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

1235»

Comments

  • Chewable C++Chewable C++ ✭✭✭
    edited April 2018
    Maybe having the percentage in the bars on the distribution rather than the number of trials?

    Yes, that's what I meant when I said "I do intend to enhance the chart a little further so that the tooltip/annotation on it will show the cumulative percent chance to reach that voyage length".

    The vertical axis values are not helpful, so they're not shown, but the tooltip can be customized to be more meaningful.
    The problem with distributions like that is it's almost too much information to get an adequate feel for exactly how it's structured. In order for me to figure out where my "middle 50%" might land, I would need to check the quantities on each bar, back and forth from the middle until I added up to n/2.

    Yes, that's true, if you want to calculate a specific figure that isn't already shown directly. I just feel that this use case is unlikely. What difference would it make to know the 75th percentile value?

    If it's to alleviate confusion about why something seemed to differ from the estimate, there's already a 90%, a 99%, and the graph. If yours is within the 98% range, the first set of output would tell you that, otherwise you could look your time up in the graph using the horizontal axis.
  • Peachtree RexPeachtree Rex ✭✭✭✭✭
    Cumulative % on the bars would get me the information.

    Thanks.
  • Thanks for having this. It is a big help.
  • Peachtree RexPeachtree Rex ✭✭✭✭✭
    edited April 2018
    Something I'm a little surprised is that the variance for each of the humps doesn't increase as refills are added. I would have expected each of the successive refills to have a wider and flatter spread since earlier runs might have gone longer/shorter.

    Edit: And I suppose that it could be getting wider, just not in a way that is particularly human-visible.

    Edit 2:
    I suppose this input best illustrates my concern. Here, I have intentionally skewed my skill inputs pretty severely, which should significantly increase the variance. That increased variance is visible on the first hump (this is much flatter and wider than, say, a 8k/8k/4k distribution).

    I would have expected the 2nd and 3rd hump to be at least as spread out as the first one. I can't really think of any reason the distribution should become more consolidated over time. If anything, each successive distribution should become more spread out over time.

    q1fkzexgpbxj.png
  • Chewable C++Chewable C++ ✭✭✭
    edited April 2018
    Something I'm a little surprised is that the variance for each of the humps doesn't increase as refills are added. I would have expected each of the successive refills to have a wider and flatter spread since earlier runs might have gone longer/shorter.

    Edit: And I suppose that it could be getting wider, just not in a way that is particularly human-visible.

    Edit 2:
    I suppose this input best illustrates my concern. Here, I have intentionally skewed my skill inputs pretty severely, which should significantly increase the variance. That increased variance is visible on the first hump (this is much flatter and wider than, say, a 8k/8k/4k distribution).

    I would have expected the 2nd and 3rd hump to be at least as spread out as the first one. I can't really think of any reason the distribution should become more consolidated over time. If anything, each successive distribution should become more spread out over time.

    q1fkzexgpbxj.png

    Very interesting observation.

    At first I too thought that this can't be right. It seems unintuitive. But, there is an explanation:

    The refill shapes are affected only by the variation in AM from passing or failing hazards that determines how many hazards you can fail before running out (because for those curves, you will always fail hazards).

    The first shape is additionally influenced by the possibility of running out of AM while you are still capable of passing hazards. This is negligible if your chance of passing hazards when running out is low, but at a certain point, it starts to become noticeable (at around this point is when the script warns you)

    The "wasted" skills neatly explain why the first curve is farther to the left.

    The reason for the greater width is less obvious, but it's basically due to an increased number of possibilities - in addition for rolling for hazards, you kind of meta-roll for how much of your skills actually get used - that is, there is an additional risk of ending "early", which expands the graph leftward.

    Edit: also to your suggestion that subsequent refills should have wider distributions - that's not true. After you get past the "might still pass hazards after running out of AM" point, the further curves simply inherit the shape of the previous curves because after that point, there is zero RNG.
  • Peachtree RexPeachtree Rex ✭✭✭✭✭
    I think I need to do a better job explaining my thoughts.

    I'm 100% OK with how the first hump looks. I would expect a wider distribution for a more uneven allocation of skills. Sometimes, you'll luck into more primary/secondary, sometimes less. The modeling there looks spot on.

    I'm suspicious of the subsequent ones. I would expect the following characteristics of the 2nd/3rd hump:
    1. The time between the peak of 1/2 should be greater than 2/3 because some skills are still passing after the first peak is achieved. The model DOES display this (~3hrs between 1/2 and ~2hrs between 2/3)
    2. The distribution of peak 2 should be wider than peak 1. Given 10k primary/secondary skills, hazards are still being passed between peak 1 and 2. Thus, for the same reason peak 1 is widely distributed, peak 2 should be even more widely distributed (it is based both on the distribution of peak 1 as well as some conglomeration of the hazards passed/failed).
    3. The distribution of peak 3 should be identical to peak 2. As you appropriately point out, eventually, it's deterministic (all hazards are failing). As all of peak 2 end points are well beyond the "full fail" stage, I would expect peak 3 to have an identical distribution as peak 2, which is what appears to be modeled.
  • Sounds like we agree on your points 1 and 3 and the current model seems to be working as described if I understand you correctly.

