Hey guys, I am taking a crack this December at making a Box Office Prediction game. Oh, I've said this before, and it's never really taken flight, but this time should be different. Previous times have often been derailed due to cost, but now I'm doing it myself, so that's not much of an issue.

My problem is, I don't play the game. I never have. So I am afraid that while I understand the basic concept, I don't know the dirty little details.

If you guys could do the following for me, I'd really appreciate it.

- Explain the basic prediction process. What do you do? Everything from logging in, etc etc.
- Aside from predicting, what else do you do?
- What's the interface like?
- What's the formula for calculating percentages?
- When are the results given?
- Is there anything you don't like?
- Is there anything you wish the game did, that it currently doesn't?
- Are there any comparisons, etc, that you wish you could do, but can't?

That's about all I can cook up for now. If you have a technical mind, and want to help, you can always start thinking about the database structure!

Thanks guys, with any luck, we'll have some semblance of a game in January.

From the user's perspective, each of us will have a username and password to log in, then we get to the page where the predicted movies are listed, enter our predictions (the text field should check the validity of the input), and submit. Then the set of predictions should be associated with a user, and the interface should allow us to retrieve it and modify it up to the deadline.



Nothing really, but more statistics the better, since we like to compare against each other; so both personal stats and overall stats regarding the averages, rankings, etc.



I think this could be flexible. The BOM one generally is ok, but like their forum board, it feels old.



sum(1 - min(1, abs(predicted - actual) / actual)) / the number of movies



Monday after the weekend actuals are known; it could become Tuesday sometime if we're predicting a long holiday weekend.



One thing I think we don't need with our game is to make the averages of everyone's predictions available before the deadline like BOM does. This is more important for us because unlike BOM, we have quite fewer crazy/dont-care players, at least at the beginning, so the overall average for each movie could be a lot more accurate, and it'd be more fair if it's not available until after the deadline each week.

The way the score is calculated could be tweaked. It has a couple of issues:

1) Underpredicting is penalized less than overpredicting. You can't get 0% for underpredicting unless you predict 0, while you could get 0% for overpredicting by 100% or more. One solution is to change the formula to sum(1 - min(1, abs(predicted - actual) / min(predicted, actual))) / (the number of movies);

2) You don't get penalized more once you reach a score of 0 on a movie. So if a movie earned $0.5m, you would get 0 for predicting either $1.0m or $3.0m, although clearly the former more correctly predicted the degree of the film's failure. But it also won't be fair to say because you overpredict it by 500%, you should get -5 (1 - 3.0/0.5) points on it. The severity should be somewhere in between. Don't have a formula to propose yet.

Thanks!

That helped a lot. Two things stuck out at me:

One being, that I need to find some way to 'end' the prediction time period. Obviously, I could do it where someone manually 'ends' the prediction time period, but I would rather do it automatically, and I'm not sure how exactly to go about that. Any ideas?

The second issue is the equation. I want to make sure I understand it correctly:

Assuming I predicted a movie would make 29 million, and it really made 32 million.

I take 29-32.
I get -3.
I divide -3/32.
I get -0.09375
I take the absolute value to get 0.09375
I subtract 1-0.09375
I get .90625.

I assume that would be the percent based on the current BOM method for one movie. How do I take all 10 and mold them into one percentage? Do I add all the absolute values, in the above example the 0.09375, and then divide by ten. Then take that answer and subtract it from one?

I assume that is correct, I didn't understand how you propose changing the formula. I think that's because the whole min(x,y) is lost on me, so I am unsure what the result should be there. But I do agree, logically, what you suggests makes much more sense.

Theez1 helped me out with the first problem. He suggested to compare the time of the page load to a pre-set deadline time. Which should work out well.

This would be wonderful. Definatelly would bring me back to predicting. And winning ofcourse.

The next problem becomes how to set the required information for the new week. For example:

Past weeks actuals
New weeks movies to predict
Prediction deadline

It's going to have to be entered manually each week I suppose. In all honesty, I just don't know enough about it to do it week to week. I can set up the system and add it to our current back end so that someone can login and update everything from there, but I'll still need someone to do it.

