Topic

Better Learning V Math from a Non-Math Major

I just finished learning Better Learning V. With that skill complete, I have exactly 37 skills learned. I started a brand new skill, and time was increased across the board, which proves that Better Learning skills ARE included in the no-time-penality count.

Anyhoo, here were my times before Better Learning V completed:

30m 36s
1h 31m 49s
2h 2m 25s
4h 4m 50s
8h 9m 41s
24h 29m 2s
97h 5m 6s

Here are my times AFTER Better Learning V finished:

24m
1h 12m
1h 36m
3h 12m
6h 24m
19h 12m
76h 48m

For a difference of:

6.6m
19.82m
26.42m
52.83m
1h 45.7m
5h 17.03m
20h 17.1m

For a total savings of:

1d 5h 5m 30s (29h 5.5m)

The problem of course is that each new skill I learn tacks on another 3.0009% worth of time to ALL of the remaining skills. So if I learn the 7 skills posted above, I effectively lose 6h 6.555m of previous time saved.

If this pattern continues, the benefits gained from Better Learning V will be completely LOST by the time you learn your 71st skill (there are 93 skills in total).

So while it appears you are training "faster" with Better Learning V, once you learn your 71st skill, you are at exactly the same point as someone who did not spend ~4 days learning the skill.

But then again, I wasn't a Math major in college. I could be wrong. :)

Thoughts?

Posted 14 months ago by Bashere Subscriber! | Permalink

Replies

1 2 Next
  • @Bluigi, yeah, that would be nice wouldn't it.  Too bad I already did it.  BAM.

    http://agent86.nfshost.com/glitch/skill/

    Also times are now in days/hours/mins/seconds, like glitch's skill view does.
    Posted 14 months ago by agent86 Subscriber! | Permalink
  • Yes, Bureaucratic Arts 3 is also in the API's skill list.  I'm filtering that and the Unlearning skills currently.

    If you're curious about them, you can pull up the data on them at the API's website:

    http://developer.glitch.com/api/explore/#!skills.listAll
    Posted 14 months ago by agent86 Subscriber! | Permalink
  • That agent86 sure is good, faster than the speed of light.
    Posted 14 months ago by Bluigi Subscriber! | Permalink
  • It's so pretty now.  Plus it takes into account any skills you've started.

    http://agent86.nfshost.com/glitch/skill/

    I think I'm pretty happy with this now.  Moving on to the next project...
    Posted 14 months ago by agent86 Subscriber! | Permalink
  • I heart you so hard right now, agent 86 :D
    Posted 14 months ago by Prof. Woland Subscriber! | Permalink
  • Oh my, you really are good at this. Thanks, cause this is great. :)
    Edit: Nevermind the other part, it's perfect!
    Posted 14 months ago by Bluigi Subscriber! | Permalink
  • I was testing the math a bit today and found it to be somewhat off, so I recomputed my personal penalty percentages based on what the official skill tree viewer was telling me.  I think the math is more accurate now.  

    It's been said elsewhere that the Better Learning skills don't count against your skill total, but I'm not sure I believe that based on the math I'm seeing.

    As always, I'm guessing at the algorithm based on its output, so this might still be way off.  The time estimates are there for entertainment purposes only, folks :)
    Posted 14 months ago by agent86 Subscriber! | Permalink
  • Was working for me yesterday, and appears to not be working for me today (I wanted to see this new logic where it helps me plan).

    Anyone else having this issue? I'm on IE9 if it matters.
    Posted 14 months ago by Evadrepus Terramere Subscriber! | Permalink
  • Can you try it again?  Late last night I re-added a fix for IE 8 & 9, but it might be that they fixed the bug in more recent versions of IE9 and the fix breaks something.

    I added a check for the IE version and only included it in IE versions less than 9.  It worked fine in IE8 before, and it works fine in IE8 now, so hopefully IE9 will also work fine :)

    Note that this should work fine regardless in Chrome and Firefox, although Opera doesn't yet support the feature I'm using to make the Glitch API request.  

