The Blazers were 100% to make the playoffs a month or so ago. Shouldn’t we be 100% for the rest of the season if the formula used is correct?

The playoff odds are calculated daily based on the current standings, stats, and their relative rankings. They are saying that if everything continues exactly as it is right now, the statistics say that Portland has a 97.4% chance of making the playoffs. Since that assumption is statistically nearly impossible, the odds must be recalculated with every new data point.

It’s not a question of whether the prediction is right or wrong. In fact, it’s not really a prediction at all, since the underlying assumptions will not hold true for the rest of the season.

What would be more interesting to me is the variability of the odds for a given team over the course of the season. If the Blazers’ playoff odds were really high in the early season, then dropped to middlin’ in the middle season, then rose again in the late season, there could be some really interesting statistical analyses of what changed in the Blazers’ play to cause that change. I also don’t know how Hollinger’s system deals with a very small number of games remaining. I don’t remember what happened at the end of last season and what patterns there were in teams’ separation into categories of 0% and 100%. That’s interesting stuff, too.

You cannot assess the validity of a predicted probability with a single event
In general, you cannot assess a predicted probability with a single event. You need many events.

Here’s a slightly simplified example: suppose I predict that teams that are favored to win by 10 points in Vegas have a .70 chance of winning. Now, if the favored team loses, does that invalidate the claim or not? There’s no way of saying. But, if you look at a year of games (or so) and see that teams favored by 10 points in fact win 70% of the time, then the prediction seems about right.

So, how did you assess the validity of Hollinger’s playoff odds. IAt first blush, it seems like you can’t because he has more predicted playoff odds than actual seasons. But, something like the following might be a good way to assess the system:

Re-run the system and the half-way point of the season. This gives you a predicted probability for each season. For each season give the prediction a 1 if correct and a 0 if not correct. That is, if the predicted probability was .50001, he predicts they do make the playoffs and that prediction gets a 1.

Next, calculate the percentage correct for each predicted percentage (probably need to make “bins” of about 2 percentage points 0-2, 2.00001-4,4.0001-6, etc). Then you could see how close the predicted probabilities match the percentage correct. You’d like to see teams predicted to make the playoffs with a probability of .6 to make the playoffs about 60% of the time.

In fact, this is roughly how I would create predictions, rather than the method that Hollinger uses.

Here is all you would need to do: calculate a power rating for each team for each day of the season. Run a probit/logit regression on making the playoffs using the power ranking and number of games played as your independent variables (interacted). Use the results to predicted probabilities in the current season.

67% chance, not 66%.

warning any team that get’s the 8th seed in the event of the Blazers not making the playoffs will face legal consequences.

an ‘arch form’ if you will? I don’t think that they will miss the playoffs though.

at about 77% not long after the 13 game winning streak, but then it fell back down to 0% pretty quickly.

To me the biggest take away from the playoff probabilities is wether or not a teams probability stays fairly consistent. The Blazers have been around 90% since Thanksgiving while Dallas, Phoenix and Utah have been all over the board since then.