New Derby Website

[AndyLL]

I'm rewriting the Derby website.

 

Right now it's mostly a new technology update with better flow, better historical game data and better mobile support.  

 

But I'd like to hear every ones ideas on ways to make the Derby better.

 

That includes not only the way the website works for entering/displaying data but also the game itself.

 

 

 

 

[XXRDJisDoctorDoom]

@Legion in Boots @TalismanRing @8wombi7 @M37 @datpepper @glassfairy @Grebacio

[Shawn Robbins]

Awesome to hear, @AndyLL!

 

One (probably controversial) thought/suggestion that has come up recently is the deadline of the game:

 

We went with late Thursday night when first launching back in 2015/2016 because that was the standard during the Mojo days dating back to the early 2000s.

 

I think everyone would agree things have changed a lot since then. More information is available and Thursday previews give a better indicator of weekend guesses than was ever possible 20 -- or even 5, 6, 7 -- years ago.

 

It might be worth considering these new realities in terms of earlier Derby deadlines, by as potentially as much as 24-36 hours, in part to help even the playing field and hopefully expand the player base.

 

The challenge to this is the complication an earlier deadline creates with the game's feature to enter predictions based on % drops from the most recent actuals. 

 

If that can be overcome, it might help the game evolve more as Andy works on his other redesign elements.

 

Granted, shorter deadlines could also mean fewer players -- though the base is small right now as is. Perhaps games could also open earlier than they currently do in order to offset the reduced backend deadline window.

 

Loading new games on Sunday, for example, should be feasible since there will be a decent idea of what the next weekend's top 10 films may be at that point.

 

Anyway... just food for thought. This is a game, after all, so the goal is to keep it fun while also embracing new ideas and hopefully bringing in new players!

[XXRDJisDoctorDoom]

@Shawn Maybe it could be pushed forward to Noon CST Thursday? For the most part that would give everyone 48 hours (or more) to get in numbers.

[Grebacio]

I think Thursday 23:59 Pacific Time could be a good deadline 

[kayumanggi]

Please consider overseas players as well. I'm looking forward to playing again.

[krla]

When do theater counts usually come out? I think I've typically noticed them on Thursday afternoon. 

 

Maybe a nice 'feature' would be the option of predicting based on PTA?

[XXRDJisDoctorDoom]

Just now, krla said:

When do theater counts usually come out? I think I've typically noticed them on Thursday afternoon. 

 

Maybe a nice 'feature' would be the option of predicting based on PTA?

 

Usually the The-Numbers put out theater counts around 3-6PM EST Thursday.

[M37]

I have some more general thoughts written down that I’ll add, but as for an earlier deadline … personally I’m very much against it 

 

The info that comes out on Thursday - Wed dailies, theater counts, and yes, preview tracking - is just so valuable to forecasting that moving the deadline earlier just feels like taking a lot of the skill out of the process and replacing it with feel, or even just luck. (And leaves open the possibility that some have access to it before it’s public, while others do not)

 

Even with all that info in hand by the current deadline, there are many weeks where the best score is still sub-90%, meaning that’s there enough nuance and variance to keep it interesting even when we’re all working with the same advantages 

Edited January 10, 2023 by M37

[AndyLL]

1st look at the new site.  This is still very much in the alpha stage.

 

You'll only be able to look at data for right now... you'll have to use the normal site to enter predictions.

 

Feedback is welcome

 

https://testderby.boxofficetheory.com/

 

[XXRDJisDoctorDoom]

5 minutes ago, AndyLL said:

1st look at the new site.  This is still very much in the alpha stage.

 

You'll only be able to look at data for right now... you'll have to use the normal site to enter predictions.

 

Feedback is welcome

 

https://testderby.boxofficetheory.com/

 

 

Ohhh that's nice. Love the data. Off the dome just 2 things I noticed/question: You have "prediction" as "predition" and is there a way to toggle between "dark" mode and "light" mode? I know for me personally I don't really care for a black background.

[AndyLL]

16 minutes ago, XXR the Conqueror said:

 

Ohhh that's nice. Love the data. Off the dome just 2 things I noticed/question: You have "prediction" as "predition" and is there a way to toggle between "dark" mode and "light" mode? I know for me personally I don't really care for a black background.

 

I just did a quick test and it's doable.  Probably take me some time.

 

[AndyLL]

I know you guys keep a spreadsheet where you can track estimates between prediction close and actuals.

 

On the new site I added an 'estimates' page that allows people to enter estimates and see the results.  

 

Open to suggestions on how to improve that functionality.

[XXRDJisDoctorDoom]

4 minutes ago, AndyLL said:

I know you guys keep a spreadsheet where you can track estimates between prediction close and actuals.

 

On the new site I added an 'estimates' page that allows people to enter estimates and see the results.  

 

Open to suggestions on how to improve that functionality.

 

@Legion in Boots put the spreadsheet together so he may have the best ideas for any additional inputs or functions that could be useful. 

[krla]

For the issue about deadlines, you could keep the deadline as is, but maybe have extra points for people who locked in early. Could be something similar to how scoring works for the derby. If 6 people are in the 'early' group, then you'd go 6 points to first, 5 points to second. Though that's a lot of points, so you could halve them. Then it goes 3 points, 2, 1, 0, 0, 0. And you wouldn't want early birds to be guaranteed these points, so you'd only award them based on their position in that weeks derby. So if first place isn't an early bird, the 3 points don't get awarded. If second is an early one, they'd get an extra 2. 

