THE PRESS CONFERENCE ISN'T ON IN JACKSONVILLE!!!!!
THE PRESS CONFERENCE ISN'T ON IN JACKSONVILLE!!!!!
Perhaps the recent history of spectacular moderator meltdowns is convincing JI to move in a different direction.
New releases of board software are currently undergoing beta testing. Board ownership are being close-mouthed about what exactly is being tested but it is believed that selected moderator perogatives can be unlocked by the JI faithful simply by being here a lot, clicking on lots of the ads and generally kissing up to management.
"The aim is to be completely a moderator-free environment before the wild card round of the playoffs this year!" confided one person close to the situation. "moderators are just so full of themselves and anything that would help the board collectively flick this piece of snot of it's fingernail is a good thing" said another. The new unlocked moderator priviledges can only be used by a poster on themself so the Ban priviledge for example, can only be used by a poster to ban themsleves and not to ban a 3rd party. Checking your own IP is expected to be a real winner and will be linked to a click-through advertizement for Enzyte.
I'm watching the Finals Game 4 postgame coverage on ESPN tonight, and it's clear at some point that the desk of Wilbon, Magic, etc, is wrapping up and they are about to send it back to the studio. As the conversation ends, Stuart Scott gets ready to do the send off. He looks at the camera like he has something terribly important to say, a real zinger.
He says, "Game 5 upcoming is crucial...in NBA playoffs history when the series is tied 2-2, the team that wins Game 5 goes on to win the series 73% of the time"
My first thought: OK, interesting, it is important to win Game 5!
Wait a minute.
Let's analyze this. All else being equal, a team should have a 50% chance to win each game. Stuart in his statistic did not mention home court advantage, momentum, or team skill, so let's work within the same context of the stated 73% statistic. A 50% chance to win each game is roughly accurate.
If a team wins Game 5, they must then win 1 of the remaining games to win the series. However, Game 6 and Game 7 both carry roughly a 50% chance of a loss for the team that won Game 5. Since we are not considering momentum or home court (as per Stuart), Games 6 and 7 are independent probabilities.
Therefore, the chance that the team who won Game 5 will then go on the lose the next two is approximately: (1/2) * (1/2) = 1/4, or .25, or 25%.
Let's fit this back to what Stuart said. He stated proudly that Game 5 was critical because the winner would win the series 73% of the time, or in other words lose the series 27% percent of the time. When looking at the raw probabilities, this actually means that if you win Game 5 you are slightly mathematically worse off than if the remaining games were played by chance alone.
Yes, I get that 73% is probably within the standard error for the sample size, and the 73% figure proudly quoted by Stuart Scott can roughly translate to: the average mathematical odds prevail - there is no clear advantage to winning Game 5 other than the win itself.
But then, why even make a point of this, ESPN? Why make it seem like you are revealing some huge mathematical wormhole that applies to the winner of Game 5? I just can't believe that people watch this crap religiously. What's sad is that a lot of Dallas and Heat fans heard that and went to bed tonight dreaming of the wonders that their team will reap if they can manage to win the magical Game 5. It truly is the dumbing down of America when our leading sports channel can freely apply any sort of misleading statistics to any story they'd like, and nobody gets put off by it.
No, as I stated, I know that in actuality each game is not 50/50, but I am working within Stuart Scott's confines. In his 73% statistic he did not reference if it mattered if the home team won Game 5, he did not mention superstars, nor momentum. He looked at the aggregate data (and with a large enough data set one would assume those things would all cancel out) and gave a broad summation of his findings. So in a way we agree, ten-not-done, that Stuart was using a relative frequency approach, looking at series winners and not game by game winners.
But you simply have to use a game-by-game prediction model for prediction of future outcomes in sports...it's how all web based generators work, from WhatIfSports to ESPN's own NBA playoffs predictor. Since I don't have a data set to work with, I'm assuming that the rules of Stuart's set (aggregate, over time each game would be 50/50 because of all the cancelling factors) will apply to my game-by-game approach...that's why I roughly surmised it to be 50/50.
My gripe is exactly this: Stuart stated his relative, aggregate findings as if it was a tell-all mathematical compass for who would win the series....but by operating within his guidelines, if you use a classical probability approach (as one naturally would for purposes of prediction), the numbers reveal that Stuart's 73% tells us exactly the opposite - that if the Game 5 winner wins the series 73% of the time it is actually below the likelihood that they should win, given all else equal.
I recognize margin of error, but the insignificant difference between 73% (derived from aggregate) and 75% (classical approach) begs the question: why is ESPN even wasting their breath with such a meaningless statistic, and then trying to pass it off like they are revealing some great piece of wisdom? Actually I can guess the answer to that, but it's still pathetic.
Again, you can't use classical approach here because the outcomes are not equally likely. It's a useless statistic.
