Understanding probability

Consider six people sitting around a table, with their chairs numbered from 1 to 6.

A fair six-sided die is rolled.

Each person has one number, so each has a 1 in 6 chance of their number appearing on any individual roll.

It is tempting to imagine that, after six rolls, every number should have appeared once.

That is not how probability works.

Number 2 might not appear for 20 rolls. It could then appear three times in succession.

Both are perfectly possible.

Most importantly, after number 2 has failed to appear for 20 rolls, its probability on the next roll is still 1 in 6.

The die does not remember what happened previously.

Probability describes tendencies, not schedules

Over a very large number of rolls, we would expect the proportion of each number to move towards approximately 1 in 6.

That does not mean the results will be evenly distributed along the way.

Over 10, 20 or 50 rolls, the results can look very different from the long-term probability.

There can be long sequences where a particular number does not appear at all, followed by several appearances close together.

This variation is normal.

What this means for horse-racing statistics

The same principle matters when using the historical statistics shown throughout HRBC.

Suppose a particular group of historically comparable runners has won 20% of its races.

That means approximately 20% of the runners in that historical sample won.

It does not mean:

• one of every five future runners will win;
• a winner is due after four losers;
• five consecutive losers make the next runner more likely to win;
• the next five runners will contain exactly one winner; or
• a historical 20% win rate guarantees anything about the next race.

You could have a long sequence without a winner and then several winners close together while still producing approximately the same percentage over a sufficiently large sample.

Historical evidence is not a prediction

HRBC uses historical racing information to help you research and compare runners.

Statistics can show what happened previously when particular conditions, rankings or characteristics were present.

They cannot tell us with certainty what will happen in an individual future race.

A percentage based on thousands of historical runners may be statistically interesting while still telling us very little about exactly when the next winner will occur.

The size of the sample also matters. A percentage calculated from 20 historical runners should not automatically be treated with the same confidence as one calculated from thousands.

Losing and winning sequences are normal

Random variation produces sequences.

A run of losers does not necessarily mean that a historical statistic has suddenly stopped being meaningful.

Likewise, a run of winners does not prove that a particular angle has suddenly become exceptionally powerful.

Short-term results can move substantially in either direction.

This is one reason HRBC provides different historical periods and, depending on your package, different ways of examining the underlying evidence.

The purpose is to give you information to investigate — not to turn historical percentages into promises about future results.

Use probability as evidence, not certainty

When looking at HRBC statistics, ask:

• How large is the historical sample?
• What percentage of comparable runners won or placed?
• Over what period was the statistic measured?
• Are the historical conditions genuinely relevant to today's runner?
• Could the apparent pattern simply be short-term variation?
• What other evidence is available for this race?

No individual statistic should be treated as a guarantee of the result of a race.

We rank the runners. We give you the evidence. You make the decision.