Re(2): ChatGPT Follow Up Posted on May 26, 2025 at 09:34:31 PM by Craig G
I think it's good overall, but it also illustrates what I said about being wary of garbage out.
It gives you an example bet. Normally, when you give an example, you want it to be 'exemplary'. (LOL) But including 5/68 as a trikey is a major FAIL. It doesn't know that for those numbers to have a prayer, you need the monster pools of yesteryear. The base probabilities of 568 and 586 are about 1 in 8400 and 1 in 4200, respectively. Multiply by 2 for your target payoff to break even, and there's the problem. You will never get that, and the best case scenario is when you get the entire pool, which these days is what?
Absolutely horrible example.
So it's giving you generic betting advice without an awareness of the warping impact of Spec 7 scoring. Or knowledge of today's severely shrunken tri pools, which tend to push all low-prob tri's into the unplayable category.
It also advised you to look for cases where choice 2 has similar credentials to choice 1, but pays 4 times as much. Well, A, that's almost never gonna happen, and B, how in the world could you ever deduce that if you are following its advice and playing exclusively in the tri pools?
Without beating this to death, it might be interesting to see what it has to offer wrt, "a sample set of trifecta odds and payout estimates to practice spotting value".
I don't want to be a pessimist, but I think the WOPR from WarGames (1983) would assess betting today's jai-alai as being in the Global Thermonuclear War category.
Maybe the truth is in the middle.
But overall, I love ChatGPT and DeepSeek. It's a 'baby with the bath water' kind of thing. The 'baby' part is great. Replies:
Tiger's Jai-Alai Chalk Talk
