To clarify my contribution, stats on 20,000 races might show that, for example, last start winners from barriers 1-3 are good things. And of course there is a certain predictability to this that you can rely on. But every race is a unique event all the same, and while such runners might be good things statistically they might not be in any given race. That's my point.

The question to ask about any stat is: in what context? It is the context that is unique. You can narrow it down but the best instrument for accounting for the full context is human judgement, in the end. I reckon.

My main point is that any mechanical system can be improved by the application of wise judgement, especially judging which races to shun.

At least I see it that way because most mechanical systems I devise either fall over or break even eventually and the best way to push them into the black is not to throw more stats and filters at it but to apply some judgement. Departing from your "system" you might decide, as the race shapes up, scratchings are made and real money is laid down and the context changes, to increase a bet. Or pass the race. These deviations from the rules are where I make the money that pushes the mechanical system into profit.

I regard mechanical systems as guides. Most of them break even at best, long term. For me. There comes a point in the development of a system where throwing more statistical power at it won't achieve much more than drastically reducing the volume of bets. That's the point at which I think its best to drop the science and apply some well-honed human judgement which is best able to comprehend and synthesize the unique context in which the mechanically chosen selection is situated.

Hermes