    2. Nope. For the reasons I explained above, but it really is unintuitive. Peaks 2 and 3 behave differently because they recoup the wasted AM from peak 1. You can think of them as waveforms that interfere destructively. The recouped AM cancels out the sides the raises the middle.

    There's RNG in the script too since it actually simulates the voyage. The RNG is more obvious when you draw with finer increments. I can't tell from your screenshot, but if your increment isn't divisible by 3, then you're going to get additional artifacts because the increments won't align with the tick duration.

    If you look at the chart in this case with default (or consistently ramped up) draw settings, they look very similar, as expected:

    me72ydqheirv.png
  • Peachtree RexPeachtree Rex ✭✭✭✭✭
    edited April 2018
    So, I thought on this last night and I think I understand what you were getting at as to why it consolidates and I 100% agree that your model has it right. I think you may have described this in your earlier posts, but I'm going to do my best to put what I'm thinking into my own words.

    Essentially, voyages that fail earlier have more potential to "catch up". The fact that they failed early means they had an "unlucky" run of hazards. If they saw a bad distribution (like 30/20/12.5 across primary/secondary/other) then it is very unlikely that they would continue to see a bad distribution (at least, they are AS likely to have the unlikely distribution as they were the first time around, which is to say, not likely).

    Since they failed earlier (Say, 5 hours), they get 3 hours of Primary/Secondary hazards passing where a voyage that lasted until 7 hours only gets 1 hour of Primary/Secondary hazards passing. That would make the first refill last significantly longer for the "earlier failing" voyage and significantly less for the "over performing" voyage, consolidating the distribution by some measure.

    So, while there might be SOME trials that would deviate more (they were un/lucky and continued to be un/lucky), it is MORE likely that the outliers would see some regression towards the mean.

    I kind of love the subtle complexities of this.
  • Right.

    The "potential to catch up" is important to remember if you're willing to refill AM. The tool will tell you quite accurately at that point how far it will get, though with some RNG left if you had some high skills.

    Btw, at first glance, it almost sounds like you are making a gambler's fallacy (or, fittingly, Monte Carlo fallacy): https://en.wikipedia.org/wiki/Gambler's_fallacy

    But, you're not - your parenthetical saves you :p
  • Finally got around to the tooltips:

    m15vey5lsysn.png
  • Important PSA:

    I've written a voyage crew optimization algorithm for this tool:

    https://forum.disruptorbeam.com/stt/discussion/4964/tool-crew-management-desktop-tool-and-google-sheets-add-on

    This will automatically recommend a crew assignment that will produce the longest (or very close to it) voyage.
  • Fist of all I am in awe at the level of thought and attention to detail i see in the discussions on the forum. Ive been using voyages to boost my chronitons. After all, if a 52 crystal refill gains me 100 or more chroniton it is a deal compared to spending 150 crystals for 50 chroniton. Saw lots of good advice in this thread to maximize my voyages! Still eying that 20 hour voyage...
  • As a lot of people have said before this is a great a mod/program. I was wondering though if it could have an extra segment focusing on just 2 hours. I'm in a position where i can get 6 with my crew all of the time but still struggling to get 8 hours each time. So could there be a segment where you enter in your stats and how much AM you have and it can make an estimate that you will get it or dont. I know the first bit does give you a total estimate but by focusing on just a 2 hour segment it might be more accurate ?
  • As a lot of people have said before this is a great a mod/program. I was wondering though if it could have an extra segment focusing on just 2 hours. I'm in a position where i can get 6 with my crew all of the time but still struggling to get 8 hours each time. So could there be a segment where you enter in your stats and how much AM you have and it can make an estimate that you will get it or dont. I know the first bit does give you a total estimate but by focusing on just a 2 hour segment it might be more accurate ?

    Maybe I don't understand what you're asking, but you can already specify your remaining AM to get an estimate taking that into account. The accuracy is the same regardless. You can increase the number of simulations to make it slightly more accurate.

    If you're on PC, you can use this desktop tool that will compute the best voyage crew out of the crew you have for you: https://forum.disruptorbeam.com/stt/discussion/4964/tool-crew-management-desktop-tool-and-google-sheets-add-on/p1
  • I'm no mathmetician or computer expert, so attempting to decipher these posts is very difficult. However, I gleaned enough from them to learn how to more successfully set up my voyages. Because of that, my past 4hr voyages have turned into 5-6 hr voyages, usually. So thank you to all who have shared. But, please, try to simplify some of this into simple English for us simple folk, lol.

    Also, I'm trying to understand the significance of this 21 number you all keep talking about. Do I divide my total skills by 21? Do I divide each skill by 21?

    Is 21 the new 42?
  • Synthetic CommanderSynthetic Commander ✭✭✭✭✭
    I'm no mathmetician or computer expert, so attempting to decipher these posts is very difficult. However, I gleaned enough from them to learn how to more successfully set up my voyages. Because of that, my past 4hr voyages have turned into 5-6 hr voyages, usually. So thank you to all who have shared. But, please, try to simplify some of this into simple English for us simple folk, lol.

