Having no fun with super AI's discussing descending W-P-S patterns for heavy parimutuel favorites Posted on July 6, 2025 at 01:55:32 PM by Craig G
Note: this appplies to any parimutuel event with a large field.
I'm starting to wondering about the AI's with respect to original thinking that involves ideas outside of their humungous data sets.
I asked ChatGPT what it thought of the descending results pattern for heavy favorites. I expected it to spit out the underlying logic as if it were child's play.
I gave an example of 547 races with, 185, 96, 72 for WPS.
And for good measure a second case, ultra-conspicuous, 200: 61-43-21.
I wanted it to consider if that pattern was the result of some inherent attribute of the dominant entities. Or what?
Here is the kind of junx I received.
"These results show a declining scale of payouts." (Who said anything about payouts?)
"The odds reflect the probability of an event occurring (e.g., a dog winning a race). The favorite is considered the most likely winner, which means more people bet on it, creating a larger pool for that outcome. However, this larger pool gets divided among the winners, so the payout per winner (the odds) goes down.
Here's why the payouts tend to decrease for favorites:" etc.
Conclusion:
The descending pattern in the results you’re seeing is very typical in parimutuel betting. It reflects the odds, betting volume, and the way payouts are calculated. The more people that bet on a particular outcome (typically the favorite), the lower the payout is per bettor, leading to the kind of results you're seeing with heavily favored dogs or horses.
Holy Toledo, I rephrased and hinted repeatedly, but it did not get the point. Had to spell it out.
OK, here is my key takeaway re this descending pattern. At face value we see, {3.3, 4.7, and 9.5}, and can rightly ask, "What's up with that?" But we need to understand that if the key dog won 61 times out of 200 races, that implies that there were 61 races where it could not possibly place or show. So out of the 139 races where it could place , it accomplished that at a 1 in 3.2 rate. If we extend that reasoning to show, we get 1 in 4.6. So now, our w-p-s pattern {3.3, 3.2, 4.6 } makes a lot more sense. So, we see that the extreme descending pattern is a somewhat misleading artifact of the w-p-s system, and a super dog who wins every race will have the absolute worst place and show percentages record at the track.
Even after dumbing it down to the max, it still kept telling me how "the place and show payouts diminish."
Bottom line = I expected it to see the underlying logic right away, and observe that as win % goes up past a certain point, P + S, go down, with an inverse correlation.
Never happened.
OTOH, if I need some complex SQL query that eludes me, or an instant 3 yards of Python code, not a problem.
And an even bottomer line is that this has profound betting implications for people who box heavy faves in tri's and super's. They are hoping that the sharply reduced freq's for show and fourth are compensated for by proportionately higher prices. But are they? Replies:
Tiger's Jai-Alai Chalk Talk
