You can never know the "accuracy" of a probability by looking at what actually happens. Winning a 1 in a million lotto jackpot does not make the 1 in a million odd inaccurate. The accuracy resides in the algorithm that the probabilities are derived.
I am wanting to judge the long term accuracy of said algorithm.....it is just interesting to me. More interesting than adjusting the forecast each and every game...at least to me. DD
The Jazz getting 8th was about a 3.8% chance, but it really was a longshot for Jazz to get 8th because the Mavs were so far behind at the time. All of the other probabilities weren't too bad. The Spurs had a high probability of finishing second, but then again they've had a pretty bad end of season (and its not like they had a low probability of finishing 4th or 5th. Rockets and Blazers had high probabilities around #3 and #4. I'd say this program did a solid job.
How accurate does it have to be though? These types of programs are going to look "wrong" when longshot things happen. When San Antonio loses Ginobli for the year. When Dallas suddenly starts playing great after playing like crap all year. When Utah loses at home to Golden State.
That is what I want to see, how accurate probabilities are when dealing with the human condition of atually playing the games. DD
As someone brought up earlier, you flip 10 coins, you expect heads to pop up 50% of the time and tails to pop up 50%. But when the first 3 coins come up heads, do you still expect the probabilities not to change? Does that mean the initial "prediction" was inacurate? It doesn't. Or if you ever watch poker on ESPN and they show someone with a 65% chance to win, and his opponent with 35%. When the last card flips, if the underdog wins, do you say the prediction was inaccurate? No, you don't. Just because the lesser probability won out, doesn't mean the probability was wrong. You aren't going to be able to determine how accurate the probabilities are unless you can somehow "run" the last few months of the season over and over again and collect the data on how often a team gets what seed in the playoffs. You can't determine the accuracy of the probabilities by just looking at one outcome. A probability can't be wrong. Maybe the math behind it is wrong (which I dont think is in this case, but you could make the argument that the math is bad).
Yes, I understand......I am just concerned with the formula itself...the variables used to determine who is favored, are injuries factored in, etc..etc..etc...and whether the initial projects held up. Just intersted in a specific part.....why does it bother you so much? DD
Your language makes it sound like this simulator (if the probabilities don't hold up) is a bunch of ****, and I have to respectfully disagree with you there. Thats why it bothers me. And you act like if the probabilities had for example Utah with a 3.8% chance of winning the 8th seed initially, and they still get the 8th seed despite the low probability, then that somehow proves that the simulator sucks. All it proves is the long shot happens. And if thats what you believe, then I couldn't disagree with you more. Again, I take you back to the heads/tails or the poker analogy. One outcome doesn't prove probabilities wrong.
When is DD going to challenge BigBigRed to prove that he isn't actually Carmichael Dave using another account? I'm kidding. Well, sort of.
Well some statistically unlikely things had to happen for us to end up where we are right now. The j*zz lost to Minnesota and Golden State at home. The Spurs lost Ginobili. The Mavs got really hot. So if the results end up showing that something that had a low probability happens that are related to those factors, I would say that's pretty acurate. DD, as long as you understand that the end result, which is likely to vary with the initial prediction, does not undermine the system itself. This isn't fortune telling and no one has taken this as gospel.
The point is that there's so much variation across a season that trying to run probabilities is only going to be of limited usefulness. I think this is pretty cool, but do the numbers really tell you anything that your own common sense couldn't?
Well assuming the math is generally acurate, it gives you a general idea where you and other teams stand. It does it from an unbiased standpoint. I agree, the usefullness is limited. I don't think during my day, "dang, the Rockets chances of teh 2 seed fell to 12.5% today. Oh well maybe next year." But I do look at it just to kinda help gauge the season along with the common sense that comes from watching the game.
Are you telling me that there is only one tenth of a percent seperating us playing the hornets and the jazz? Wow. But as i have said in earlier threads if we match up with jazz.........so be it.
I hope you realize that simulation was done using results from March 31st. I'd skip to the page before this one to get the latest April 11th results. Updated ones should be done tonight.
Even better his one sentence question is sandwiched between 500 lines of quoting and 200 lines of sig. lol.
Your quote implies that: You didn't use the words I used, but this to me implies that... if something that he put a probability of like 5% a month ago happens, then that means this simulator wasn't that good. I don't see how your words can be interpreted any differently. And you just said it again too "I'm just curious as to how close the initial predictions are." You don't specifically say "if the predictions are off, then this simulator is bad", but no one says "I'm curious how close the predictions are..." only to follow it with "it doesn't matter mean anything if the predictions are off, I'm just curious". And I go back to what I said before... just because the predictions are off, doesn't mean anything in terms of how good the simulator is. It just means that the longshot happened and that skews the results.
