After some thought, I've used the words “test
results” rather than “past
results”.
I started working on UK horse racing towards
the middle of 2003 so clearly
I didn't produce
live results from 1998 onward.
Essentially what I did was to take a model
that I had developed in South
Africa and apply
it to UK & Irish racing, with adaptation to
allow for
jump racing. There are, however, a few
differences between SA and the
UK so some
changes were necessary.
The test results can be interpreted as
follows :
1. Calendar 1998 starts with absolutely no
history and can be ignored.
    I attribute the fact
that top selections made a profit in 1998 to
pure
    chance. Second to fifth selections did
poorly.
2. Calendar 1999 and 2000 were used mostly to
test and adapt the model.
    As such it's probably
best to treat these results with a pinch of
salt. In
    model building jargon, they can be
described as “within sample results”.
3. Calendar years 2001 to 2004 are a better
guide to what can be hoped
    for in the long term
and are “out of sample” or "hold-out" results.
Flat Turf
|
Year |
Races |
Top1 |
Top2 |
Top3 |
Top4 |
Top5 |
ROI % Top Selection
|
Avrg field Size
|
|
1998 |
3314 |
25.4 |
40.8 |
53.2 |
62.9 |
71.4 |
+3.01 |
11.15 |
|
1999 |
3443 |
25.1 |
42.2 |
55.0 |
65.7 |
74.8 |
-0.76 |
11.30 |
|
2000 |
3355 |
25.6 |
42.8 |
56.5 |
67.4 |
74.7 |
+7.76 |
11.46 |
|
2001 |
3415 |
25.6 |
42.5 |
54.8 |
64.8 |
72.9 |
-2.70 |
12.19 |
|
2002 |
3419 |
25.4 |
42.9 |
57.3 |
67.7 |
75.3 |
+4.39 |
11.56 |
|
2003 |
3625 |
26.9 |
43.4 |
57.6 |
67.9 |
75.8 |
-1.18 |
10.99 |
|
2004 |
3690 |
26.0 |
43.0 |
55.6 |
65.7 |
73.6 |
+3.80 |
11.35 |
|
Total |
24261 |
25.7 |
42.5 |
55.8 |
65.9 |
74.1 |
+2.04 |
|