Jump to content

DarkAlf

Member
  • Posts

    182
  • Joined

  • Last visited

Profile Information

  • Interests
    Fire Emblem, mathematics, spreadsheets
  • Location
    Elibe

Previous Fields

  • Favorite Fire Emblem Game
    Blazing Sword

Member Badge

  • Members
    L'Arachel

Allegiance

  • I fight for...
    Order of Heroes

Recent Profile Visitors

1,778 profile views

DarkAlf's Achievements

  1. @DefyingFates The algorithms for both predictors basically work like this: start with the hour 1 totals. This is actually a relatively good assumption, as usually in the first day it's mostly casuals spending flags, and army sizes proportionally in each match usually don't vary an incredibly huge amount. (The second day is a different story, especially if there's a "Team Feathers.") Then for each hour after hour 1, hourly gains are predicted for each remaining hour by taking the hour 1 gain and multiplying by these values: "Score vs. hour 1" multiplier: basically this changes the baseline value, as an army gets larger with more folks joining in. This attempts to account for differences regionally - such as U.S. vs. Europe vs. Japan vs. other regions. "Flag spending" multiplier if army is weaker: whenever an army has the Disadvantage (weaker) multiplier, there will be more flag spending, so this attempts to account for that. I try to account for "Team Feathers" effects with this multiplier here. Current multiplier in effect: this is either the Disadvantage (weaker) or the Endurance (stronger/same) multiplier. The first two multipliers are from a table, where I calculated these values using previous VGs. How these are calculated is a bit of a long story with some statistics involved, so I'll leave it at that. In the main predictor (version A), whenever a prediction is about 10% or more off, the first two multipliers - "Flag spending" and "Score vs. hour 1" - will adjust proportionally upward/downward if the prediction is too low/too high, with the "Flag spending" multiplier adjusted if the prediction is too far off for an army that has the Disadvantage multiplier, and the "Score vs. hour 1" multiplier adjusted if an army has the Endurance multiplier. (It's a bit tricky to try to adjust both at the same time, so I figure it'd be easier to adjust one at a time.) If the predictions are within 10%, then no multipliers are adjusted. In the backup predictor (version B), the first two multipliers are always adjusted no matter the accuracy, and there's actually a little variability built in as well with the multipliers using an assumed normal distribution. I've found that this doesn't work very well, so I removed that in the main predictor. There are long-term plans to try to incorporate machine learning/AI to predicting future values, but that's a long ways off, if it happens. Other ideas include trying to account for longer periods where it's the "same" on both sides to adjust the Disadvantage multipliers, adding an "afterglow" effect, especially for the final hours, among others. Anyway, to cut a long story short, I use past history to attempt to predict each hour's gains based on army size and the multiplier in effect.
  2. Yep - you beat me by literally a few seconds! We posted on the same minute. I was caught off-guard because I didn't think the calendar would arrive for like 3 more days since the previous one hadn't run out yet. Not including orbs from this TT currently running, probably somewhere in the 120s, at least if I projected everything right in my orb/banner spreadsheet (you can view here). I wonder what this special heroes banner is going to be.... There probably will be something coming for the half-anniversary like last year, I reckon.
  3. A little earlier than anticipated, a new calendar is out! What are you looking forward to? Dreading? Enjoy!
  4. A new event calendar has arrived! What are you looking to? Dreading? Have at it!
  5. We have a new event calendar! What are you looking forward to? Dreading? Enjoy!
  6. A new event calendar has arrived! What are you looking forward to most? Dreading? Speculate away!
  7. We have a new event calendar! Who do you think is coming in these banners? Anything you're looking forward to? Dreading? Enjoy!
