beyond complexity ?
 

Say I have a race positional running line as follows (an actual 1400M race @ Sha Tin):

8 8 8 1
7 7 4 2
1 3 1 3
9 9 9 4
12 12 10 5
6 6 5 6
11 11 12 7
4 4 2 8
2 2 6 9
3 5 3 10
10 10 11 11
5 1 7 12

I can measure the 'complexity' of the race (the level of positional changes) using a formula but it won't 'reveal' that the horse that finished 3rd ran the best race given the setup. A huge race, in fact.

Obviously, I can look at the chart for the race and see this. And more broadly, that pace horses did not fare well.

But perhaps there's a way to automate this and not have to do as much work. Or maybe train a model to do the work for me.

Any ideas?

How can the positional line of the horses mean anything without referencing the pace fractions of the race?

Shockingly, there is correlation to future performance even with no reference to the ratings involved.

Time-decayed positional gain/loss is not as good as lengths gained or lost, but they all have value.

Years ago I bought an expensive black hardcover book written by a Canadian "pro horseplayer", whose name eludes me at the moment (Jayce Killian?). I think he proposed a method which dealt with the horses' positional moves without taking the figures/ratings into account. Perhaps I should have paid more attention to the book than I did...

It's not enough to win, so I'd say not.

And the calcs are monstrous for someone using paper & pencil.

But your point is not too far off in the sense that doing what everyone else is doing isn't enough.

I used to think that meant "do something different."

Now I believe it means "do more." That is, add another step of analysis.




.

Have you looked at this in terms of surface (and field size)?

Any of the ML algorithms able to do something like this? Would Random Forest spit out useful categories if passed vectors?

Not a fan of random forests.
Yes, my new AI uses it for one of its handicapping approaches known as "Best Horse."

Best Horse - in its simplest form, is just what it sounds like. It assigns a probability to each horse based upon using the factors that push the most winners to the top.

However, there is a second version of it that breaks it down by odds ranges:

• <2/1
• <3/1
• <9/2
• 3/1 up
• 4/1 up
• 8/1 up
• 17/1 up

I've been able to look at what it typically uses in each category - and it varies widely. The lower odds ranges are dominated by what we would mainstream factors.

But as the odds get longer, the standard stuff just won't put us on the winners - because it already puts us on the favs.

For example, in the 17/1 and up range it is almost always centered around gains between calls, some obscure form factor (like times on track in last 30 days), or other esoteric approach.

BTW, I should be ready on the weekend to begin building demos of the deTerminator.

exactly!!!

I'd say horses are more lkely to run back to their typically preferred running styles and positions than their prior fractions. The fractions vary depending on the makeup of the field and how many horses are being used (and how hard) to get their preferred position.

This might be true, but it’s not what the original poster is looking for. He wants us to come up with the horse with the most impressive performance, simply by looking at the field’s positional running line.

You can certainly flag anomalous. Anyone who is in fact able to make money at this game is surely a master of it and in several areas.

Edit: You asked how, well for me in this example I'd use a custom function in VBA / Excel. I suspect you're liking a few things here, you might like that the horse finished on the board, was the only one in the first flight that was on the board as all of the other first flight horses finished in the rear half, the horse also lost the lead and regained it. You could even add the odds in as another input/parameter because maybe the first flight contained lower odds horses. That might strengthen some early pace assumptions a little anyway. No reason why you can't include times, vs par times and process all tracks to solidify it even more, even if it's just for replay follow-ups IMO, it doesn't necessarily have to be an automatic next out spot play or something like that.

In the example given, if the pace was very fast, would that enchance the 3rd place finisher and downgrade the winner?

Or if the pasce was very slow, just the opposite.

Got it.

You also need some objective measurement of the strength of the field to go along with it. You can either look fractions and final time or the quality of the horses in the race (comparative class) and compare similar trips within the race, but just the positions and running lines are not enough even if they have some predictive ability. Ideally you'd be looking for horses that ran well against the general flow of the race.

We also need some objective statistics that show our chosen approach has validity.

"I think" does not necessarily equate to reality.

Finally, some rational thinking about a nonsensical approach to find some sort of plausible meaning by just comparing past running lines.

These type of questionable parameters just prove one thing: How far will someone go to create an argument with Logical Fallacies. By definition: Are deceptive or false arguments that may seem stronger than they actually are due to psychological persuasion but are proven wrong with reasoning and further examination. These mistakes in reasoning typically consist of an argument and a premise that does not support the conclusion.