    If you're curious, this is called CORS, and the browsers that support it are:
    http://caniuse.com/cors
    Posted 14 months ago by agent86 Subscriber! | Permalink
  • Your math is right, Better Learning do count toward their own total, people have done the math to prove that.
    Posted 14 months ago by Bluigi Subscriber! | Permalink
  • Need to check the IE9 version when I get home, but just checked here on the Droid and it works without a hitch. Nice tweak!
    Posted 14 months ago by Evadrepus Terramere Subscriber! | Permalink
  • @agent86  A couple skills are missing their prereqs, such as Saucery 1 needing EZ cooking.

    This came up with a slightly different way than I calc'd, mainly in putting Master Chef sooner than I had planned.  However it came up with a longer time than I very roughly estimate so I think I'll stick with my order.
    Posted 14 months ago by Hameigh Subscriber! | Permalink
  • @agent86: It's not using the algorithm it says it is using.  Or if it is, I have subtle misunderstood what you meant.  Currently, I have everything in the animal kinship line except the last one, which is by far my biggest cost item, and it's not the first thing it picks.  Instead, it puts the entire telepoting series first, which ends up being slower by about an hour.  I don't care much that it isn't optimal (even without being perfect, the tool is a great way to get an approximate order and a time estimate), but I am curious what you did that would produce the current behavior.  Is it the following:

    (1) For each skill (currently available or not), calculate the time it will take when you get to it if you start doing all the prereqs now.
    (2) Use that time in the greedy algorithm.

    If so, this runs into the problem that an expensive skill that you can get immediately will be under prioritized if there is a medium skill that will be more expensive in the future.

    You can get better (but I don't think optimal, especially not with BL in the mix) by doing a bubble type sort on your initial result: compare adjacent elements and if a switch would improve the total time, make the switch.  Unfortunately, the behavior of BL is decidedly non-local, so such a bubble sort won't get you to the "right" place unless you start with BL at the very beginning.

    Actually, I think the following will be an optimal algorithm:

    (1) Do a sort just of the priorities and their prerequisites.  Essentually use step (3)
    For what remains:
    (2) Do a topological sort with the BL series front-loaded.
    (3) Do a bubble sort (making sure to respect the topology)
    (4) Concatenate the priorities list with the remaining list.

    Note that step (3) might be a little subtle, as the topology can mess things up.  Therefore, I think you have to start at the front of the list and only try to move things further towards the front.  You could probably combine step (3) into the topological sort (are there constrained topological sorts that work in one pass? A topological insertion sort?). 

    In any case, if you decide to go through with changing the algorithm (I know you said you're moving on), let me know, I'm curious to what extent it will change the results.  Also, if you make the changes and let me know, I'll give you a diamond.
    Posted 14 months ago by Pazza Subscriber! | Permalink
  • @Hameigh - The Saucery thing appears to be a bug.  I'll have to figure out why that is happening, because it works fine if you mark Saucery I for priority - it correctly identifies the prereq's.  It's definitely NOT an API bug.  

    @Pazza - The algorithm is the quickest way I could write it, and I've already thought of the case you're mentioning - I don't build the prereq tree until I've already picked the skill to learn.  It's 'greedy' in the sense that it always chooses the most expensive skill left and then tries to figure out its prereq's.
     
    Therefore, if you're almost to the end of a long chain, it may prioritize learning a whole 'nother chain because the capstone skill takes the longest out of the currently available unlearned skills.  There are probably 'more optimal' ways to do it, ie, by doing two or more passes over the data, but I think this is probably close enough for most people to get the idea.  You can modify the list it creates to an extent, in order to experiment.  