Nothing wrong with using what happened in the past to try to predict future either. It is how it is done almost everywhere. It is a valid piece of data to consider.
Also nothing wrong with looking at it game-by-game. That happens all the time everywhere including the pfail strip. That's like if the Jets were to make it to the playoffs as a wild card team and they were playing the Colts in the AFC Championship game. You could look at historical data and say the Jets only have a 15% (or whatever) chance of making it to the super bowl because in the past only 15% of wild card teams have ever made it that far because of not having home field advantage, playing an extra game, etc. (relative frequency approach). But wait a minute... What if Peyton Manning just broke his pelvis and is out for the game? That changes things. (game by game analysis).
Both are valid ways of trying to analyze what might happen in the game. And what else is ESPN (or posters on JI) going to talk about nonstop leading up to the game. What would be a completely invalid argument though, is saying they have a 50% chance of winning or losing. Just because these are the only two outcomes does not mean each one is 50% likely. If that were true I would have a 50% shot of kicking Roger Federer's ass at the US Open. I think you are over analyzing this stuff.
Eh, I think you are still misunderstanding what I am saying. Let me try to put it one more way, and then I think I'm done with enough math to last me through the lockout.
According to Stuart Scott, 73% of teams that win Game 5 from a 2-2 tie go on to win the series.
Now, what we are trying to find out is if that 73% figure is drastically different than what one would expect to happen, over the course of the past 40 years of NBA playoffs. The underlined part is especially important.
Now, here is another important part: over the course of the past 40 years, with all the Game 5's from a 2-2 tie, do we think there is any significant pattern in the Game 5 winners? No of course not! There are too many variables over that many years (including some of the very variables you mentioned, such as injuries) to claim anything significant about the skill of the Game 5 winner relative to the other team.
Therefore, what we would expect to have happened over 40 years is for the team that won Game 5 to have roughly a 50/50 shot in each remaining game. Again, I'm not saying that any one particular series had 50/50 odds for any game, I'm saying that over the course of 40 playoff seasons, 50/50 is roughly what you would expect if you knew nothing about the two teams....And we just stated that knowing the winner of Game 5 tells us nothing about the two teams.
Then, you follow through with the math and find that a 75% series win rate after winning Game 5 (what we would roughly expect) is not very different at all from 73% (the actual number), which brings us to why ESPN considered it a useful statistic at all.
-----------
Or, in other terms, I'm sure you've heard before of Z-tests and T-tests, Null and Alternate Hypotheses. One way to phrase this whole exercise would be:
Ho: There is no benefit to winning Game 5, other than the win itself
Ha: There is additional benefit to winning Game 5
You use the test formula to find your value, look up how many standard deviations it is from the mean, and can say at what confidence level the outcome of Game 5 is statistically significant or not.
Again, using 50/50 is not to determine any particular series with accuracy, it is to give us a baseline from which to judge how exceptional Stuart's 73% really is. 50/50 is our Null Hypothesis. And with the two X values being 75% and 73%, the 73% figure is not very exceptional at all, and the Null Hypothesis would be proven correct.
Ok that is all, I know all the diehard math fans that have been closely hitting F5 will thank me for bumping this one last time.
Of course winning game 5 is a huge physiological advantage as the 73% figure indicates. You cannot discount this. If you want to talk about hypothesis testing I can tell you without even running the numbers that 73% is going to be statistically significant. Use any confidence level you want.
I think I see how you are coming up with this 75% figure now and I can explain why that is not right. you keep trying to use classical approach to probability here, and as I've explained, that cannot work because the outcomes are not equally likely. Classical approach only works for things like rolling dice and picking lottery numbers. For everything else you have to take a sample and try to approximate the true probability by calculating the number of successes out of the total number of trials. You can never know the true probability, but you can be damn accurate if your sample size is large enough. This is how they got the 73% figure. By looking at the percentage of teams that won game 5 and then went on to win the series.
Here's what you are doing... First of all you can't state that each team has a 50% shot at winning each game because that's too simplistic and it doesn't work that way for the reasons I've stated above. Second, assuming you could use the classical approach, your math does not pan out.
You stated in a previous post that a team has a 25% chance of LOSING both game 5 and game six.
.5 * .5 = .25
And so the probability of WINNING both games would be 75%.
1 - .25 = .75
Nope. Doesn't work that way. If you are going to use classical approach (which you can't but just for illustration purposes I will play along) then you have a 50% chance of winning OR losing each game. So if both game five AND game six have not been played yet, you have a 25% chance of losing both games, true. But you also have the same 25% chance of winning both games. The same .5 * .5 = .25 still applies, NOT 1 - .25. You would also have the same 25% chance of winning game five and then losing game six, or losing game 5 and then winning game six.
Do you see why that doesn't work now?
