Home The Bridge

(Experimental): Next 4 Events Forecast

5000 Quatloos5000 Quatloos ✭✭✭✭✭
edited January 2019 in The Bridge
Updated 1/5/18 with info based on 1/10 Event actual type, and model version 0.3.  This is tricky with a small data sample!   :-)

Updated 1/2/18 11:18 EST with improved and much more parsimonious model.   Use with caution, still under development!   Let's call this version v0.2

I've built a forecast tool to predict the next 4 weeks' event types. The predictions become less accurate the further out in time.

I'll try and keep these listed in reverse chronological order and update with better predictions and actuals each week.

It would be interesting to compare this with folks' heuristic rules of thumb...

Feb 7 Event:
Actual : TBD
1/11: Updated forecast delayed while I work out some software issues...
1/5: Faction (50%), Faction/Galaxy Hybrid (50%)
Is a Rerun (0%)
Is a Mega (100%)

Jan 31st Event:
Actual : TBD
1/11: Updated forecast delayed while I work out some software issues...
1/5: Galaxy (88%), Any Other (4% each)
1/1: Galaxy (67%), Faction (33%)
Is a Rerun (17%)
Is a Mega (62%)

Jan 24th Event:
Actual : TBD
1/11: Updated forecast delayed while I work out some software issues...
1/5: Faction (46%), Faction/Galaxy Hybrid (42%), Galaxy (8%) Skirmish or Skirmish Hybrid (4%)
1/1: Faction/Galaxy Hybrid (67%), Skirmish or Skirmish Hybrid (33%)
Is a Rerun (17%)
Is a Mega (12%)


Jan 17th Event:
Actual : Skirmish/Faction Hybrid :smiley:
1/5: Skirmish or Skirmish Hybrid (92%), Faction (8%)
1/1: Faction/Galaxy Hybrid (67%), Skirmish or Skirmish Hybrid (33%).
Is a Rerun (0%)
Is a Mega (0%)
Has a faction phase? (8%)
Has a galaxy phase? (14%)
Has a skirmish phase? (91%)


Jan 10th Event:
Actual : Faction :smiley:
1/1 Forecast: Faction (67%), Skirmish or Skirmish Hybrid (33%)
Is a Hybrid? (10%)
Is a Rerun (0%)
Is a Mega (0%)
Has a faction phase? (67%)
Has a galaxy phase? (5%)
Has a skirmish phase? (33%)

For the geek-ishly inclined it's a Bayesian network incorporating event types since last June (start of skirmishes), with prediction nodes including number of weeks since the last faction, galaxy, skirmish or expedition component, number of weeks since last rerun, and mega status (in/out and week #). Very first-cut right now but will refine. Once I get some software licensing issues worked out I may publish a DIY version.
Accepted. Mark them, Galt.

Comments

  • Dirk GundersonDirk Gunderson ✭✭✭✭✭
    I tip my nerd hat in your general direction and can’t wait to see how this tool compares to existing prognostication methods.
  • (HGH)Apollo(HGH)Apollo ✭✭✭✭✭
    I would switch the hybrid and skirmish events. We know the jan 31 event will be faction because it will be the beginning of the next mega event month and that is usually a faction event.
    Let’s fly!
  • 5000 Quatloos5000 Quatloos ✭✭✭✭✭
    I would switch the hybrid and skirmish events. We know the jan 31 event will be faction because it will be the beginning of the next mega event month and that is usually a faction event.

    The stats-only tool favors 2/7 as the start of the next Mega rather than 1/31, since 9-week intervals have been a little more common than 8-week intervals. But the sample size is still small so this is highly sketchy!
    Accepted. Mark them, Galt.
  • Dirk GundersonDirk Gunderson ✭✭✭✭✭
    I would switch the hybrid and skirmish events. We know the jan 31 event will be faction because it will be the beginning of the next mega event month and that is usually a faction event.

    The stats-only tool favors 2/7 as the start of the next Mega rather than 1/31, since 9-week intervals have been a little more common than 8-week intervals. But the sample size is still small so this is highly sketchy!

