re: Season Record Prediction Using Mathematical Analysis of Point Spreads 2015

When a math problem wins a football game let me know, till than this is just a bunch of numbers.

quote:

---

Can we just all agree that this thread has gotten really stupid, none of you are mathematicians

---

Do you have any idea how stupid this statement is?

quote:

---

none of you can produce a formula that will absolutely 100% accurately predict the outcome of the season

---

Of course not. Nobody ever said that anyone can absolutely 100% accurately predict the outcome of the season. I'd even go so far as to say that preseason predictions are notoriously inaccurate.

And yet there are hundreds, perhaps thousands, of people who attempt to predict the outcome of the season before the season begins. There are entire magazine, webpages, and hundreds of hours of sports programming that are dedicated solely to that task. Why? Because it's fun and it gives people something to talk about before the season begins.

LINK

quote:

---

poor

---

Why?

Care to offer any analysis, or are you just a mindless zombie?

quote:

---

Well, you claimed your prediction is based on mathematical analysis and I am just pointing out where the analysis doesn't make much sense mathematically. Ever heard the quote "averages without variances are meaningless - or worse misleading"? You were trying to predict using mean value but totally ignored the small variance assumption this methodology based on. In your case, one single event's variance for many events is large enough to make the prediction based on mean value useless. For example, in 50% odd case the variable is as large as .25. What makes things much worse is that you are taking summation of these variables. Based on your assumption that these events are independent, if you ever heard of Bienayme formula you can calculate the variance of the summation variable which is the summation of all independent variance - that's huge! The variance of summation variable is large enough so that its mean value means a little better than nothing.

---

quote:

---

It seems I overestimate your level and you seems didn't understand what I was talking about. I was not arguing about whether an odd/line is correlated to event outcome at all. Let's put it in easy to understand way - after game X happened, will the the odd/line be adjusted for future games? Do you think using adjusted odd to calculate then leads to more actuate prediction at that point? If your answers to both problems are yes, then can you tell the difference between your independent events model and a different model that takes account for prior probabilities? Ever heard of the word random process?

---

quote:

---

It seems you cannot even understand plain English without math. Your number is predicted using current, today's, 8/22/2015's odd for all games, and do you think the pregame odd before each game is exactly the same as current odd? If your predict is totally based on current odd, then the number doesn't mean much since the odds change a lot.

In summary, I don't have any problem with you doing some high school math exercise and call it your prediction. But I wouldn't calling it mathematical analysis - the method and model have some systematic flaws and leads to meaningless results. To me the model performance is probably similar to a dummy model that picks a random number from a few numbers.

---

When you have some semblance of an understanding of the words you are using, come back and join the discussion.

quote:

---

I appreciate the analysis.

---

Thank you.

quote:

---

I do have a few thoughts on it.

---

I'm all ears . . . or all eyes.

quote:

---

1. While there may not be a better Mathematical basis to plug into your formula I have little faith that lines generated in July can be considered good data. Kinda garbage in garbage out.

---

The analysis remains the same with updated lines. The more current the data, the more accurate the prediction. But really, the larger point is that predictions are inherently difficult to make. If you compare the thousands of predictions made prior to last year, I think you'll find that most were wildly inaccurate. Nevertheless, there is no end in sight for preseason predictions.

quote:

---

2. "Adjusted for subsequent information". Can you elaborate?

---

I used the latest lines I could find, and I used 100% for the cupcake games. I don't think I made any other adjustments for subsequent information when performing the analysis this year.

quote:

---

3. "Adjusted for % chance to win". Is it wise in any mathematical predictor to assign 100% to anything? If even 90 or 95% is used for those 3 games wouldn't it drop the predictor closer 8 than 9 wins?

---

I hear what you're saying. Is any football game a 100% lock. No, probably not. I used 100% for three reasons, in decreasing order of importance. First, I could not go below 95% because I expected a reasonable point spread for all three of those games to exceed 21.5. (The line for Western Kentucky and Vanderbilt opened at 17.5) ) Second, the difference in the conclusion that results from moving the three games to 97% from 100%, leaves the conclusion at 8.51 wins instead of 8.6 wins. Pretty negligible. Third, the numbers were cleaner.

quote:

---

Just seems to me 100% is not a good number to plug in in any predictive model.

---

I completely agree with your thought process.

quote:

---

That said I appreciate the work you put into it.

---

Thank you.

quote:

---

When a math problem wins a football game let me know, till than this is just a bunch of numbers.

---

That has always been the way it is with stats, betting lines, and season records.

That's all a preseason prediction is too. It's just a bunch of numbers.

quote:

---

Your use of non-sense jargon that you don't understand almost hid how confused you were. Thanks for bringing this out to clear it up for everybody in plain terms.

---

I assumed the posters here are with college level education and some stats background. Apparently I was wrong. Sorry for the confusion it may caused for you.

quote:

---

Why use pre-game odds? Why not half time odds? Pre-game odds would be different and thus inaccurate by that time by your logic..... His prediction is being made right now so of course he is going to use today odds. Jesus, this is truly remarkable. So many buffoons.

---

See, I was not even talking about whether he should use today's odd or the calculation based on his model is wrong. I was talking about different ways to model the process and a overly simplified model cannot lead to meaningful results. Certainly you can use the half time odd to adjust the expectation at half time, and that leads to yet another model superior to OP's.

Although it is not exact, you can view the whole season as a process to estimate some hidden parameters about 'true' power of teams. After each game's outcome the odds are adjusted in favor to better estimate the parameter value, and the estimates converge at end of season. If you plot the estimate values over time it would like a damped oscillation curve. OP's prediction is a peek at the starting value of the curve and it has little information about the end value. Again if you didn't ever heard of the term random process this may not make sense to you at all. Sorry.

Even though you threw some attacking words at me, I won't behave at your level and simply call you stupid. You may be a smart person but are lack of relevant education and background knowledge to discuss the topic. It's not your fault though.

quote:

---

tripe in all preseason prediction threads

---

My fun is predicting preseason results after the season ends.

quote:

---

What are the college or graduate level stats courses you ever took

---

Took this guy three years to do his arithmetic demonstrating another moral victory for Les Miles.

quote:

---

That's not how it works.

---

Yes it is.

quote:

---

If your odds of success are accurate, LSU has a better chance of winning 10 games than 8 games.

---

No, he's right. If you have an 80% in each of 10 games, then you should win eight of the ten. It's like a free throw shooter who shoots 80%. Any given time he shoots, he's 80% likely to make it, so you would always bet on him to make it, but 20% of the time he will miss. If you bet on him every time, you will win 80% of the time, so it's good. But if you have to predict how many out of 10 he will make, your better guess is 8, not 10.