[TOOL] Voyage Estimator (+help us gather data!)

245

Comments

  • For those that do longer missions I have found and normally do.. More than 12 hours mission and I will not finalize for 12 hours (overnighters) I make sure I have 30 plus a few extra since it takes 30 am every time...for longer I do 60 plus as I had one mission elapse with 36 left and had to refill ...
  • Kirian wrote: »
    Some data that I don't think I've seen elsewhere, that may help you.

    There is a fixed cycle to Voyages.

    If we call each event a tick, there's a 28-tick cycle. Every 7th tick is a reward. Every 4th tick is a hazard, except that tick 28 is a reward, not a hazard. In a cycle, you will encounter 6 hazards, 4 rewards, and 18 filler events.

    Each tick is exactly 20 seconds. It does not automatically display as such because display is delayed by various things. Each cycle is therefore 9 min 20 sec.

    The first two hours consist of exactly 360 ticks (2x60x3). This is 12 cycles and 24 ticks. The 24th tick of cycle 13, which would normally be a hazard, is replaced by a dilemma. This then repeats; the 4 hr dilemma replaces the 20th tick of the 26th cycle.

    This explains the 21.2 AM/min number for failing everything. Rewards do not lose an AM. Each 9m20s cycle has 6 hazards (180 AM) and 18 filler (18 AM). 198 AM / 9.333 min = 21.21 AM/min. This is the average, though any given 1 minute may lose as little as 2 or as much as 32 AM.

    These numbers are exact. You can verify in your own log, and I encourage you to do so. This does not account for passing or failing hazards, which almost certainly involves RNG.

    21.2 is not accurate. As i said, i recharged at 1h and 53 min after the dilemma. It was well above the 10th hour, so every hazard was a failure. I had 2800 AM. With 21.2 it meant 132 min(2h 12 min). That would have given me 2 dilemmas, the one that came in after 7 min, the one after 2h(127min) and have 5 min to recall the ship. After the second dilemma i had only 7 AM remaining, so no recall was possible.
    if you divide 2800 by 127 it gives you exactly 22...
  • Chewable C++Chewable C++ ✭✭✭
    edited October 2017
    Kirian wrote: »
    Some data that I don't think I've seen elsewhere, that may help you.

    There is a fixed cycle to Voyages.

    If we call each event a tick, there's a 28-tick cycle. Every 7th tick is a reward. Every 4th tick is a hazard, except that tick 28 is a reward, not a hazard. In a cycle, you will encounter 6 hazards, 4 rewards, and 18 filler events.

    Each tick is exactly 20 seconds. It does not automatically display as such because display is delayed by various things. Each cycle is therefore 9 min 20 sec.

    The first two hours consist of exactly 360 ticks (2x60x3). This is 12 cycles and 24 ticks. The 24th tick of cycle 13, which would normally be a hazard, is replaced by a dilemma. This then repeats; the 4 hr dilemma replaces the 20th tick of the 26th cycle.

    This explains the 21.2 AM/min number for failing everything. Rewards do not lose an AM. Each 9m20s cycle has 6 hazards (180 AM) and 18 filler (18 AM). 198 AM / 9.333 min = 21.21 AM/min. This is the average, though any given 1 minute may lose as little as 2 or as much as 32 AM.

    These numbers are exact. You can verify in your own log, and I encourage you to do so. This does not account for passing or failing hazards, which almost certainly involves RNG.

    This is excellent info.

    It confirms and refines my 21 figure which was based on a variety of consistent data, but I never attempted to refine it down to this level.

    But actually, plugging in your numbers, I get 21.21/min, if I don't factor in dilemmas. Factoring in dilemmas, I get 20.96, which is actually closer to my previous value of 21.

    So I would say that 21 is basically close enough for providing a single estimate without consideration of context (never mind the weird variations people have observed that are likely some kind of server and/or sync glitches).

