KV,

ok to avoid confusion I did the following :

Took some data I had stored from Dec - Jan and essentially backfitted the results so that it only had horses which had won or placed. Any horse which didn't at least place I found a reason to exclude ( hence backfitting ).

I then checked it over the previous 10 days and found it had 10 selections for 10 Winners. But I didn't test these real time ( but they were not part of the data which was used for backfitting).

I haven't bothered going back further.

The backfitted results were :

Races : 45 ( 2 months of data)
Wins : 30
Places : 45
WIN ROT : 183%
PLACE ROT : 153%

I don't believe the above figures are worth anything except to show the system might have some potential. Hence I only put up the last 10 days worth as these were not part of the backfitted data.

The first "live" selection where I actually picked the winner before the race was the first one I posted here. Regardless of that the system had selected 11 winners in a row.

Therefore new stats are (with those 11 winners and a second) :

Races : 57
Wins : 41
Places : 57

Hope that clears it up.