Seriously? $150 horse?
 

PA,

Seriously? $150 horse?


Your system never ceases to amaze me.

:ThmbUp::ThmbUp:

Yeah, but I needed that...I was on the verge of being even ROI-wise at that point after about 180 races posted...

Frankly, I'm disappointed I haven't hit more mega payoffs...during testing, they seem to come up more often then they have lately...

Enough to pay for 74 consecutive losers! :eek:

Trust me...I'll probably need every one of them.

Congrats....Had a couple bucks on horse. Was top contender.

Those are pretty much few and far between nowadays and less tempting, because of very low hit rate. Still can lose 75 in a row before you give it all back. Has to look really good to play those.

The "top contender"...at odds of 74-1. And people say that this is a tough game to beat. :confused:

It is still tough. Not something I would want to count on for steady winners. More of a contest horse :headbanger:

Let them pay $150...and I don't care if they are "unsteady".

long shots over 50-1 per year
2017 117
2015 118

2016 139

2011 181


seem to be fairly steady recently

PA,

I am still astounded at how quickly you adapted to HSH.

I mean, it is one thing to get up and running but to be this competitive in the short time you've been using it is... astounding.

Did you come in with an idea already in place?

Without giving away your secrets, could you just share a little of the development process you used?


Dave

Without that $150 winner, I'd be barely ahead after 195 races on this current run.

So let's not get too excited.

With that said, my whole point of wagering these days is to HIT longshots. My average payoff before this $150 horse was over $20. So all I am betting, basically, is longshots. That's the approach I came to HSH with.

To try and effectively "hunt" for longshots.

This incorporates both looking at races "like this one" that have run in the recent past and seeing if they tend to pay off big. And also to look at the upcoming race itself and see what horse HSH comes up with (based on how I have set it up to handicap a race) and if that horse is projected to pay a decent price.

With this most recent run, I have leaned toward the latter and discounted the former...totally. Meaning, currently, I am not paying attention to how races "like this one" have paid off in the past...and am instead just concentrating on what THIS RACE looks like it will pay off. On my prior, profitable 500 race run, I discounted the latter...totally...and went with the former. Meaning I wouldn't bet a race if past races "like this one" tended not to pay off big.

If I am presented with a top choice that HSH projects to be a short price (one of the top three favorites on HSH's line), then I pass the race. Otherwise, I'll bet it, if it's an overlay on the HSH line.

And by top choice, it doesn't necessarily have to be the top probability horse. But it does have to be the top optimal bet.

After you run through a million or so races you will get a good idea of what it takes

These are absolutely great concepts to live by.

I especially like the "tone" you promote with this. There is a certain conviction. There is also a sense of personal responsibility coming through.

Fantastic stuff.

:ThmbUp::ThmbUp:

Looking for races "like this one" is certainly nothing new and this is what most of the handicappers are trying to do (consciously or not).

In problems like a horse race, where you have to deal with non structured data associated with a lot of noise and incompleteness discovering "similarity" can very well represent an impossible to solve problem.

The dimensionality of the problem is so high that a sample of 500 (or I dare to say 50,000) races is extremely limited to derive any kind of a meaningful result and conclusion.

In machine learning , looking for races "like this one", is called clusterization and is implemented using metrics like Euclidean or Manhattan distance and unsupervised learning algorithms like K-means to discover cluster centroids.

Although this approach might work great in linear separable data and in many cases scale well to cover more complicated cases trying to scale it beyond a certain level of complexity results to an impossible to solve problem. I have valid reasons to believe that trying to discover similarities when it comes to horse racing falls in the latter category!

Having said these, I would like to hear more about the "hows" of the solution to the problem; an example (not necessary for horse racing) and a description of the algorithms used used might be helpful and spread some light to what you are saying here.

First you fart in his general direction, then ask him how he does it. Funny stuff.