What are we measuring here? Posted on December 29, 2023 at 01:02:42 PM by Craig G
This is a tricky area.
For one thing, there is an extra superfecta game this season. That will automatically bump up the median and average tri payoffs a bit.
More importantly, what is the meaning of the price when the winning tri had less that $2 bet on it?
Case in point = Dec 6 Wed mat game 5 --> 871 for $6,445.40.
Approximating the takeout, do we think that there was $8,600 in the tri pool? Obviously not.
Since 871 has a native prob of 1 in 1283, and you state '220 games', that means ~ 147 Spec 7's. If we take the sum of all the probs of Spec 7 tri's that are 1 in 1000 or worse, it is about 3.8%.
Using the binomial calculator at Stat Trek, it looks like having about 5 of those bad boy numbers hit in 147 trials is typical. So now the question is, 'In that handful of hits, how likely is it that one or more had only 50 cents bet on it?'
My point is that it looks like we are measuring the likeliness of some freak occurrence, when what we are really trying to measure is the health and vitality of the typical tri pools.
IMO, a better metric is needed.
We could just ignore N outliers.
Or we could borrow from the Dow Jones Dirty Industrials, and take the top 30 tri's based on descending prob. Those numbers are almost guaranteed to be covered at least $2. I think that since my DB has the Dania prices from 2001 -> 2018, I should put a little effort into producing some yearly figs for that top 30 group. Might be revealing. We almost have to predict a steady decline in ROI, right?
As an aside, I do have a copy of 'Calculated Bets' where Prof Skiena has a table showing that the top 30 tri's apparently ranged in ROI from 1.78 for 142 to 1.24 for 325. If that is true, no wonder everybody here has happy jai-alai memories from yesteryear.