Xiayun, once I had the game running, I was hoping I could convince you to be WoKJ's official predictor. If you have an interest in that, maybe you could also run the game? I could give you full access to the back end, so you'll be able to add and delete users, thus if you need or want help running the game, entering information and the like, you can add someone of your choosing to help you.

If you have any interest in that, let me know. Otherwise, can you suggest someone that may be interested in helping? I know I can't do it week to week on my own.

That works. Or you could add all the 10 scores from how you get the 0.90625 and divide the sum by 10. They're the same. Steps wise, yours is simpler.



What I proposed is, using your example, instead of dividing -3 by 32, we divide it by 29, so the final score will be a little less than 0.90625. This way, if one predicts 16 for it, he would get 0. It may be too severe, but it's a way to discourage the "to-be-safe" way of approach.

No problem, I'd love to take on the responsibility, and hopefully we could find one or more persons to help running the game.

I could help a bit, if it is within my abilities.

A small suggestion I have would be that we shouldn't have the same 10 movies to predict in our game that BOM has. If you could tweak the formula so that overpredicting by 100% doesn't eventually have to end with a 0% score, one could maybe included 1 or 2 limited releases every week.
Or another way to make the game more interesting could be to award bonus points if you have the ranking of the movies in the correct order...

Well Xiayun, I don't know enough about the formulas. I trust your judgment, if you think it works better that way, then we can do it that way. We can always tweak them later on.

well, it's way easier than that, isn't it?

29 / 32 = 0.90625

I love the idea of a new game, but without the incompetence of the careless (non-KJ) Derby players, I may finish dead last!

Sadly, no.

Think about if the person predicts 33 instead of 29.

Loving the idea! Like Nebs, maybe it will bring me back to predicting as well.

Any updates about this?

Also, I'd gladly help with anything needed.

Would be great if the game is ready at new year.

Why do I have a bad feeling this will be another idea that just gets pushed to the side and forgotten?

Heh, calm down. My finals just ended today!

My priority list is:

1) Create and Install new CMS System Started Work
2) Add ability to comment on articles Coded, Needs Modifications
3) Start work on the box office game. Haven't Started

It won't get pushed off to the side. I can't say for sure I'll be able to get it working, but I will be making an attempt this month. I'll update this thread with details as I get started.

So... where on that priority list is quick reply?

Below many other things.

I won't even think about quick reply until the board goes Gold, it's currently in RC8.

Progress Update:

CMS

I now have 4 out of 11 desired functionalities operational. More importantly, I got validation working to guard against SQL injections and other nasty things. Not so important to the CMS (unless one of our writers has their login info hacked), but it will be important to the BO game.

I hope to finish the others tomorrow. My goal was to have the CMS and the ability to comment on articles finished by this weekend. That is still a possibility. Things are going well so far, though taking a bit more time than expected.

BO Game

Nothing really new to report, but as I said above, I have done significant work enhancing the security of forms interacting with our databases. The code written will all be re-used in the box office game, so I suppose I've started, in a way!

I am aiming for January 1st Beta, but I have no way of knowing the reality of that before I get started. As of now I'd say that's ambitious.

Forum Update

PHPBB decided to release the gold version of the forum, so it slots in at priority number 3 right now. I'll upgrade sometime in 2008.

Main Site

I had hoped to finish things like:
- Creating a new Box Office page to fit the new design
- Transitioning search results and archive pages to the new design

As of now, I don't think I'm going to have the time over the holiday. The CMS, while going within reason, has cost me 2 days work already, and countless hours. These updates are likely getting moved back to late January.

UPDATE:

CMS

The new content management system is now fully functional, working, and finished. I have some additional functionality I want to add to it, but that will come later.

What's Next:

Commenting on Articles

I am hoping to add the ability to comment on all of the site's articles at some time tomorrow.

Gradient text.

UPDATE:

Commenting on Articles

The ability to comment on articles is now finished. It works just as I wanted it to, and the implementation went pretty smooth, though it took about two days longer than I imagined.

What's Next:

Box Office Game!

It's about time to get down to the big project of my Xmas break. I'll be making a new thread to detail progress so that anyone interested can follow along and help out. I'll be including active links to what I have working, your input will help alot, so please check it out.