  8. Congratulations on the +10 Elise!!! That's gotta be very exciting to finish her off at long last!!! I guess I'll share my own +10.... Last April, legendary Alm was announced, and I at the time tried to save to +10 several legendaries at once. I realized that was rather untenable as a F2P, and I didn't have a +10 colorless unit yet at the time, so I decided to focus on him. 100 orbs in April gave me 1 copy (plus legendary Hector and Fjorm); 425 in July for 2 copies (plus 3 Eirs and Caineghis); 304 in October for 1 copy (plus 3 legendary Leifs and 1 fallen Tiki); then insane luck with this last banner - 513 orbs netted me the final 7 copies (plus 2 Dumas and 2 female fallen Corrins). 1342 orbs total to +10 a legendary is amazing luck, honestly. Here he is! Here are my 14 other +10s I've completed since I started playing: Aversa: https://imgur.com/svqZT47 Caeda: https://imgur.com/6yutymY Cordelia: https://imgur.com/ep3B5rQ Eliwood: https://imgur.com/SWyve84 Fae: https://i.imgur.com/4xf8CBI.jpg Gwendolyn: https://i.imgur.com/c57s4JG.jpg Hana: https://imgur.com/sQSB0FJ L'Arachel: https://imgur.com/IFJv6Yk Marth (Masked): https://imgur.com/ZyD3Fej Nino: https://i.imgur.com/XS1B9vj.jpg Peri: https://imgur.com/zzZe0Wa Priscilla: https://imgur.com/sSzsG6j Sheena: https://i.imgur.com/0uVlBCS.jpg Tiki (Adult): https://imgur.com/l9cYMul
  9. The bracket is announced! There is more at stake than normal for the Voting Gauntlet - the winner of this edition will be gifted to everyone who logs in after the conclusion of this event. Bracket: Altina (1st in votes) vs. Winter Sothis (5th) Legendary Leif (4th) vs. Edelgard (8th) New Year's Alfonse (2nd) vs. Legendary Azura (6th) Sothis (3rd) vs. Legendary Alm (7th) Predictors: Version A Version B Who will win? Who will you support? Good luck!! (P.S. I hope I'm not stealing your thunder just this one time, @Diovani Bressan!)
  10. I'm a bit surprised to see a second TT to go with this banner, but happy to get more orbs! This could very well be Farfetched Heroes, but we'll see.... After digesting this calendar a bit, I've come up with 404 (+5 for arena tier 21 and 950 score in Allegiance Battles) or 412 (+5) orbs for this month (link to my orb calendar), depending on if this is Chapter 4 or Paralogue 42 coming up in case this is Farfetched Heroes or something similar. This is the second highest total in a month ever, only to February 2018, which had 437 orbs thanks in part to the 1-year anniversary. I wonder what they have planned for the 3-year anniversary....
  11. We have a new event calendar! What are you looking forward to? Dreading? Anything you wish to speculate? Have at it!
  12. I'm glad to hear it! Thank you and @Maaka and @daisy jane for your input! I finally figured out how to read the data basically instantly, and you'll see that the next edition of the Voting Gauntlet. So no more waiting up to 10-12 minutes for the predictors to update - hopefully they'll be done by 3-4 minutes after the top of the hour consistently. If anyone else has ideas or input, please let me know!
  13. Okay - I had some time to digest this. So December will have a total of likely over 360 orbs (my orb sheet) and January should be very generous since we have a New Year's Event, which probably will include some combination of Daily maps/log-in bonus/other free orbs. Rally Skills banner: Some combination of Lugh, Ophelia (Rally Up skills), Eliwood (Brave), Dream Corrin (female) and Quan - my guess is Ophelia, Dream Corrin and Quan Brazen Skills banner: The only 5* exclusives are Tiki (fallen), Mikoto and Hardin, although I wonder if they'll sneak in Norne, Echidna or Ares, which I doubt Special Heroes banners: Your guess is as good as mine - I have no clue. New Heroes banner: Same - no idea. When January 4 and 5 come, we will have a record 12 (maybe 13) banners minimum those days: 6 special heroes, Brazen Skills, Weekly Revival, Tempest Trials+, Bound Hero Battle (Fae & Idunn), Legendary heroes, and at least 1 (maybe 2) for Voting Gauntlet.
  14. We have a new event calendar! Who is coming in the special heroes banners and the new heroes banner? Have at it!
  15. Hour 1 totals: Barst: 1,140,876 Helbindi: 2,917,349 Mordecai: 1,825,342 Ike: 20,798,708 Hawkeye: 831,245 Duma: 6,947,828 Dorcas: 2,683,558 Effie: 11,705,139 Poor Ike and Duma might not get many multipliers - if at all - this round. Mordecai and Hawkeye are Team Feathers. To anyone using my predictors: I'd like to hear your input. Is there anything you'd like me to add or change? Please PM me with your thoughts (don't reply here so that the thread is not derailed)! I'm open to ideas!
×
×
  • Create New...