    Here's the strange thing. Based on the data I've collected (from a variety of sources, not just my own voyages), the actual hazard count / voyage time figure varies in a way that's not well explained by the above figures. The median is 38.18 (and the discretized mode is also very close to that). But your figures predict 38.07 = (360/28*6-1)/2.

    The reported data I have is usually accurate to within a minute. I don't know if that's rounded down or up more on average as I didn't request that amount of accuracy (keeping it simple to encourage more reporting), but this alone wouldn't account for the difference. So, something is up.

    Also, do you know if Dilemmas subtract 1AM?

    EDIT: my original post forgot to consider the refinement this info has on the "filler AM" - instead of 120/hr it's closer to 115/hr
  • Again remember that the 21.2 is an average. Shifting a window of N minutes by 20 sec either way could be up to a 60 AM difference, depending on N and where in the cycle you are. Which dilemma did you recharge after, and what was the timestamp down to the second?

    I should say that the 21.21 is not exact, you're right. The other numbers are, however.

    Also, 7 after a dilemma is plenty of time to recall, you're guaranteed 1:20 before a hazard starts. Just recall immediately, assuming no bugs.
  • Chewable C++Chewable C++ ✭✭✭
    edited October 2017
    Kirian wrote: »
    Some data that I don't think I've seen elsewhere, that may help you.

    There is a fixed cycle to Voyages.

    If we call each event a tick, there's a 28-tick cycle. Every 7th tick is a reward. Every 4th tick is a hazard, except that tick 28 is a reward, not a hazard. In a cycle, you will encounter 6 hazards, 4 rewards, and 18 filler events.

    Each tick is exactly 20 seconds. It does not automatically display as such because display is delayed by various things. Each cycle is therefore 9 min 20 sec.

    The first two hours consist of exactly 360 ticks (2x60x3). This is 12 cycles and 24 ticks. The 24th tick of cycle 13, which would normally be a hazard, is replaced by a dilemma. This then repeats; the 4 hr dilemma replaces the 20th tick of the 26th cycle.

    This explains the 21.2 AM/min number for failing everything. Rewards do not lose an AM. Each 9m20s cycle has 6 hazards (180 AM) and 18 filler (18 AM). 198 AM / 9.333 min = 21.21 AM/min. This is the average, though any given 1 minute may lose as little as 2 or as much as 32 AM.

    These numbers are exact. You can verify in your own log, and I encourage you to do so. This does not account for passing or failing hazards, which almost certainly involves RNG.

    21.2 is not accurate. As i said, i recharged at 1h and 53 min after the dilemma. It was well above the 10th hour, so every hazard was a failure. I had 2800 AM. With 21.2 it meant 132 min(2h 12 min). That would have given me 2 dilemmas, the one that came in after 7 min, the one after 2h(127min) and have 5 min to recall the ship. After the second dilemma i had only 7 AM remaining, so no recall was possible.
    if you divide 2800 by 127 it gives you exactly 22...

    21 has been accurate in my experience (as long as you pass through a dilemma), but I've seen quite a few outliers for effective hazard/hr that have gone in either direction. I was thus far assuming that there is some kind of sync issue and the server sometimes doesn't provide accurate hazard count statistics at the end, but I'm not sure.

    Maybe something did changed recently or there is some factor we're not considering, but our math seems to work pretty well in most cases. Unfortunately, it seems to occasionally not work, so "your mileage may vary" :( Sorry it failed you in that case.

    In your case, you seem to have had 40 hazard checks per hour during your refill span. In the data I have, about 10% of reports had 40 or more per hour (overall). I had earlier theorized that perhaps the rate increases over time, but this does not seem to be the case. I have few reports beyond 8 hours, but many of those do exceed the prediction, although a few do not. I also have a ~12 hour report that fits perfectly.

    One of the outlier reports we also had a manual hazard count from the log, and the manual count didn't match the count reported on the stats screen, so this seems to be at least one hard data point that strange things can happen on the server.

