Databases for those who don't want to do it themselves.

[acorn54]

just got cx wong's book "precision". mr wong is a member of a syndicate in hong kong. very interesting and enlightening book. he has in, one of the last chapters, how the typical syndicate is structured. there is a statistical modeler, ( no pre-packaged software from vendors), and they have a trip handicapper, and an expert of horse body language, and an "insider", (yes, there are betting coups). the syndicates use all of this info.
i highly recommend this book, worth every penny.

[DeltaLover]

Originally Posted by acorn54

just got cx wong's book "precision". mr wong is a member of a syndicate in hong kong. very interesting and enlightening book. he has in, one of the last chapters, how the typical syndicate is structured. there is a statistical modeler, ( no pre-packaged software from vendors), and they have a trip handicapper, and an expert of horse body language, and an "insider", (yes, there are betting coups). the syndicates use all of this info.
i highly recommend this book, worth every penny.

I find it terribly antiquated by today's standards.

[acorn54]

Originally Posted by DeltaLover

I find it terribly antiquated by today's standards.

i would be very interested if you would elaborate on a more precise method of placing a probability on a horse winning than using multinominal logistic regression,in combination with factoring horses trips, and physicality in pre-race appearance.

[DeltaLover]

Originally Posted by acorn54

i would be very interested if you would elaborate on a more precise method of placing a probability on a horse winning than using multinominal logistic regression,in combination with factoring horses trips, and physicality in pre-race appearance.

Start here:

https://en.wikipedia.org/wiki/Bayesian_network
https://arxiv.org/abs/1603.04467
http://ufldl.stanford.edu/tutorial/s...tmaxRegression
https://en.wikipedia.org/wiki/Softmax_function

Some books you might find helpful:

https://www.amazon.com/Artificial-In...l+intelligence

https://www.amazon.com/Machine-Learn...ng+theodoridis

https://www.amazon.com/Hands-Machine...ing+tensorflow

Again the book you are talking about (and Benter's approach in general which is based in multinomial logit) is good only as an introduction.

[acorn54]

delta thanks for the lincks.

[DeltaLover]

You need to understand that logistic regression is a special case of a more general non-linear paradigm (neural network) which can solve problems that are impossible to be solved by linear relations. This does not mean that NN is the silver bullet as linear approaches are still used heavily to solve many problems; what is important though is to have the knowledge and experience to apply the most applicable algorithm for each case.

To get a simplified and intuitive understanding of the multinomial logit you can think that the initial assumption is that the ability of each horse (for example expressed in speed figures) is following a normally distributed curve whose mean and sigma depends on the features of each individual horse.

For example, for two horses that are facing each other using the logit method you can imagine that each of them is assigned a mean speed figure and a sigma. Let’s say that these values are as follows:

A: mean 98 sigma 6
B: mean 102 sigma 12

If you plot both of them in the same graph, the probability of A beating B equals the probability of A running a larger figure that B and vice versa:



Note that in this example, although the 102/12 is faster than the 98/6, you can visualize that the latter is still winning many times (when the blue are is higher than the orange).

Try to understand how this can be extracted from the graph and think of why this approach might be wrong and you will see that this approach has a lot of room for improvement.

[acorn54]

i understand that multinominal logistic regression is not the silver bullet for paths to profits. the 64k question is when to know what model to use for best evaluation.

[DeltaLover]

Originally Posted by acorn54

i understand that multinominal logistic regression is not the silver bullet for paths to profits. the 64k question is when to know what model to use for best evaluation.

There is no analytical methodology to decide "A priori" whether logistic regression will be sufficient for your problem or not. Usually the only way is to try several different approaches and pick the best among them (or create an "ensemble" of the most promising solutions and use all of them based on some type of a voting mechanism. There exist problems that fit very well in linear (or logistic) regression and there are others that are impossible to be solved without the introduction of "hidden" processing layers as it happens in Deep Learning NN for example. An example of the former, is the creation of a par-times model that predicts the final time of a race based on factors like distance, classification, age and sex while a typical example of the latter can be found in the solution of the XOR operator, something that is impossible to achieve using logistic regression without the addition of a hidden layer (perceptron).

[Mdnspecialist]

Originally Posted by Partsnut

My choice would be BetMix.
It is intuitive, insightful and accurate.
It is probably one of the best data mining tools out there and offers good value at the price. You can look at a multitude of tracks per day and selectively pick the races that appeal to you. They offer lots of handicapping tools.The support and video tutorials that are second to none and you can easily get past the learning curve.

I've been using betmix for about three months now. Angler is what hooked me. It took some time to figure it out. Still learning. I'll work on an angle then take a break. It's very time consuming. I created an angle last month that showed a profit of $4,015 from January 1st, 2013 through October 15th, 2017. Profitable every year. It's 493 winners out of 2609 races which is 18.89%. It shows all the races associated with the angle you've created. There are some mind blowing losing streaks that would bring most handicappers to their knees. Still tinkering with it, though. Going to try and bring the win percentage up. You do that by eliminating more losing races than winners, obviously. And at the same time, not eliminate too many long priced winners. Because this is a long shot angle for sure. It probably took me three weeks to create this angle. I'd say a 100 plus hours. It currently is on one of those losing streaks. 2 for it's last 28..-$37.40. There has been worse losing streaks as I look over all the races. 37 straight losses. 2 for 42. But it always seems to pull out of it. A 2,000 unit profitable angle is something that you just don't see. What I'll do is build an angle with as many races as possible and breaking even. Then start to break it down. This angle was 951/6276 15.2% win rate and -$243. Then slowly, I eliminated races. I think it's possible I can create one that shows even more of a profit. I'll wait another month to see if this angle pulls itself out of the current losing streak. Based on the past 57 months. I just don't see why it won't.