That is very impressive, thanks Dave.

Question: The A/A+ grades were based on the final tote odds line, correct?

I am interested in this also.

@Atakante I am sure Dave will be weighing in here shortly, but in the meantime:

Yes, The Neal uses the real time tote odds, which are of course a better estimate of the final tote odds than the AI Line is, so are better for estimating real time $Net value, I agree.

Yes, it would be fantastic to have an idea of how well the AI Line and AI BH do at predicting final odds and win percentage respectively. In effect we are each responsible for doing our own research on how it does for different scenarios, if I understand Dave correctly.

Assuming AI Best Horse Pct gives a reasonably accurate estimation of a horse's win probability, and that the AI Line is some kind of estimate of the final tote odds, the grades are the most important thing of all, in my opinion.

As you may know, Grades are just an expression of the $Net. For example, A+ means the $Net is greater than $2.40, where $Net = $2.00 is break even.

I have noticed that $Net is a function of AI Line (AILineOdds) and AI Best Horse Pct (AIBHPct) according to the usual formula for ROI:

$1ROI = Odds * Probability - (1 - Probability)

where $1ROI is the Return On Investment for each $1 bet ($1ROI = 0.0 is break even).

in DET
$1ROI = AILineOdds * AIBHPct - (1 - AIBHPct)

For example for AILineOdds = 0.34 and AIBHPct = 36.2% = 0.362,

$1 ROI = 0.34 * .362 - (1-.362) = -0.51

To convert $1 ROI to $Net,
$Net = 2 * $1ROI + $2

In this example
$Net = $2 * (-0.51) + $2 = $0.97.

My philosophy is to never bet horse to win that have $Net less than $2.00. Assuming the AI Line and AI Best Horse Pct are accurate, betting such horses will always lead to losses in the long run. I focus on horses with $Net > $2.40, i.e., A+ horses.

<< I believe the Grades (A+, D, etc.) were initially portrayed as ways to find who the whales are likely to bet>>

I am not sure that is correct. My understanding is the AI Line is how the public (mostly whales) should bet, whereas the AI BH Pct is how the AI predicts the win probability for each horse. $Net/Grades are a function of both of these.

I see it mainly as a way of identifying that "somebody (whale or not) might know something and therefore I have to consider the horse

Ok, fair enough, but I don't think it's the whales, since Dave says they bet it all when they are loading in the gate.
To me it's mostly the horsemen's money, though I guess some is from sharp handicappers.

I have a question @Neal Freedman...

Is this primarily intended to measure how much insiders (not including whales) are betting a horse down?

Thanks for your response, @Dave Schwartz

The AI Line is literally how the entire public should bet the horses based upon past history.

Ok.
I am guessing that what you are asking the AI to do, is to create an odds line based on all relevant historical data factors, that estimates what the final odds for be for this horse and race, on average.

It's not a prediction of what the actual odds will be, since that depends on what nuances the whales see in this race - what they see as the overlays based on their unique knowledge of the actual odds milliseconds before the tote closes.

That is, the AI Line is your AI's a priori best estimate of the final odds.

That sound right?

Tom

Neal, when are you recording tote odds, given that they change so much after post time these days?

[edit: DS] Adding a ping of @Neal Freedman so he'll more likely see it.

Newbie here, looking forward to having fun working with you all.

Dave, what plans do you have for deTerminator documentation in general?