    I'll look into adding stronger disclaimers into my tool for this and consider adding a range estimate like the regular estimate.
  • Chew, I think the discrepancy is in display, perhaps. The ticks are every 20 s, but the display may be off by up to 10 s or so at times.

    Hazard count per voyage time should be 6/9.333, which comes to 38.57 per hour, ignoring dilemmas. Dilemmas remove 0.5 hazards per hour, the 38.07 you post above.. Variability is based on exact timing, however. 30 seconds either way gives a variation of +- 0.4 or so.

    The first hour should have 39 hazards exactly, but that 39th hazard is at the hour mark proper, so might get recorded as 38. The second hour should have 37 hazards if I'm counting correctly. I'll test with my next Voyages, recalling immediately after the first three dilemmas.
  • Kirian wrote: »
    Some data that I don't think I've seen elsewhere, that may help you.

    There is a fixed cycle to Voyages.

    If we call each event a tick, there's a 28-tick cycle. Every 7th tick is a reward. Every 4th tick is a hazard, except that tick 28 is a reward, not a hazard. In a cycle, you will encounter 6 hazards, 4 rewards, and 18 filler events.

    Each tick is exactly 20 seconds. It does not automatically display as such because display is delayed by various things. Each cycle is therefore 9 min 20 sec.

    The first two hours consist of exactly 360 ticks (2x60x3). This is 12 cycles and 24 ticks. The 24th tick of cycle 13, which would normally be a hazard, is replaced by a dilemma. This then repeats; the 4 hr dilemma replaces the 20th tick of the 26th cycle.

    This explains the 21.2 AM/min number for failing everything. Rewards do not lose an AM. Each 9m20s cycle has 6 hazards (180 AM) and 18 filler (18 AM). 198 AM / 9.333 min = 21.21 AM/min. This is the average, though any given 1 minute may lose as little as 2 or as much as 32 AM.

    These numbers are exact. You can verify in your own log, and I encourage you to do so. This does not account for passing or failing hazards, which almost certainly involves RNG.

    21.2 is not accurate. As i said, i recharged at 1h and 53 min after the dilemma. It was well above the 10th hour, so every hazard was a failure. I had 2800 AM. With 21.2 it meant 132 min(2h 12 min). That would have given me 2 dilemmas, the one that came in after 7 min, the one after 2h(127min) and have 5 min to recall the ship. After the second dilemma i had only 7 AM remaining, so no recall was possible.
    if you divide 2800 by 127 it gives you exactly 22...

    21 has been accurate in my experience (as long as you pass through a dilemma), but I've seen quite a few outliers for effective hazard/hr that have gone in either direction. I was thus far assuming that there is some kind of sync issue and the server sometimes doesn't provide accurate hazard count statistics at the end, but I'm not sure.

    Maybe something did changed recently or there is some factor we're not considering, but our math seems to work pretty well in most cases. Unfortunately, it seems to occasionally not work, so "your mileage may vary" :( Sorry it failed you in that case.

    In your case, you seem to have had 40 hazard checks per hour during your refill span. In the data I have, about 10% of reports had 40 or more per hour (overall). I had earlier theorized that perhaps the rate increases over time, but this does not seem to be the case. I have few reports beyond 8 hours, but many of those do exceed the prediction, although a few do not. I also have a ~12 hour report that fits perfectly.

    I'll look into adding stronger disclaimers into my tool for this and consider adding a range estimate like the regular estimate.

    Don't get me wrong, i'm not blaming you. If there were someone to blame, it would be DB cause they won't give us any help and we have to learn even the basic mechanics ourselves.

    I think you're doing an amazing job!
  • Hey Chewy... If you're looking for more data, I have another 16 voyages on my spreadsheet (#24 - #39). that you can add to your "STT Voyage Time Estimator - [C.Jello]" sheet.

    Thanks! I'll copy them over now.
  • Don't get me wrong, i'm not blaming you. If there were someone to blame, it would be DB cause they won't give us any help and we have to learn even the basic mechanics ourselves.

