Bris Prime Power -Any Software Use it?

[Flysofree]

With all the talk about this figure, I haven't seen any handicapping programs advertized that use it specifically .

[DeltaLover]

Originally Posted by Flysofree

With all the talk about this figure, I haven't seen any handicapping programs advertized that use it specifically .

I do not know about commercial software but in my own algorithms I am using BPP a lot.

[Capper Al]

Originally Posted by Flysofree

With all the talk about this figure, I haven't seen any handicapping programs advertized that use it specifically .

If they did use it, they wouldn't tell anyone. Some software claim to make their own prime with a similar hit ratios, but better ROI since it's private.

[DeltaLover]

Originally Posted by Capper Al

If they did use it, they wouldn't tell anyone. Some software claim to make their own prime with a similar hit ratios, but better ROI since it's private.

Doable, but why complicate things when you have such a consistent figure as the BPP for minimal cost and for any track you like?

[Dave Schwartz]

I have looked at the BPP via someone else's database. It is comparable (perhaps a little weaker) to HDW's PSR.

My statistical experience is that both get lots of winners but neither has much true value to the handicapper because they correlate with the the tote board too closely.

Example: Top-ranked horse will be 1st or 2nd favorite around 83% of the time.

[DeltaLover]

Originally Posted by Dave Schwartz

I have looked at the BPP via someone else's database. It is comparable (perhaps a little weaker) to HDW's PSR.

My statistical experience is that both get lots of winners but neither has much true value to the handicapper because they correlate with the the tote board too closely.

Example: Top-ranked horse will be 1st or 2nd favorite around 83% of the time.

Sure, you cannot show a profit by blindly betting the top bpp, but having such a consistent and accurate figure can be useful in several aspects of building a model.

[Capper Al]

Originally Posted by Dave Schwartz

I have looked at the BPP via someone else's database. It is comparable (perhaps a little weaker) to HDW's PSR.

My statistical experience is that both get lots of winners but neither has much true value to the handicapper because they correlate with the the tote board too closely.

Example: Top-ranked horse will be 1st or 2nd favorite around 83% of the time.

Maybe 63% of the time. 83% sounds high.

[MitchS]

Originally Posted by DeltaLover

I do not know about commercial software but in my own algorithms I am using BPP a lot.

Absolutely!... And I'm sure there are commercial software applications that use BPP in conjunction with other numerical ratings. % of BPP + % of other fig equals +ROI. I could mention a couple, but then I'd have to kill you

[Capper Al]

Originally Posted by Capper Al

Maybe 63% of the time. 83% sounds high.

My apology, Dave. You said it will be the favorite not won or place.

[Elliott Sidewater]

I don't have actual stats to back this up, but when I was playing a lot more often than I am now, the Prime Power figures performed worse in turf races than on dirt. I agree with Dave Schwartz's statement about correlation between PP rank and actual betting favorites and second choices, but when the top ranked PP horse (a big betting favorite) has a big PP score advantage over the second choice and is a negative or suspicious drop, and you're right in that assessment, the second and third PP horses have value.

[MJC922]

Anytime we're studying a ranking a rating or whatever it may be, if it's upwards of 30% wins it seems to me it would be a mathematical certainty that it would correlate strongly with the betting favorite. What matters more IMO is what it's a ranking of, i.e. if it's a 30% ranking which doesn't take the pace of today's race into consideration or the physicality, or any other sometimes relevant factor which appears after the horse has crossed the finish line of it's last race, then I think there's a lot of profit potential there.

My current thoughts on the matter are the data points that are known, let's say all of those which appear in the ink up until the finish of a horse's last race represent practically all of what is predictive, the rest of the factors have predictive value which is overall minor but this is only because it's situational. The 'value' is all concentrated in the situational, at the same time I believe you still need the synergies created from pairing with the predictive.

[Robert Goren]

At one time, I looked at running a multi correlation with PP as one of the factors. Like using the odds as factor in a multi correlation, it is very had to find anything that improves it much. But maybe you can.

[MJC922]

Originally Posted by Robert Goren

At one time, I looked at running a multi correlation with PP as one of the factors. Like using the odds as factor in a multi correlation, it is very had to find anything that improves it much. But maybe you can.

That makes sense because that last 5 percent or whatever is left to make this any more predictive is comprised of a bucket of many situational factors. I spent extensive time several years ago generating these types of composite ratings and I've found that once you get to about 31% you're not going to move the needle by much more than 1% no matter what else you throw into it. So for example you're at 31% and you add an early pace figure into the mix and the early pace figure has some predictive relevance sometimes so it will move the needle maybe .5-1%. However I tend to believe you reach a point where this becomes entirely counterproductive, because the value (if there is any in a pace figure) is when it points out abnormal pace scenarios. That is, it's more valuable in certain situations, unlike ability / class which is always going to be predictive and relevant.

[DeltaLover]

Originally Posted by MJC922

That makes sense because that last 5 percent or whatever is left to make this any more predictive is comprised of a bucket of many situational factors. I spent extensive time several years ago generating these types of composite ratings and I've found that once you get to about 31% you're not going to move the needle by much more than 1% no matter what else you throw into it. So for example you're at 31% and you add an early pace figure into the mix and the early pace figure has some predictive relevance sometimes so it will move the needle maybe .5-1%. However I tend to believe you reach a point where this becomes entirely counterproductive, because the value (if there is any in a pace figure) is when it points out abnormal pace scenarios. That is, it's more valuable in certain situations, unlike ability / class which is always going to be predictive and relevant.

It is absolutely possible to increase the predictability of the BPP. Doing so, is one of the cornerstones in my models.

[MJC922]

Originally Posted by DeltaLover

It is absolutely possible to increase the predictability of the BPP. Doing so, is one of the cornerstones in my models.

Making sure we're on the same page then, out of curiosity how much have you increased it and have you confirmed this holds when forward-tested? Back-tested let's not go there, we can fit phases of the moon using an evolutionary algorithm and probably bring it up to 35%.