Most Depressing Losses of All Time

[Panama]

If that is what he is saying it is ridiculous. Why in the world would you need be an omniscient god? Too many variables to consider? If you were omniscient than probability is sort of meaningless, you know what is going to happen. This in not like betting odds or something.

If this is the line of argument than the method used to determine the probability needs to be what is criticized. That is a separate technical discussion that would need to happen with those who generated the probability statistics. That is a reasonable thing to do.

Arbitrarily dismissing probability statistics because you don't understand them and think only an omniscient god could determine them is ridiculous as well. That is not a reasonable thing to do.

No, turns out I also misunderstood and that's not what he was saying. He was saying you can't compare apples to oranges, the apples being win probability at one point in one game and the oranges being the win probability at a different point in a different game and that you can't extrapolate from that to the entire game.

In other words, because the win probability at one point in the third quarter of the Houston game was 90-something% and with 20s left of the Washington game was 98%, it's not valid to state that our win probability was the similar for these two games or that we had a 90-something% probability of winning the Houston game and 98% probability of winning the Washington game.

Now, I'd like to know what the probability of my playing games with these apples and oranges is:
:salma:

[BlazingBolt]

Sorry, I guess we were saying three different things. I misinterpreted your objection to be to what goes into determining a win probability, "cracking open those formulas."

But the argument you were actually making is equally valid. For BlazingBolt's benefit, that is: you can compare win probability for each game at the same moment in time (say, with 21 seconds left, or at the same point in the 3rd quarter, or whatever), but it's not valid to compare the win percentage at some random point in the 3rd quarter of one game with the win percentage with 21 seconds left in another. Gotcha.

This I do not agree with. The probability takes in to account the time left in the game already. I think I made that point repeatedly. Having a 10 point lead in the first quarter and having a 10 point lead with 21 seconds left can be compared....with 21 seconds left your probability is much much higher.

[BlazingBolt]

As someone who builds statistical models and algorithms for a living, to suggest that I was at any point dismissing probability as a valid statistical tool is laughable. I never said any such thing.

Well it sure seemed to me that you did. You said it was meaningless because the probability changes throughout the game based on the plays that happen in my understanding. I disagree that the unknown outcome of plays changing the probability invalidates anything, that is how it is supposed to work. Positive plays increase your probability of winning. The time left in the game is a huge factor as well, that is why I don't understand how you can say looking at the probability at various points in the game invalidates anything either.

[ArtistFormerlyKnownAsBKR]

This I do not agree with. The probability takes in to account the time left in the game already.

You made that point and it is incorrect. A 93% win probability in the second quarter your confidence interval is much, much lower than it is with 21 seconds left. Moreover, to extrapolate trends using these probabilities at different points in three different games is not statistically valid.

This is what I wrote in post 61:

When you start to apply it across multiple games and your read is at different points in the game it becomes even more useless. While I will give you that the percentage with 21 seconds left is a pretty sound number, those taken in the other games are at various points. They've self-selected the high-points but that's no more useful or meaningful than selecting the low points. Totally arbitrary.

[BlazingBolt]

Sorry, I guess we were saying three different things. I misinterpreted your objection to be to what goes into determining a win probability, "cracking open those formulas."

But the argument you were actually making is equally valid. For BlazingBolt's benefit, that is: you can compare win probability for each game at the same moment in time (say, with 21 seconds left, or at the same point in the 3rd quarter, or whatever), but it's not valid to compare the win percentage at some random point in the 3rd quarter of one game with the win percentage with 21 seconds left in another. Gotcha.

Not sure why. I think they are comparing the point in the game where we had the highest probability of winning, whether that was at a point in the 3rd quarter of with 21 seconds left it is apples to apples because the probability takes in to account the time left in the game.

[ArtistFormerlyKnownAsBKR]

Not sure why. I think they are comparing the point in the game where we had the highest probability of winning, whether that was at a point in the 3rd quarter of with 21 seconds left it is apples to apples because the probability takes in to account the time left in the game.

You can't do that because the margin for error is different in each context, even if the probabilities are similarly high. It's arbitrary. As I said before, you could cherry-pick the low points and come out with aa very different read on the outcomes and it would be equally invalid, statistically speaking.

[BlazingBolt]

You made that point and it is incorrect. A 93% win probability in the second quarter your confidence interval is much, much lower than it is with 21 seconds left. Moreover, to extrapolate trends using these probabilities at different points in three different games is not statistically valid.

And how in the hell is a 93% win probability in the second quarter even possible? Confidence is a separate thing.

I don't see what is being extrapolated. Comparing the highest probability achieved in 3 separate games seems perfectly valid. Comparing the probability with 21 seconds left in 3 separate games would be arbitrary.

Reaching such a high probability in 3 separate games that you then lose is mind boggling. You don't even need the statistics to know that, it is plain to anyone to see. It means we probably should have won those games don't you think?

[ArtistFormerlyKnownAsBKR]

And how in the hell is a 93% win probability in the second quarter even possible? Confidence is a separate thing.

