Hi Kenny Victor,

Yes a large database helps in formulating systems. I never would disagree with that. But winning systems not necessarily. You can make curves about known facts like your above example and they are useful. It's the unknown ones that become the problem. We get into probability here and chance. The greater the sample of unknown 'events' rather than facts the greater the error in interpreting them correctly to be of future use.

Lets look at a probability example:

In any case of probabilities of an event, the individual probabilities of the situation must add up to one. However, in horse racing the odds often add up to more than one. How come?

Let me first clarify the difference between probability and odds, just to be sure we're using the same terminology. The probability of an event is:

(Chances for)
P(x) = ---------------
(Total chances)

So, for example, the probability of drawing an ace in a single deck of 52 cards is 4/52 = 1/13 (or about 0.077 = 7.7%).

Odds, on the other hand, are given as:

(Chances for) : (Chances against)

Of course, (Total chances) = (Chances for) + (Chances against), so we can determine (Chances against) as (Total chances) - (Chances for).

When the odds are converted to probabilities, they usually add up to more than 1 to give the house or bookies their "edge" - that's how the Casinos and bookies make their money.

Let's take a simple example. Suppose we have a series of 12 races with 4 horses, given the following betting odds (for simplicity, we'll only consider the odds of winning each race):

Horse Odds Probability (from odds)
----- ---- -----------------------
Horse A 1:1 1/(1+1) = 1/2 = 6/12
Horse B 2:1 1/(1+2) = 1/3 = 4/12
Horse C 3:1 1/(1+3) = 1/4 = 3/12
Horse D 5:1 1/(1+5) = 1/6 = 2/12
-----
Total Probability = 15/12 > 1

Now suppose I were to bet $1 on each horse for each race. In order for me to break even on each horse, horse A would have to win 6 of the 12 races - then I'd win +$6 on the races A won and lose -$6 on the races A lost. Similarly, horse B would have to win 4 of the 12 races for me to break even -- I'd win 2 * $4 = +$8 (because of the 2:1 odds) on the wins but lose -$8 on the losses. Horse C would have to win 3 of the 12 races (3*$3 = +$9 on the wins, -$9 on the losses), and horse D would have to win 2 of the 12 races (2*$5 = +$10 on the wins, -$10 on the losses). Of course, this means that all together, they have to have 15 wins in 12 races, so somewhere they're going to fall 3 short of my "break even" requirement.

If, for example, horse A only wins 5 races and horse C only wins 2 races, then I've lost -$2 on horse A (+$5, -$7) and -$6 on horse C (2*$2 = +$4, $10).The bookie has just collected $8 from my pocket.

As long as no horse wins more often than its "probability" (based on odds), the bookie wins. Of course, it is possible that horses D and B will win 4 races each, horse B will win 3 races, and horse A will only win 1 race. In this case, I will lose -$10 (+$1, -$11) on horse A, break even on horses B and C, and win +$12 (4*$5 = +$20, -$8) on horse D for a net winning of $2 - But you can "bet" that that won't happen too often. ;-)

Systems have a notorious lack of selecting their odds that are in their favor, and very few punters really understand them anyway. So big problem number 1. Next, with relying on data that includes what, 30 to 40% max. known facts [like your example] and up to 60% or even more, unknown facts collected from thousands of races, you are really up the creek because the more the unknown facts collected the greater the chances of the conclusions being incorrect for future prediction.

Hope this helps, but I really don't want to get into this crazy system business.
Zoe