    I think you're doing an amazing job!

    To be fair here, that's standard for every game ever. Blizzard didn't hand-hold us through Molten Core, Neverwinter Nights didn't come with a complete guide to the game, and Nintendo didn't tell you where all the stuff you needed in Legend of Zelda was.
  • DralixDralix ✭✭✭✭✭
    edited October 2017
    Kirian wrote: »
    To be fair here, that's standard for every game ever.

    Besides, the information you worked out here is far beyond "basic mechanics."

  • Kirian wrote: »
    Don't get me wrong, i'm not blaming you. If there were someone to blame, it would be DB cause they won't give us any help and we have to learn even the basic mechanics ourselves.

    I think you're doing an amazing job!

    To be fair here, that's standard for every game ever. Blizzard didn't hand-hold us through Molten Core, Neverwinter Nights didn't come with a complete guide to the game, and Nintendo didn't tell you where all the stuff you needed in Legend of Zelda was.

    First of all, you are comparing a 90's console game to a game that came out 20 years later.
    Second, i'm on the "cadet manual" right now, and there is no mention of voyages...they didn't update the arena to tell what the hell does positioning even mean...and more....and those are basic mechanics.
  • I think we ought to try to stay on topic here ;)
  • Peachtree RexPeachtree Rex ✭✭✭✭✭
    edited October 2017
    Since you've hammered out most of the "how frequently each action occurs" stuff, it looks like the next step to take here is to figure out (more precisely) how skill requirements increase over time.

    I tried out a napkin math estimation of a 1000 skill point starting point, escalating 250 points every 15 minutes, and that got me a lot closer than I though. However, that estimate ended up being about 30-40 minutes optimistic by the end of my voyage.

    Could you add a few more fields on your data collection to note the AM level at each of the dilemma points? I'm not sure if that would be granular enough, but it's a start.

    Edit: The reason this is important is that I think the biggest downfall of the current tool is that it continues to value stats even past their real usefulness. For example, the tool estimates a nearly 32 hour voyage for a primary stat of 100k (and 0 on the others), which is not at all how it would really play out.

    If we have more information about how the AM rates track over time, we can better quantify how difficulty escalates.

  • Since you've hammered out most of the "how frequently each action occurs" stuff, it looks like the next step to take here is to figure out (more precisely) how skill requirements increase over time.

    I tried out a napkin math estimation of a 1000 skill point starting point, escalating 250 points every 15 minutes, and that got me a lot closer than I though. However, that estimate ended up being about 30-40 minutes optimistic by the end of my voyage.

    Could you add a few more fields on your data collection to note the AM level at each of the dilemma points? I'm not sure if that would be granular enough, but it's a start.

    Edit: The reason this is important is that I think the biggest downfall of the current tool is that it continues to value stats even past their real usefulness. For example, the tool estimates a nearly 32 hour voyage for a primary stat of 100k (and 0 on the others), which is not at all how it would really play out.

    If we have more information about how the AM rates track over time, we can better quantify how difficulty escalates.

    I think it's 1500 skill starting point.
  • Peachtree RexPeachtree Rex ✭✭✭✭✭
    I think it's 1500 skill starting point.
    I don't think it's that high because that would mean my AM rates would start going down earlier than I've observed.

    I've made the following spreadsheet up real quick that gives you a little play. I have a local copy, but have left this one wide-open for editing so you don't have to copy it locally:

    https://docs.google.com/spreadsheets/d/1Bm0emOz-eNHU-_YCH741vG4DxRltoyS4tWvA6cee3V8/edit?usp=sharing
  • I think it's 1500 skill starting point.
    I don't think it's that high because that would mean my AM rates would start going down earlier than I've observed.