    Also, I'm trying to understand the significance of this 21 number you all keep talking about. Do I divide my total skills by 21? Do I divide each skill by 21?

    Is 21 the new 42?

    Since you're running 5-6 hour voyages, the 21 number people are talking about is after your 4 hour dilemma, take note of the remaining antimatter on your voyage after the 4 hour mark, then divide this number by 21.

    This will give you quite an accurate estimate of the remaining number of minutes on your voyage before you run out of antimatter, and you can set an alarm for yourself to make sure to recall your voyage before it runs out of antimatter to maximize the time.

    Note that this estimate is based on an assumption that you'll fail almost every remaining hazard, while if you're almost able to reach 6 hour dilemma you probably won't be failing all hazards shortly after 4 hours. So, it'll probably be a bit conservative (a good thing).
  • Thank you Synthetic Commander! That was an understandable answer. I appreciate it! :)
  • muwoomuwoo ✭✭✭
    I'm no mathmetician or computer expert, so attempting to decipher these posts is very difficult. However, I gleaned enough from them to learn how to more successfully set up my voyages. Because of that, my past 4hr voyages have turned into 5-6 hr voyages, usually. So thank you to all who have shared. But, please, try to simplify some of this into simple English for us simple folk, lol.

    Also, I'm trying to understand the significance of this 21 number you all keep talking about. Do I divide my total skills by 21? Do I divide each skill by 21?

    Is 21 the new 42?

    Since you're running 5-6 hour voyages, the 21 number people are talking about is after your 4 hour dilemma, take note of the remaining antimatter on your voyage after the 4 hour mark, then divide this number by 21.

    This will give you quite an accurate estimate of the remaining number of minutes on your voyage before you run out of antimatter, and you can set an alarm for yourself to make sure to recall your voyage before it runs out of antimatter to maximize the time.

    Note that this estimate is based on an assumption that you'll fail almost every remaining hazard, while if you're almost able to reach 6 hour dilemma you probably won't be failing all hazards shortly after 4 hours. So, it'll probably be a bit conservative (a good thing).

    According to this I'm 8 minutes shy of getting a 6 hour voyage! So close! :neutral:
  • Ambrosius77Ambrosius77 ✭✭✭
    edited August 2018
    21 is old value, you have to divide the remaining antimatter by 30 to get a safe number of minutes left of voyages.

    That's because you get double hazard check very frequently (2 hazard check in 1 minutes). You can check that every 3rd or 4th hazard check followed by an other hazard check (only 3 or 4 filling event with -1 antimatter between them and no rewards at all)

    Also there are hazard checks that are more harder than have to be. Usually skill divided by 1200 shows that how far in hours you can get with that skill, but sometimes even this is not enough and you fail the hazard.
  • 21 is old value, you have to divide the remaining antimatter by 30 to get a safe number of minutes left of voyages.

    That's because you get double hazard check very frequently (2 hazard check in 1 minutes). You can check that every 3rd or 4th hazard check followed by an other hazard check (only 3 or 4 filling event with -1 antimatter between them and no rewards at all)

    Also there are hazard checks that are more harder than have to be. Usually skill divided by 1200 shows that how far in hours you can get with that skill, but sometimes even this is not enough and you fail the hazard.

    old value? in my final voyage tier (50/50 6 or 8 hours now for me), I divide by 21 every single day, and it's never been wrong for me yet. dividing by 30 would be very conservative, but would probably leave 10 minutes on the table at least every time.
  • muwoomuwoo ✭✭✭
    21 is old value, you have to divide the remaining antimatter by 30 to get a safe number of minutes left of voyages.

    I recently tested it and 21 came out almost exactly correct.
  • Has anybody found this calculator accurate lately? I’ve found that it is pretty accurate with a balanced crew (primary and secondary a couple thousand ahead of other stats) but way over estimates if you try to min/max the primary and secondary traits.

    As an example, if my primary and secondary traits are around 10k and the other traits are 2-4k and starting antimatter is around 2700 the calculator says I will hit 8 hours no problem. In reality my voyages barely hit 6 hours and one I had to recall before 6 hours it was doing so poorly.

    Anyone else experiencing a change in the math?
  • I've found this tool to be very reliable but it is still subject to RNG. You'll notice there are 3 different time estimates, one is with 90% certainty then the times get shorter when you get to 99% certainty time estimate.

    Make sure you're entering your starting AM correctly and don't accidentally choose a different ship, but occasionally you'll get a voyage that just hits your low stat skills a bunch against the odds.

    Also, you'd be surprised how often 3-400 AM can just keep going and going if you start getting lucky RNG.

    If the tool's only estimating 8 hours, then you don't have reliable 8 hours yet. I did not start getting almost 100% 8 hour voyages until it was estimating 8.5-9 hours, and I still (VERY rarely) get one that fails to reach 8 hours from time to time when I have a lot of voyage crew busy during faction events for something like MED CMD where I'm heavily dependent on a few key crew.
Sign In or Register to comment.