I think you will find that ranking the horses win percentage to the others in the same race will give a clearer picture rather that a single percentage on its own.
e.g. 20% for a horse does not mean much on its own if its ranked 10th in the rankings for that race.

One has to try to compare apples with apples where possible.

For example the top ranked win percentage accross all races is approx 20% & it the SR drops slightly as one goes down the list.

If one chases the top ranked place percentage, that is around 22% & is considered to be more accurate that the win percentage.

Then there is the top ranked API (Avereage price) also accurate where majority of winners will come from the top 5 in the rankings.
This also has a 20% SR with low divs.

There is a commercially available system heavily promoted based on combining top win & API together who claim massive POT .
When I ran it through my program its SR was approx 20% for a LOT with consistant very low prices.

These top ranks are usally over bet as well, in other words , are usually shorter in price to what they should be, if betting with the TAB.

If one is going to chase top rank percentage horses , its a strong idea to bet with Betfair so as to attain a stronger price.

A rule I find that seems to inject more value to the final outcome, is to only target the selections that are shown to be $3.20+ in the pre-post newspaper market.
This rule reduced the LOT considerable when applied.

Cheers

__________________
Cheers.