I see the error in your thinking. It works and I will tell you why. I didn't arrive at the 75% figure by determining the probability that the team would win BOTH games. That would be moronic, because if there is a win in the first game the series is over, and obviously the percentage of winning BOTH would never be higher than the percentage of losing BOTH if the odds are 50/50 for each game. Nobody, I think, would make that mistake.
The 75% is derived from the odds of NOT losing both games. In the NBA playoffs, you either win a game, or you lose. If there are two games remaining, and you DON'T LOSE BOTH, you win the series. The odds of losing BOTH are 25%. If you don't lose BOTH, the only other outcome is that you win at least ONE. Therefore you win the series at a rate of 1 - (odds of losing BOTH).
That is where the 75% comes from, and I am shocked you are just getting to the point where you are understanding where that number comes from. It makes me wonder why I've been taking the time to rationalize the subsequent steps of the process to you if you didn't understand the basic parameters.
Look, it's sheer craziness for you to claim that time sequence hypothesis testing using real data sets is impossible. Statisticians do this every day. To use the most textbook example, let's say you start a beverage company. You claim you will have 12 ounces in each beverage bottle. After one year, we take a random sample of your bottles and determine there was actually 12.05 ounces in each bottle in the sample we took. A Z-Test is now used - this is essentially what we are doing with the NBA - to determine if this discrepancy between the stated 12 ounces and the 12.05 in the sample is statistically significant. This depends on sample size, and standard deviation. If the extra .05 ounces is not significant, you can go on claiming to have 12 ounces per bottle. If it IS found to be significant, you need to change your claim on the bottle.
Here's how this relates to the NBA. The "claim on the bottle", so to speak, is the idea that, over time, NBA playoff games have roughly 50/50 odds. With a large enough population this is undoubtedly true. (To put it one last time, if I took all historical NBA playoff teams and divided them completely randomly into Group A and Group B, you'd find that the overall win percentage for Group A would be roughly around .5, and for Group B roughly around .5. I don't know how to state this any clearer, but that is the truth. The only argument can be whether the population size is large enough, and I would argue that over at least 40 years of playoffs, it would be.)
What we are doing is taking Stuart Scott's actual findings - much in the same way we found your bottling sample to actually have 12.05 ounces per bottle - and seeing if his findings are statistically significant compared to the 75% "rough expectancy."
This, to me, proves that we are not even on the same page in understanding the ramifications of what Stuart said. You are saying the 73% figure is evidence of some type of momentum after the Game 5 win. But it's clear that even using a 50/50 estimate for both remaining games - an extremely conservative estimate that DOES NOT add in any momentum - the team that won Game 5 should still be expected to win the series 75% of the time. So Stuart's 73% figure is basically saying there has been negative momentum for the Game 5 winner, because historically they have fared worse than 50/50 in at least one of the remaining two games. This is from his real data!
To put it another way, again, you'd EXPECT that the Game 5 winner would have better than a 50/50 chance in both remaining games, because of momentum or whatever. But 73% indicates that they historically have not had that advantage, because it is below the 75% threshold that you would get IF every game truly WAS 50/50.
That's what got me so interested in his ridiculous statistic in the first place. If the number would have been 77%, or higher, instead of 73%, I'd say, OK, slightly higher than what you would expect from a 50/50 draw alone. Somewhat useful stat. But for ESPN to state 73% without realizing that it actually implies negative momentum is ridiculous. Hannah Storm said it AGAIN last night. It's like their story for Game 5. And it is, at best, completely useless, and at worst implying the opposite of what they mean,
Any truth to us offering HAM a vet min deal to return?
Is shakin supposed to replace 32??? He's been a great mod with very good numbers for many years, but he's OLD now. I know 32 is just a solid 3rd string mod and he feels disrespected as we keep adding new mods into the mix as he still thinks he's young and good enough to at least be a 2, but he's a great team guy and is one of the oldest tenured posters... Jeeze losing him AND HAM is going to really mix up the face of our team. That's two fan favorites and Sooth didn't apparently care to keep either. Next you'll tell me we aren't going to make an offer to the versatile and very dangerous WCO. Even with the new changes to the posting rules he still adds so much value running the wildbat(****crazy) threads for us. Let me guess he's just going to sign with the Bills' board as a mod while we chase down another top flight nice guy Jetswin for huge money when we already have the best in the biz. It's too much money locked up in one position!!!
Is there any suspicion that 32 was looking to better-deal his JI buddies from the start and that all of this here was just a stepping stone on the road to bigger and better things?
I haven't liked 32g since he slept with Cotch's wife or girlfriend.
You guys pick on Stokes a lot.
I'm glad he's being the bigger man.
This actually completes the JetsInsider-PatsFans trade which brought me here...32g was just named the 'future considerations' part of that deal.
You guys got screwed......
Bookmarks