    For the sole purpose of predicting mega event starting dates, how hard would it be to go back to the Borg mega?
  • 5000 Quatloos5000 Quatloos ✭✭✭✭✭
    edited January 2019

    For the sole purpose of predicting mega event starting dates, how hard would it be to go back to the Borg mega?

    Not hard, I can create a tool variant. The cost would be that skirmish predictions will start to get "leaky" since the sample will be biased down and the "weeks since last skirmish" input will have to start imputing missing values prior to June.

    Also an update: I updated the predictions above by forcing Jan 10-24 to be non-mega weeks. It completely reversed the Jan 10th prediction to favor faction (ugh, small data sample). This looks much more like the forecast uncertainty I would have expected ...
    Accepted. Mark them, Galt.
  • Flemming wrote: »
    Just a footnote, Convergence Day is just around the corner.

    I wasn’t playing the game then. I just looked up the event. Is this right, the event packs offered two legendaries?
  • RaraRacingRaraRacing ✭✭✭✭✭
    Flemming wrote: »
    Just a footnote, Convergence Day is just around the corner.

    Oh man ... if they re-run this event it'd be awesome. We didn't have it last year (2018).

    I still have a 1/5 CD Quark sitting around ... and COP Founder Archer (not part of the first event) just dropped for me in the last voyage. :) The community rewards were a Hugh boost to my mediocre crew at the time.

    But, I doubt they re-run this event.
  • Travis S McClainTravis S McClain ✭✭✭✭✭
    This could be an intriguing thing to observe. Neat idea, @5000 Quatloos!

    Regarding mega-events, my gut reading is that 1/31 is the likelier start date for the next one. That would have all four events take place within the month of February. A 2/7 start date would have the final event begin on 2/28 but primarily run throughout the first few days of March. There's nothing to say that DB wouldn't let a mega spill over into an entire weekend of another month, of course, but it seems more in keeping with what they've done before to try to keep megas contained withing a single month.

    As regards the anniversary event, I think since we got "Infinite Contentions" last year instead of "Convergence Day 2", it's likelier we'll get a third anniversary event rather than a replay of either. And I'd bet on it featuring Q, since both previous anniversary events included him.
  • AviTrekAviTrek ✭✭✭✭✭
    Agreed. The augment mega event started May 31st with the rest of the event in June.
  • Mirror CartmanMirror Cartman ✭✭✭✭✭
    Flemming wrote: »
    Just a footnote, Convergence Day is just around the corner.

    Convergence day event 10 Jan.

    First Discovery season two event 17 Jan to coincide with first episode of DSC S2. Expect known DSC2 crew for the event. e.g. Captain Pike, Beardy Spock

    24 Jan either a re-run of a previous DSC1 event, of something unconnected with DSC.

    31 Jan, first part of the first DSC2 mega event .
  • DavideBooksDavideBooks ✭✭✭✭✭
    Flemming wrote: »
    Just a footnote, Convergence Day is just around the corner.

    Convergence day event 10 Jan.

    First Discovery season two event 17 Jan to coincide with first episode of DSC S2. Expect known DSC2 crew for the event. e.g. Captain Pike, Beardy Spock

    24 Jan either a re-run of a previous DSC1 event, of something unconnected with DSC.

    31 Jan, first part of the first DSC2 mega event .

    2 separate Discovery events followed by a Discovery mega event? Never happen. I've seen Discovery and liked it well enough, but you have to know there are Die Hard Anti-Discovery people here and their numbers are not few. 6 Discovery events in a row would invite torches and pitchforks. No, we will have the mega and understandably, but no Discovery-specific events until then. After the mega, we can expect an uptick in Discovery crew throughout the season, but they won't go crazy.
  • My prediction:

    If it becomes too accurate, then some tool at DB will start to make last minute changes to the schedule to ensure that it no longer reflects the prediction (Cause and Effect).
  • Travis S McClainTravis S McClain ✭✭✭✭✭
    If DB did release Discovery content leading up to a Discovery mega, it would not be altogether inconsistent with how they have circulated crew to date:
    erfbqx2d7x4m.jpg