    I don't mind fixing bugs in it, and I'll probably invest more time at some point working on the quirks in it.  I just might mess around with some other API ideas first :)  I also don't want to invest undue time in something that has only limited interest, so part of the purpose of hijacking (to an extent ;) ) this thread and the rapid prototyping is to gauge interest in a particular idea.
    Posted 14 months ago by agent86 Subscriber! | Permalink
  • @agent86 Fair enough.  Actually, the algorithm I proposed isn't optimal, as it will never move one skill past enough unless the shorter skill was first or the two skills have very nearly the same time.  In particular, it will never moved one tree past another unless everything in one tree takes longer than everything in the other.  For my own personal interest, I will continue thinking of an algorithm, but yes, this is off topic for the original thread, so I will keep my mouth shut unless I have something truly constructive to add.
    Posted 14 months ago by Pazza Subscriber! | Permalink
  • Nah, I don't think we're that far off topic.  The thread started out as a way to figure out if a particular skill ordering was optimal, and we're still discussing that.  We've just geeked out completely :)

    I guess my position is that there are likely more optimal ways of doing a skill sort, but this one's easy, and since there's no money/jobs at stake, I kind of like easy :)
    Posted 14 months ago by agent86 Subscriber! | Permalink
  • Not working for me still on the IE9, but getting an error message this time.

    "Cannot contact the Glitch servers"
    Posted 14 months ago by Evadrepus Terramere Subscriber! | Permalink
  • So with that API thing... I'm done with the rest of my skills in 30 days. In EVE Online... I'm done with the rest of the my skills in about 6 years. 

    Does this mean I win the game in a month?!
    Posted 14 months ago by Bashere Subscriber! | Permalink
  • If your definition of win is learning all the skills then yes. But new skills will be added so it's unlikely that you will really "win" (using your definition) the game.
    Posted 14 months ago by gamecharacter Subscriber! | Permalink
  • IE9 support should be fixed now.  The IE workaround was apparently great for IE8 and was missing something for IE9.  I had to upgrade to IE9 to test and fix this, so let's hope I didn't break IE8 support in the process.  I checked it with IE9's IE8 mode, but lord knows if that's accurate...

    Yay for Microsoft not only not following standards but changing the way they support their version of the standard in different versions for no particular reason.  Not that I'm bitter or anything.  Sigh.

    This version should also fix the problem with EZ Cooking being too far down the list.  I was pruning for duplicates in reverse, therefore leaving the last instead of the first instance of a particular skill in the prereq tree.  That's backwards, if you're keeping score.

    RE: winning, if that's how you define winning the game, then sure!  You don't even need to log in again, and soon you will have won completely, congrats!  Honestly I don't even know what 'winning' this game is, I just like playing it from time to time.  I think making tools and understanding the algorithm is winning more than anything I actually do when playing, so I'm a bad one to ask on the subject, I suppose.
    Posted 14 months ago by agent86 Subscriber! | Permalink
  • Works fantastic on all the ways I can access it now. Thanks Maxwell Smar...er Agent 86!
    Posted 14 months ago by Evadrepus Terramere Subscriber! | Permalink
  • I just did a spreadsheet run of all the remaining skills I have left (not counting BL5), and... well, it looks like I still save almost a WEEK compared to completing the rest of the skill tree without it, even accounting for extra time accrued.

    I would've probably saved even more if I had just started it a couple days ago, when it took even LESS time... albeit it was probably worth picking up Intermediate Admixing between now and then.
    Posted 14 months ago by Jigsaw Forte Subscriber! | Permalink
  • I have a question: I didn't know to learn the long skills or BL skills early, so at this point BLV would take me 12! days to learn. Is it still worth it?
    Posted 14 months ago by Illamasqua II Subscriber! | Permalink
  • Illamasqua,

    It is likely that more skills will be added in the future.  I would say it's still worth it, since it may be useful in the future.
    Posted 14 months ago by WindBorn Subscriber! | Permalink
  • Hmm not a big difference overall. I won't beat myself up over not getting those first.
    Posted 14 months ago by Serra Subscriber! | Permalink
1 2 Next