    I've made the following spreadsheet up real quick that gives you a little play. I have a local copy, but have left this one wide-open for editing so you don't have to copy it locally:

    https://docs.google.com/spreadsheets/d/1Bm0emOz-eNHU-_YCH741vG4DxRltoyS4tWvA6cee3V8/edit?usp=sharing

    I ran a short voyage with low stats and watched every event, to try and see if there was a starting point, but I couldn't discern one, and it seemed to tie in with the estimator's 1200 an hour (20 per minute), starting at 0.

    129 Sci passed at 1 minute and 5 minutes, and failed at 16 minutes
    170 Med stat had its first check fail at 11 minutes.
    389 Cmd passed at 4 and failed at 27 minutes.
    432 Dip Passed at 3 + 8 minutes, its next check was at 32 when it failed
    794 Eng (Silver) passed 5 times, last at 32, 33 and 39 minutes, then failed at 40, 41, 43, 44 and all after.
    678 Sec (Gold) passed 7 times until 21, 23 and 29 minutes, and then failed at 49, 51 and all after.

    This seemed to tie in so well with the 20/minute that I didn't push it for further checks. almost all my checks confirmed the voyage estimator value until around 6/7 hours when there was very occasionally a fail or pass that should have been the other, typically within 5-10% of what it should be though, and in both directions. soon it moved to all fails anyway.
  • Peachtree RexPeachtree Rex ✭✭✭✭✭
    I updated the spreadsheet with a 0 initial and 1200/hr climb. It's only 100 AM off of what my current voyage is at 5:15. Well within the realm of reasonable.
  • the biggest downfall of the current tool is that it continues to value stats even past their real usefulness.

    Yep. I'm well aware of that. I'm sorry I haven't had a chance to improve that aspect, though at least the situation is somewhat rare in practice anyway. I do acknowledge the deficiency in the disclaimers on the tool page:

    "If your pri/sec skills are low or one of them is very high, the estimate will be too high due to that skill still being able to pass more hazards when the voyage ends."

    Modeling this more accurately would require an iterative or stochastic (Monte Carlo) solution. I've been reluctant to do that so far partly because of limited time but also because that would not work in a Google Sheet.

    I really should in the near term at least add a warning if a skill is too high, though.
  • Peachtree RexPeachtree Rex ✭✭✭✭✭
    I'm not sure a MC would be any more useful than just assuming the averages hold true (as I've done in my spreadsheet there). Like I said, the biggest unknown right now is how difficulty changes over time which helps us really dial in on a "6hr" or "8hr" skill cap that we try to hit before stacking the "other" stats.

    From what I've seen, players with primarily 4* FF crew can easily hit 6 hours, but struggle to hit 8hrs (this is at least the boat I find myself in), so having better information on where that cap is can help min/max crew choices.
  • Chewable C++Chewable C++ ✭✭✭
    edited October 2017
    I'm not sure a MC would be any more useful than just assuming the averages hold true (as I've done in my spreadsheet there). Like I said, the biggest unknown right now is how difficulty changes over time which helps us really dial in on a "6hr" or "8hr" skill cap that we try to hit before stacking the "other" stats.

    From what I've seen, players with primarily 4* FF crew can easily hit 6 hours, but struggle to hit 8hrs (this is at least the boat I find myself in), so having better information on where that cap is can help min/max crew choices.

    I think my 1250 is relatively accurate at this point. Pretty sure it's at least accurate +/- 50. And I see no evidence that my assumption of linear difficulty scale is wrong either, as the graphed out estimate error vs voyage time don't seem to have any obvious patterns - it looks relatively flat (just added a trendline to mine and it's only slightly downward sloping - which is actually the opposite of what I expected, but I think that slope is essentially insignificant anyway).

