I agree that track to track times can differ, be it shape of track or the surface, But, as far as class goes I am with Duritz. It all depends on the time.

If a maiden dog runs 31.00 seconds to win a race, and then in a free for all over the same distance, the winner also runs 31.00 seconds. To me the performance is equal, and the class of the race is irrelevant.

About lengths beaten and time, I don't at all agree that 1 length equals .066 seconds. In a short race with the dogs running at 18 m/s 1 length is less time than in a 700 m race when they are going at about 16 m/s. I call 1 length 1 metre.
eg. A dog wins a 520 metre race in 31 seconds. A dog that gets beaten by 10 lengths, I consider to have ran a 510 metre race in 31 seconds. Convert that to m/s and then back to the original 520 metres.....

My 3 criteria are :
The fastest speed in m/s (using the last 3 starts, and converting them to current distance).
The fastest average speed in m/s(3 starts).
The current race distance must be within 10% of the average distance of the last 3 starts.
I will add in a fourth criteria which may appear obvious, every dog in the race must have had at least 3 race starts (not trials).

The reason I do this is because I figure that the dogs singled out will be quick out of the box and less likely to get into trouble. A fall in any start is considered as a 15 length loss. I do not want to back a dog that gets midfield and into trouble.

As far as boxes are concerned, I think the middle boxes are underbacked so will make up for any bias. I always bet to return a figure rather than have a level stake.

The whole point of having a system, is to have it as simple as possible with as little work. While I am still learning about data input from the web, this is as challenging as I feel it needs to be.
I do however treat others ideas with merit, as long as people have the means and the knowledge to put it into practice.