I don't see what is being extrapolated. Comparing the highest probability achieved in 3 separate games seems perfectly valid. Comparing the probability with 21 seconds left in 3 separate games would be arbitrary.

Reaching such a high probability in 3 separate games that you then lose is mind boggling. You don't even need the statistics to know that, it is plain to anyone to see. It means we probably should have won those games don't you think?

It means we agree to disagree.

[Panama]

This I do not agree with. The probability takes in to account the time left in the game already. I think I made that point repeatedly. Having a 10 point lead in the first quarter and having a 10 point lead with 21 seconds left can be compared....with 21 seconds left your probability is much much higher.

Again, that's not the point. Of course, the probability of winning is much higher if you have a 1-point lead with 1 second to go than if you have a 1-point lead with 45 minutes to go. Hello, McFly! And no one will dispute that.

What it sure sounds like you're saying, though, is that the stats that TTK cited:

they pulled out stats that had the Chargers winning the Texans game (95% chance), Titans (98% chance) and Redskins (99%) chance when they were ahead

are valid. Yet the objection is that those probabilities were taken from different and incomparable moments in the games. Just because the Chargers had a 95% probability of winning at one point in the Houston game and a 98% chance at another point in the Titans game doesn't mean that they had a 95% and 98% chance of winning the respective game. I don't know at what point the probabilities were determined, but a more valid comparison would be to compare the win probabilities for the same moment in each game. We know the 99% figure for the Redskins game was based on having the ball inside the Washington 1 yard line down by 3 and holding 2 timeouts with 21 seconds to go. What was the win probability for the Chargers with 21 seconds to go in the Houston and Tennessee games? That is a more valid comparison.

I think there is a 100% probability that this discussion has grown tiresome.

[BlazingBolt]

You can't do that because the margin for error is different in each context, even if the probabilities are similarly high. It's arbitrary. As I said before, you could cherry-pick the low points and come out with aa very different read on the outcomes and it would be equally invalid, statistically speaking.

All talk. If you were cherry picking low points i would assume you are looking at a team that has made improbable comebacks. Seems equally valid to me.

The entire point of referencing the probability is to show how unlikely the actual outcome was. I don't see any reasonable discussion of saying wow this team had such and such probability and they ended up that way. Arbitrarily looking at random points and cherry-picking is going to give you what? Give an example?

[BlazingBolt]

Again, that's not the point. Of course, the probability of winning is much higher if you have a 1-point lead with 1 second to go than if you have a 1-point lead with 45 minutes to go. Hello, McFly! And no one will dispute that.

What it sure sounds like you're saying, though, is that the stats that TTK cited:



are valid. Yet the objection is that those probabilities were taken from different and incomparable moments in the games. Just because the Chargers had a 95% probability of winning at one point in the Houston game and a 98% chance at another point in the Titans game doesn't mean that they had a 95% and 98% chance of winning the respective game. I don't know at what point the probabilities were determined, but a more valid comparison would be to compare the win probabilities for the same moment in each game. We know the 99% figure for the Redskins game was based on having the ball inside the Washington 1 yard line down by 3 and holding 2 timeouts with 21 seconds to go. What was the win probability for the Chargers with 21 seconds to go in the Houston and Tennessee games? That is a more valid comparison.

I think there is a 100% probability that this discussion has grown tiresome.

I am assuming the stats are valid. I don't think Ben Higgins is putting out some non-sensical stats.

The moments are comparable because they are the highest points in the game. The moments where the Chargers winning was the most likely. They then lost all those games.

It is the entire point of Higgins quoting those probailities. The argument that they are at different points in the game just doesn't make sense to me.

The point is the Chargers lost 3 games that they very nearly had locked up, for each of those games the Chargers gained a huge advantage (21 point lead in second half, possesion with a 4+ point lead with little time left when your opponent has no timeouts, Down 3 with first and goal inside the 1 with multiple timeouts) to the point that the opponent coming back to beat us (or being able to stop us from winning) was very unlikely. Higgins put a number on each game, but regardless of the number or the accuracy of the statistic it is a perfectly valid point. Losing 3 games in such an improbable fashion is kind of a crazy thing. We should have won all 3. Can we agree on that?

[ArtistFormerlyKnownAsBKR]

I am assuming the stats are valid. I don't think Ben Higgins is putting out some non-sensical stats.

The moments are comparable because they are the highest points in the game. The moments where the Chargers winning was the most likely. They then lost all those games.

It is the entire point of Higgins quoting those probailities. The argument that they are at different points in the game just doesn't make sense to me.

The point is the Chargers lost 3 games that they very nearly had locked up, for each of those games the Chargers gained a huge advantage (21 point lead in second half, possesion with a 4+ point lead with little time left when your opponent has no timeouts, Down 3 with first and goal inside the 1 with multiple timeouts) to the point that the opponent coming back to beat us (or being able to stop us from winning) was very unlikely. Higgins put a number on each game, but regardless of the number or the accuracy of the statistic it is a perfectly valid point. Losing 3 games in such an improbable fashion is kind of a crazy thing. We should have won all 3. Can we agree on that?

I'm sure Yubaking has them in the win column. We're 7-1 in Yubaworld.