(03-16-2019 10:13 PM)H2Oville Rocket Wrote:  

(03-16-2019 02:50 PM)emu steve Wrote:  

(03-15-2019 11:45 AM)kreed5120 Wrote:  I'm 99% confident that Buffalo will beat CMU and 65% confident that BG will beat NIU. This means I'm placing a MAC East vs. MAC East title game at an 82% chance of occurring.

I'm not sure the probs were correct but now they are 100%.

I got about 64%. The probability of a joint event (UB winning .and. BG winning) = prob of UB winning (supposedly 99%) x prob of BG winning (supposedly 65%).

.99 x .65 = .64.

20 - 20 hindsight = 100% probs.

How can one assign any accurate probability of victory to an event with as many variables as a basketball game? I think only those who inhabit Maddawg’s world think that can be done. That word “supposedly” looms large in any attempt at statistical analysis.

You are correct. They are not (usually) calculated using defined metrics, e.g., based on 50 occurences of a semi-final between a #1 and a #4, the #1 won X% of the times. Ditto for #2 vs. whatever.

In these cases, someone ends up with their subjective guess as to what the probability would be.

I would have made UB say an 85% chance of beating CMU.

I would have made BG say 60% chance of beating NIU.

Purely subjective.

As far as the NCAAs, we have a lot of data. #1 has played #16 4 times per tourney for how many years (since the tourney was expanded to 64 or more teams)?

So #16 over #1 has occurred 1 time in over 100 tries, hence the odds of it happening should be say less than 1%.

We can calculate odds for #15 over #2, 14 over #3, etc.