One of the earliest investigations I carried out as a young and enthusiastic racing researcher concerned horses that had won their last start.

The exercise involved more than 2000 instances where the horse in question had run again during the same season (flat racing in the UK) and the most significant finding was that approximately one in four of the qualifiers had gone on to register a second consecutive win.

Now there is nothing even vaguely earth-shattering regarding this finding, but the follow up to this statistical model proved to be of far more interest to my inquisitive nature.

Using the original 2000 races, the result of a further check which involved the winners of two consecutive races was that approximately one in four of the races fell to these "double winners".

They became, in fact, triple winners . . . horses that won three races in a row.

This piece of knowledge could be looked at in the following way: The original 2000 qualifiers had been reduced to 500, of which 125 had proved to be successful in their third race (after winning the other two).

Although the win frequencies had remained roughly the same, the idea of attention being focused on a few (500) rather than many (2000) held great appeal for me.

Over the ensuing years, the "double winners" have always featured prominently throughout the millions of words I have produced regarding the subject of the statistical approach to racing.

What is more to the point is this: They will continue to occupy the same exalted position on my list of most suitable systematic methods.

Profits do not exactly leap from the pages of the sporting press the moment you begin to look at those horses with at least two consecutive wins. However, these horses can be held in high esteem because they provide a safe enough base for a great many more refinements to be carried out.

Some years ago, I completed an in-depth investigation of these "double winners" and divided my analysis into various subsections.

The best strike rate of 31 per cent was achieved by 2yo's with 34 winners from 108 runners, while 3yo's got 54 winners from 194 runners, a strike rate of 28 per cent.

So far as market prominence went, clear favourites had a 45 per cent strike rate with 74 winners from 166 starters.

A check of actual starting prices showed that favourites at odds-on had a 67 per cent strike rate with 29 winners from 43 starters, while "odds against" favourites had 76 winners from 356 starters, a 21 per cent strike rate (well below the average strike for favourites).

The more strongly fancied the "double winner", the more chance there is that it will win.

Now for time lapse (days between races). There was a 33 per cent strike rate for those qualifiers racing within 1 to 7 days with 20 winners from 60 runners. For those starting within 8 to 14 days, the strike rate was 25 per cent with 27 winners from 106 starters, while for qualifiers racing between 15 and 21 days since the last start, the strike rate was 21 per cent with 21 winners from 101 starters.

Lastly, type of race: The stats showed that handicap races provided a 27 per cent win strike, with 67 "double winners" winning again from 248 starters. In non-handicaps, the 38 winners had a strike rate of 25 per cent from 151 races.

Formulating winning ideas need not be dependent upon the extraction of only one profitable line of investigation.

Indeed, that is often the least effective way to go about the production of a winning system.

The best, I have found, is where one "almost profitable idea" can be dovetailed with another of similar standing to form an entirely different method.

This process of reduction will inevitably mean fewer selections will eventually remain but it will also mean a sounder database with a higher win frequency.

* The late Philip Alexander was regarded as the guru of system research in Britain. His work appeared in newspapers, magazines and in books. Philip died in 2003 but PPM  has acquired the rights to produce some of his best work of the last 30 years.

By Phillip Alexander