Predicting the winner of the Battle of the Obelisk
Want to win a boon every 2 weeks? Going after the 6 available avatars? Don't know which team you should pick? This page is for you!
To recap the basics, every 2 weeks, 3 of the 6 Obelisk teams fight it out for control of the Obelisk, hence the name. You have to choose a team to battle with, and whichever team does best wins.
Because the winners each week are not randomly generated, it should be possible to predict the winners more often than random chance would do, and thanks to the archive at SunnyNeo, I have created a model to predict the winner each round based on 2 factors that does this pretty well. These 2 factors are:
How often they win
As some teams are judged to give better boons than others, and thus are selected to win more often by the regular battlers, this is the simpler of the 2 factors to calculate. Simply take the number of times a team has won, and divide it by how often they battle. The Sway win roughly half of their battles, while the Thieves Guild and The Order of the Red Erisim win about 1 in 5 battles. Anyone can get a decent win% by picking the best of these teams each time, but that's boring, especially if you're not a fan of The Sway, or want to get all the avatars, and we can do much better. How?
How long since they last won
Because there are avatar hunter about who want every team to win, this factor has to be taken in to account. I take the number of rounds since a team last won and calculate a median win round when I would expect this team to have won 50% of the time by that time, and put this number in to a t-distribution to get a probability that the team would have won at some point between their last win and now. How do we use this data to predict winners?
Here's the exciting part, I set the team that won most recently to 0, and then I take a constant that represents a minimum % chance of winning (about 5%), and add the 2 variables (after scaling them for prediction accuracy) listed above to get my model's output. Then, once I know which 3 teams will be participating, I scale the outputs so that the sum of percentages adds to 100%. I'll share the results that my model generates every week here.
This week's predictions
|Obelisk Team||Last Win||Win %||Prediction|
|Awakened ||7 rounds ago ||33% ||43% |
|Brute Squad ||3 rounds ago ||31% ||16% |
|The Sway ||4 rounds ago ||55% ||40%|
Last updated 23/10/2017
How do I interpret the table?
The 1st column shows the names of the Obelisk teams, the 2nd column lists how many rounds ago that team last won, the 3rd column shows how often that team wins on average, and the final, highlighted column is the one that contains my estimated predictions for the next battle.
Who should I battle with?
That's a decision only you can make, whether you want to win more often, you're going for avatars or you want specific boons, I hope that my model will help you make an informed choice and that you get the result you're after.
The team you picked lost!
I can't guarantee that my preferred team will win (If I could then I'd be using my talents to make money in the real world), but I believe that if you follow my advice you'll win more often than following any other publicly known strategy.
Why don't the numbers in the prediction column add up to 100%?
The numbers are rounded to the nearest percent.
I picked the winning team but didn't get a reward?
Did you remember to win 10 battles? Don't forget to do this or you won't win anything!
Do you have a badge so I can link to this petpage from mine?
Yes I do, there's a choice of 2 here so take your pick!
My question isn't in this FAQ!
, and I'll answer it for you.