c, 2dw

This shows that Horse I which was highest overall originally on 12, when matched against its two nearest rivals actually ranks behind them both with only two mentions (ldw, lc). Horse 5 which has strong form indicators and fitness has three mentions (5f, 5dw, 5f) without recognition for class. Horse 2 with four mentions (2c, 2f, 2c, 2dw) appears to be the class horse of the trio with two superior C nominations, ability to carry weight and get the distance and current trending form.

The implied finishing order of the trio is therefore:

Horse 2 (placed ahead of a rival four times)
Horse 5 (placed ahead of a rival three times)
Horse 1 (placed ahead of a rival two times).

Whether you accept Horse 2 is the best horse to bet on is up to you. You could consider Dutching all three or take Horse 2 outright with the other two to save. Maybe a trifecta? The staking part I'll leave to you. Whichever way you bet, you can be fairly confident that if you have looked at the form detail of each pair of horses without letting yourself be influenced by favouring or prejudgement, the result of all pairings will be valid, and your top horse will be a very good prospect.

By Tony Acciano

PRACTICAL PUNTING - DECEMBER 2005