    As for the matter of a single series dominating event content, there were sixteen consecutive events featuring DS9 last year, which encompassed not one but two mega-events:
    05-24-2018 The Longest Day (5* War Correspondent Jake Sisko)
    05-31-2018 Superior Ambition 1: Catching Fire (5* Enemy Lines Sisko)
    06-07-2018 Superior Ambition 2: Turning Point (5* Bell Riots Jadzia Dax)
    06-14-2018 Superior Ambition 3: A Quiet Coup (4* Sarina Douglas)
    06-21-2018 Superior Ambition 4: The Weak Will Perish (5* Augment Commander O'Brien, 4* Mirror Leeta)
    06-28-2018 The Isthmian Games 2 (4* Bestselling Author Jake Sisko)
    07-05-2018 Kosst Kejal (5* Queen Guinevere Kira)
    07-12-2018 Haunted Vessels (4* Breen Disguise Dukat, 4* Weyoun Clone 4)
    07-19-2018 One Incredible Tale (4* Herbert Rossoff)
    07-26-2018 Mistress of the Mirror (4* Festive Lwaxana Troi)
    08-02-2018 The Hand That Feeds 1: New Providers (5* Alternate Future Jadzia Dax, 4* Diseased Changeling Founder)
    08-09-2018 The Hand That Feeds 2: An Even Field (4* Mirror Garak, 4* Admiral Ross)
    08-16-2018 The Hand That Feeds 3: Advance Directive (4* First Goran'Agar, 4* Yelgrun)
    08-23-2018 The Hand That Feeds 4: Life, Liberty, and the Pursuit (5* Changeling Martok, 4* Dominion Dukat, 4* Ikat'ika)
    08-30-2018 Education Is Power (4* Mirror Keiko)
    09-06-2018 A Changeling in Time (4* Army Chaplain Odo)

    Note also that the recurring Legendary for "Superior Ambition" was Bell Riots Bashir. Also, the week before this streak began, the themed pack 5* was Pah-wraith Keiko, making it 17 consecutive weeks of new DS9 crew.

    This is not to say that there should be a 16-week run of any series, and it's certainly not a prediction. Yes, last year was DS9's silver anniversary, but "Hey this thing started a quarter century ago" does not have the same relevance as "this thing is happening right now".
  • 5000 Quatloos5000 Quatloos ✭✭✭✭✭
    edited January 2019
    January forecast updated with a new and hopefully significantly improved model. The data now go back to May 1 2018. For the curious, screenshot of the network itself is attached.

    As a Bayesian network, this isn't a "model" per se, but rather a compact representation of the actual statistics for the last 8 months. The node "[Factor 0]" captures 10 different modes of variability in event sequences.

    Accepted. Mark them, Galt.
  • 5000 Quatloos5000 Quatloos ✭✭✭✭✭

    For the sole purpose of predicting mega event starting dates, how hard would it be to go back to the Borg mega?

    Not hard, I can create a tool variant. The cost would be that skirmish predictions will start to get "leaky" since the sample will be biased down and the "weeks since last skirmish" input will have to start imputing missing values prior to June.

    Update: I tried this and it definitely degraded performance, skewing factions and galaxies even higher. Even without going back, skirmish percentages are low. Under the current network, I would interpret any elevated skirmish likelihood as a significant chance, even if <50% ...
    Accepted. Mark them, Galt.
  • 5000 Quatloos5000 Quatloos ✭✭✭✭✭
    Forecasts at top of thread updated based on announcement of 1/10 Event type...
    Accepted. Mark them, Galt.
  • Another good prediction. Very useful stuff.
  • 5000 Quatloos5000 Quatloos ✭✭✭✭✭
    Another good prediction. Very useful stuff.

    Thanks!!! Unfortunately my next forecast will be delayed due to a software licensing issue. :-(. Will resume as soon as I can!
    Accepted. Mark them, Galt.
  • FetaroFetaro ✭✭✭
    Unfortunately my next forecast will be delayed due to a software licensing issue. :-(
    Can't Quark do something about it?
Sign In or Register to comment.