How to determine a track bias

[Robert Fischer]

small samples aren't going to be 'statistically significant'


that doesn't mean that a skilled player shouldn't use them

[classhandicapper]

Originally Posted by dnlgfnk

I wasn't unveiling Benter to the uninitiated, but asking "Why is the criteria for computer modeling (large data samples) suddenly abandoned when it comes to especially daily, but even weekly track biases?"

I think the answer is that the bigger the sample, the greater the probability of being correct. But to ignore smaller samples that indicate a bias would be to ignore information that might add value despite occasionally being wrong.

What it comes down to is personal skill at identifying actual biases.

The better you are as a handicapper, the better your techniques for identifying biases, and the better your visual skills, the more often you will be right.

It is not entirely unusual for me to adjust my play slightly after watching one or two races. I may not overhaul my bets completely on such a small sample, but when you've been watching races for 40 years you sometimes see something so unnatural about the way the race developed and how the horses responded when switching paths you know something is amiss.

[Robert Fischer]

1. Skillful observation notes a change in the model itself. Maybe you noticed the pouring rain and high winds? Maybe you noticed the tractor or report 'seal the track'? Maybe you noticed the jocks riding unusually wide? - Then you have to think, "Hey, I should consider how to weight the 'old' model information into my decision process!".

2. You notice some negative events that were unusual enough to possibly be the makings of a 'trend'. Maybe you have a particular play that uses a trainer or a trainer angle, and that barn's last 2 plays and his recent entries have been ice cold - Then you think in terms of whether 'passing' a particular wager opportunity makes sense. Passing a possible negative trend is a lot safer than betting possible positive trends. You don't risk losing your bankroll, you only risk 'opportunity cost'.

[BCOURTNEY]

Originally Posted by dnlgfnk

Perhaps. But Alan Woods didn't...

"We chat about barrier numbers, and Alan tells me about the time in November 1995 when the computer model stopped working for a month or two. Eventually Alan worked out that the last turn at Happy Valley had been re-cambered – which means the track is shaped to slope upwards from the inside rail – creating a disadvantage for inside horses as the outside horses shifted in. His team adjusted the coefficients relating to barrier position and immediately resumed their winning ways."...

https://www.themonthly.com.au/monthl...ches-story-149

Alan has prior posts on this board regarding bias.

[BCOURTNEY]

Originally Posted by AltonKelsey

I guess no one here watches MNR

Very hard to model bias with a computer, and at the very least you'd have to input path info on your own. Post position bias is influenced by track bias where one exists, and can be modeled , but without track bias info, how can you model it reliably?

Better to just play the top consensus pick .

Modeling and detecting bias is simple.
Calibration for it is hard.

[classhandicapper]

Originally Posted by BCOURTNEY

Modeling and detecting bias is simple.
Calibration for it is hard.

The thing that makes it difficult is that most people assume that racing with or against a bias on a particular day will impact all the horses on the card equally. That is often false.

If the rail is more tiring than usual (as opposed to slower than the other paths), how much it impacts each horse will be dependent on how much stamina it has and how it distributed it's energy. The result will be different for each horse, but the trend will be negative.

[pandy]

I post track bias info for NYRA on my website, my own observations, and there certainly are some days that are hard to gauge, especially when there are a lot of turf races. Another thing I look for, say that speed appears to be collapsing, I take a look at the horses that quit. If several horses that looked like strong contenders stopped, that's a good sign that the track was tiring. The same thing when speed holds up, if four odds on horses go wire to wire it doesn't mean that the track was speed biased.

As was mentioned here, one of the trickiest things is when the rider's are avoiding the rail and a few horses go wire to wire in the three path. Do you call that a closer's bias? In that case, if you are sure that the rail was dead, you still have to make a note of it because chances are a few horses did show speed along the inside that day and they may be good bet backs.

[Cratos]

Originally Posted by classhandicapper

The thing that makes it difficult is that most people assume that racing with or against a bias on a particular day will impact all the horses on the card equally. That is often false.

If the rail is more tiring than usual (as opposed to slower than the other paths), how much it impacts each horse will be dependent on how much stamina it has and how it distributed it's energy. The result will be different for each horse, but the trend will be negative.

Using Trakus data, you can independently check each horse race performance and I would think that you would calculate “work” done by the horse to understand how its energy was distributed.

[rsetup]

Originally Posted by DeltaLover

bris?

ummm, No

[DeltaLover]

Originally Posted by rsetup

ummm, No

astrology maybe?

[BCOURTNEY]

Originally Posted by Cratos

Using Trakus data, you can independently check each horse race performance and I would think that you would calculate “work” done by the horse to understand how its energy was distributed.

Cratos how do you gather the horses weight data?

[Magister Ludi]

Originally Posted by BCOURTNEY

Cratos how do you gather the horses weight data?

Obviously, I can't speak for Mr. Cratos. However, accurate remote non-contact weighing can be performed with machine vision.

[cj]

Originally Posted by Magister Ludi

Obviously, I can't speak for Mr. Cratos. However, accurate remote non-contact weighing can be performed with machine vision.

Yeah, I'm sure that is how he is doing it.

[rsetup]

Originally Posted by cj

Yeah, I'm sure that is how he is doing it.

The answer to all your data issues: write some Python code to pull data from videos.

[bobphilo]

When someone takes into account the factors such as whether the horses with the early lead are the ones that figured to win or whether they are taking advantage of a slow pace,or whether the rail was dead or fast not, then these so-called track speed biases disappear.

The silliest question I've heard is asking how the track is playing on the basis of the 1st 3 races where the winners went wire to wire without taking all these factors into consideration.

I'm still waiting to hear a reasonable explanation as to how the track surface can magically distinguish the difference between the hooves of front-runners and closers and favor one over the other.