 

Or could just do a modifier based on how early someone locks in. Start with a 2x multiplier if you lock in the second predictions open, and end at a 1x multiplier at deadline. The multiplier would decrease by roughly 0.02 every hour. Could even remove the deadline for those who come late, and accelerate the decrease to 0.1/hour after Thursday morning.

 

Maybe that's unnecessarily complicated (and exploitable), so you could just slap a gold star next to people who enter early predictions and perform above average. Or a brown star on the last person to lock in their predictions, lol. After all, it's a game about predicting the box office, not about who is most accurate the earliest. So disregard everything I've said! lol

[Cooper Legion]

I don’t think predicting earlier takes less skill — if anything I suspect it takes *more* skill. But it does take different skills, and it’s easy to see how people who are enjoying the traditional deadline might not like the change.    
 

One thing which could be interesting is a simple deadline filter when viewing results. So e.g. someone could set it to midnight and see scores based on what predictions had been submitted 6 hrs before the final deadline. Or set to Wednesday noon and see scores based on what had been submitted by super early.   
 

One issue there however is that there’s no longer a single temporal point of coordination for people to compare their performance against each other.    
 

Ultimately what I’d probably go for if we do anything in terms of the earlier prediction space is 2 deadlines — the current one+an early one (maybe midnight wed). People who prefer the classic deadline could only pay attention to that, people who like the skills involved in earlier could focus more on there, and people who are real diehards ([Antihero chorus here]) could try to refine numbers for both but wouldn’t have any incentive to make like 20 submissions per week as beliefs constantly evolved (technically the continuous filter on submission time kind of encourages this…)

Edited January 11, 2023 by Legion in Boots

[Cooper Legion]

I also find it somewhat strange how often there will be a week when there are like 3 movies that most people really care about the performance of (say, avatar, M3GAN, and puss) because they’re still making good money, but somebody can get like 95% on all 3 and then score poorly because they put 450k each for a pair of about to get dropped smallballs that actually did 380 and 520 or whatever. Basically the volatility of the low grosses tends to make them the most important to winning even as they are the least cared about or interesting to continue spending lots of time thinking about from a normal BO POV.
 

I would love a new scoring rule (or, to be less controversial, a scoring view toggle/extra column 😛 ) that gave more weight to the higher grossing movies on the weekend rather than just 10% for all 10. There are a lot of different options here though — purely weighting by gross might make medium & small ones matter too little, especially on mega OWs. Log weighting gets funky because grosses can go <1M and then end up with negative logs. Could weight based on the log of the smallest discrete input, 10k.  
 

So for a simple example with 3 movies instead of 10, if one movie made 320k, one made 2.56M, and one made 40.96M then:

log 32 =5

log 256=8

log 4096=12

 

So the weights would be:

320k 20% (5/(5+8+12))

2.56M 32% (8/25)

40.96M 48% (12/25)

 

as compared to current uniform weighting where they’d each matter for 33.3% of the score or linear weighting it’d be:

320k 0.73% (32/(32+256+4096))

2.56M 5.84%

40.96M 93.43% 

 

Edit: I added a tab to show how it would have changed things last week (though the answer is amusingly “not much” since it was a relatively evenly distributed week).

 

Edit 2: multiplying by 100 before log to eliminate possibility of negatives compressed it a bit too much in practice. I think weighting by the square root is best I’ve found so far and maybe best simple one. To use the same example from above:  .32^.5~=.56, 2.56^.5=1.6, 40.96^.5=6.4

 

weights of

320k 6.6%

2.56M 18.7%

40.96M 74.7%

 

Also added a tab with this: https://docs.google.com/spreadsheets/d/1BMEPoyDk7TN25-3VIR_jGTOPZlYulc6H6VbkXYDXkp0/edit

 

Edit 3: Squareroot gives over 50% to avatar on its OW is maybe undesirable. Cuberoot seems like a nice balance.

Edited January 11, 2023 by Legion in Boots

[Cooper Legion]

While we’re here, is there an easy way to get the only 2021 week classified under the 2022 season instead? I basically consider those the “first year back” but right now you have to remember that one of the weeks was technically 2021 and go manually account for it.   
 

Or maybe just “all time” getting a drop down menu for “all time” vs “pre-covid” vs “after return.” 2 eras with a big gap I imagine people would eventually like to see them separated without having to manually break out the different years and combine them.

Edited January 11, 2023 by Legion in Boots

[Cooper Legion]

9 hours ago, AndyLL said:

I know you guys keep a spreadsheet where you can track estimates between prediction close and actuals.

 

On the new site I added an 'estimates' page that allows people to enter estimates and see the results.  

 

Open to suggestions on how to improve that functionality.

Poking around the beta some more — looks pretty cool. Probably more thoughts later. One that immediately came to mind is that I like displaying average as it’s own row, but it would also be cool to see a row that shows how it would have done that week to submit the average on each of the 10 movies. Not sure how you’d name that vs the current avg 😛 

[Grebacio]

Regarding deadlines: I think the current deadline is perfectly fine and I don't like the idea of having extra points because of being early