Why does my score go down sometimes?


#1

I keep open the info page for the puzzle that I am working on. After awhile or when I know that the puzzle has gotten updated scores I note and switch and compare it to the info page before I refresh the info page. Many times the score on the puzzle has gone down from what the info page says.


#2

Hmm. That’s very odd. I will inquire as to why that may be. Might not be able to give you an answer until Monday, however.


#3

also I have noticed that there are long traces that I initially traced because they are easily visible that when consensus comes around there is a gap where my trace is missing. For example:

this first screenshot shows 2 blue consensus segments between orange consensus segments
https://www.dropbox.com/s/rec70ppanplgnhz/trace%201.png?dl=0

this screenshot I have clicked on the Delete Node tool so that the blue consensus segments aren’t shown
https://www.dropbox.com/s/sj9rbym1dn7mv81/trace%202.png?dl=0

there should have been my purple segments still there if mine where off from what other peoples consensus traces were. Maybe this is why my score goes down because some of my traces are being removed when the Consensus is run.


#4

I’ve also noticed both of those issues.
The gaps in my traces seem to appear after the purple traces have turned orange. If I don’t go back and repair them (very tedious), they are likely to be filled in later with blue “community” lines and dots. Or occasionally a comment box will appear urging me to fill in the gaps.


#5

To answer the first question, vogon reports:

"it’s possible for your score to drop substantially if you delete a large branch or if the consensus changes away from your reconstruction due to other players’ contributions. due to the way we reconcile changes in the consensus with your reconstruction, it’s also possible for your score to drop by small amounts (a few points) at any time.

that said, looking at your (Gil) score history, you’ve got a couple drops that seem too large at first glance to explain away with little quirks. we’ll take a look at your reconstructions to see if we can find an explanation, but we don’t currently have a timeline for when we’ll get a chance to do that."

I will work on the second question as well, though we are working on big updates today so your patience on that is greatly appreciated!

Thanks!!


#6
Gil wrote:
also I have noticed that there are long traces that I initially traced because they are easily visible that when consensus comes around there is a gap where my trace is missing. For example:

this first screenshot shows 2 blue consensus segments between orange consensus segments
https://www.dropbox.com/s/rec70ppanplgnhz/trace%201.png?dl=0

this screenshot I have clicked on the Delete Node tool so that the blue consensus segments aren't shown
https://www.dropbox.com/s/sj9rbym1dn7mv81/trace%202.png?dl=0

there should have been my purple segments still there if mine where off from what other peoples consensus traces were. Maybe this is why my score goes down because some of my traces are being removed when the Consensus is run.


Haven't forgot about this. Promise! Still working on finding an answer.

#7

on Mouse-ALM-18 noticed that traces that I would have absolutely traced are showing up with a lot of two node orange then two nodes blue and then two nodes orange, and in-between where I would have traced there is no purple showing my traces. So if you select the Delete Node you get a bunch of orange that have gaps.

example:

https://www.dropbox.com/s/fkcw4sdww5ygioi/trace%20with%20gaps.png?dl=0


#8

I’ll see what I can find out about this.


#9

Gil, the team wanted to know whether your concern is about the gaps in the consensus or whether your concern is about not being able to delete all of the nodes in that branch.


#10

about all the gaps. I didn’t say anything about not being able to delete all the nodes.


#11

Gil wrote:
about all the gaps. I didn't say anything about not being able to delete all the nodes.


Okay. Vogon states that it's possible for consensus to not pick up on a whole branch at a time which would explain the gaps.

#12

If I have traced the whole branch and it is all purple and then the consensus comes around and some of my purple change to orange meaning others have traced the same area that I have. But there are blue segments which means that I didn’t do any tracing in those areas which I did. The blue should never be on that branch that I fully traced, there should be my purple where the blue is. Basically it appears that where the blue is my traces have been removed and I’m not getting credit for those areas. There can be 50 or more of these areas on a puzzle. See glob’s post above stating the same thing happening.


#13

Looking into this.


#14

Gil and glol - This definitely sounds like a bug. Do you guys remember which challenge(s) you were working on when you observed this? Approximate times would be very helpful, too, if you can remember those. Thanks!!!


#15

from one of my earlier posts above I mentioned Mouse-ALM-18, but there have been numerous different puzzles that I have seen this in. No idea about times. I’ll keep an eye out for it and report back then.


#16

Thanks!


#17

I’ve noticed the gap issue on pretty much every puzzle since the consensus began. I referred to it obliquely on 3/20/17 under “My neuron blew up.” I concluded that it did not rise to the level of a big problem/maybe I was being too picky, etc. But as the blue traces become more invasive, it gets really annoying. The latest example I can give you would be yesterday, sometime between 5 & 8 pm PDT on V1-39 & V1-40. At some point, the consensus train zipped through leaving a number of breaks and separations in my previously intact tracing. Sometimes the orange dots remain and just the little orange lines are missing, and other times both are gone. It seems to occur more often to the tracing over fainter areas of the neuron rather than the sturdier parts. Sometimes the gaps are filled in with blue; other times they remain empty.

I’ve only noticed the scoring issue on a few occasions; I specifically remember ALM-17 and ALM-18. It occurred toward the end of the puzzle timeline, when it was nearly complete.


#18

I see something similar on TC-10.

If you zoom in, I assume that you can see some continuous consensus traces and some “dashed” consensus traces.

When I look at TC-10, on some traces I can see a mix of orange and purple, but no blue. That would imply that in the orange areas I “agree” with the consensus, but that there is no consensus on the areas where I still see purple. That does make sense - we must have a little too much deviation in those regions for the consensus program to converge.

But on other traces, I see only broken orange when I zoom in. I know I didn’t trace it that way - I would have made a continuous dot to dot. So the orange are consensus - but what happened to my traces between the consensus locations? I see no blue or purple.

I wonder if it is somehow linked to how densely traces are done. Maybe I left a bit more space between my dots in those regions, so when the consensus is built it is only picking up my dots and leaving blanks?


#19

here is a section of my TC-10:

https://www.dropbox.com/s/8mdwxe96vyr4a6p/very%20patchy.png?dl=0


#20

Thanks for all the info, everyone. We’re working on a resolution for this.