    A Monte Carlo would solve the problem of not taking hazard passing after voyage ends into account as well as be able to provide confidence intervals specific to the inputs, rather than on average like mine currently does. E.g. if you have high "other" skills and lower pri/sec, your confidence interval should be wider.
  • i use the estimator regularly and find it pretty accurate for most of my voyages. but that's only if all my skills are above 2000. when one or two skills are lower, it seems there are far more fails in the first 4 hours than the estimator calculates. currently i've got a voyage running with primary 10349, secondary 2788, others 4465 1701 3328 1325 and AM 2850. AM running out at 5:45, while estimator says 6:54. the google-docs-table got it right.
  • FYI, as someone else mentioned, being left with a very small amount of am does not guarantee a successful recall. It might be a bug, but I too have encountered being unable to recall even though I have am left. In my case I had around 150 but clicking recall only gave a cancel mission option saying that all rewards would be lost. I backed out and tried several times but the recall message kept saying that, I finally accepted and of course the mission failed.

    I did not try to restart the game... if anyone else runs into this, I would suggest trying that.
  • Chewable C++Chewable C++ ✭✭✭
    edited October 2017
    culthero wrote: »
    i use the estimator regularly and find it pretty accurate for most of my voyages. but that's only if all my skills are above 2000. when one or two skills are lower, it seems there are far more fails in the first 4 hours than the estimator calculates. currently i've got a voyage running with primary 10349, secondary 2788, others 4465 1701 3328 1325 and AM 2850. AM running out at 5:45, while estimator says 6:54. the google-docs-table got it right.

    What do you mean by "the google-docs-table got it right"? I just checked and my sheet gives the same answer as the estimator tool (6:52)

    The reason for the overestimate there is actually because your primary skill is much too high. The simplified estimator math assumes that when the voyage ends, you can no longer pass hazard checks, but your primary skill would be able to continue passing hazard checks for hours.

    I've just added a warning to the tool if a skill is too high. This is based on the assumption that the RNG range is +/- 15% and may not be entirely accurate as that variable is based on a small number of observations and hasn't been refined.
  • i ran across your tool from the wiki... thank you! this is very helpful
  • I've updated the estimator tool to take skill overages into account. It should no longer give wildly high estimates if some of your skills are too high. (for all of my test data that reported a low unused AM, the estimate error is now within 10%)
  • Could you plz explain (the % of your average proficiency roll relative to your average total roll - increase this when using gauntlet crew) a little. I'm not native english. I'm reading this string again and again and still can't understand what I should do :)
  • Chewable C++Chewable C++ ✭✭✭
    edited October 2017
    Deonmag wrote: »
    Could you plz explain (the % of your average proficiency roll relative to your average total roll - increase this when using gauntlet crew) a little. I'm not native english. I'm reading this string again and again and still can't understand what I should do :)

    I'm still debating whether to keep that setting exposed or not.

    I'm not sure it would be of significant use to anybody.

    For example, if you have 2 crew (not going to use all 12 as that's too cumbersome for an example):

    1. 900-1200
    2. 800-1000

    Then the value for that setting is: (1200-900 + 1000-800)/2 / ((900+1200+800+1000)/4) = 25.6%

    The wider the proficiency range, the higher that setting.

    The setting will only make a difference if one of your skills is high enough to still be passing hazards when the voyage ends. It's used to more accurately compute the amount of hazards that skill will actually pass.

    Although for even greater accuracy, the setting ought to be per skill - but that would be even harder to calculate :p

    The default value is roughly the average across all crew in the game.
  • I'm still debating whether to keep that setting exposed or not.

    I'm not sure it would be of significant use to anybody.

    Although for even greater accuracy, the setting ought to be per skill - but that would be even harder to calculate :p

    The default value is roughly the average across all crew in the game.

    The only way that you could do this, would be to allow entry of all 36 stats (12 crew * 3 stats). The page would get complex, but you would have more data to work with as far as calculating the proficiency and such.

  • WizzdomeWizzdome ✭✭
    edited October 2017
    how do you calculate %proficency??? the rest seems a little straight forward but I dont get that one...

    also wanted to add a thought to your data... at 4hrs 50 min a skill check of 4230 failed... could the programming for skill check be as simple as the hour 4, and the minute 50 then add a 0 to it so at 5hrs the check = 5000???
Sign In or Register to comment.