More about our new accuracy bars
Last Thursday we launched our new UI for accuracy. This brought a few changes, the most important being:
- Your ability to trailblaze cubes is now based on your last 30 cubes.
- Your profile now shows two different accuracy bars, one for Level 1 cells, the other for Level 2. These bars likewise show your last 30 cubes. The threshold for trailblazing and enfranchisement is shown with a tick mark on each bar.
- For Level 1, that threshold is 70%, whereas for Level 2, that threshold is 80%.
- The end result is a more dynamic system where trailblazing ability can be lost more often than once a week, but it can also be gained right back.
Since this launch, we’ve received quite a bit of player feedback. Here in this post we will address some of the most frequent questions and concerns; we understand that many of you would like to know things more in-depth, and we’re happy to explain. Hopefully this impromptu FAQ can help you out if you’ve felt confused!
Q: What do you mean by “last 30 cubes,” exactly? What happens if I trace cubes faster than someone can correct them? Wouldn’t that mean I could trace badly (or well) and never be penalized (or rewarded), if I traced at the right speed?
A: Right now, we’ve started with your last 30 cubes being counted in the order of their submission timestamp. So, for instance, if on a few Mystery cells you submit 30 non-TB cubes (batch A), 10 TBs (batch B), then 20 more non-TBs (batch C), and then you stop playing for a while, when you next check your Level 1 accuracy, cubes from batch B will only factor in if they’ve achieved consensus. Otherwise, you’ll be seeing the calculation from all 20 cubes of batch C, and the last 10 cubes of batch A. If all of those TBs in batch B have reached consensus, then this accuracy bar will have bumped out the last 10 cubes of batch A and factored in batch B.
Therefore… yes, right now, if someone submits 30 non-TB cubes, then 10 TBs, then 100 more non-TBs, and consensus doesn’t form for those TBs in time (which it very well may not), it’s true that those TBs will never factor into your accuracy. After some discussion of this problem, we have resolved that it makes more sense to change your last 30 cubes to be counted in the order of their scoring timestamp. This will mean that no cube is ever unaccounted for in your accuracy, and at our current cell completion rate, your accuracy should only be impacted by TBs or other sub-consensus cubes that are at most a few days old. We will implement this change very soon, and then you can trust that your accuracy bars strictly display the last 30 cubes you’ve played where consensus has solidified.
Q: Why only 30 cubes? For some players that’s an hour’s work, for other players that’s much less.
A: The margin of 30 cubes was chosen because our developers found that using fewer cubes to determine a player’s accuracy gives much better feedback than using lots of cubes. In fact, they found that mathematically the closer to 1 cube we use (meaning using only your last cube to determine accuracy), the better. However, we didn’t want to just use a single cube, thinking that would produce far too variable results for even the best players, so we thought that 30 cubes (an hour’s work for many players) would be enough to still give dynamic accuracy while also giving players a fairly stable accuracy.
The merit of using 30 cubes rather than a higher number is that it is helpful for players who are right on the TB line. Players may work harder at being accurate if they see their accuracy dipping below the line. For more seasoned players, we think that you won’t often see too much of a dip in your accuracy, even if you do a bad cube here or there (whether the fault is in your trace or others’ traces). Please note that at 30 cubes, each cube is only worth ~4% of your total accuracy. So if you usually have high accuracy, even getting 0% on one cube should not likely cost you TBing. If you are getting different results than this (routine dips below that threshold) please let us know! But statistically there is only so much variation possible for 30 cubes.
Q: Why is the TB threshold lower now for Level 1?
A: We know that it might seem unusual to lower the TB threshold if we’re concerned about accuracy. As much as we love accuracy on EyeWire, though, we also keep it in a balancing act with efficiency and fair gameplay. In introducing the new 30 cube rolling window, we want to make sure that cells still grow at a decent rate, especially the already slower-growing Mystery Cells. Likewise, we want to make sure that less experienced players have the chance to contribute to the consensus through TBs and enfranchisement (player weight), if they’re generally good but can’t always stay out of the 70-80% range. Most of all, though, we know that making real-time accuracy more dynamic certainly can run some risk of players who ordinarily stayed above 80% and 90% dropping completely below those points. Until we have a better idea of how everyone’s accuracy may stabilize under this system in the grand scheme of things, we would rather be more generous with TBs at Level 1. We’re confident that because it is fundamentally easier to fall below the threshold, it will still be more difficult for anyone to TB if they let their accuracy slip. You can think of the threshold change as us giving more rewards to those who are equipped to handle them.
As a side note, we should also mention that in terms of enfranchisement alone, EyeWire has always used a 70% accuracy threshold for Level 1. Originally, too, there was an 80% threshold for Level 2, but from late 2014 until now, a simplified global calculation was made where the threshold was at 70% for all cells. We have not noticed a significant difference in the quality of the consensus from that change, and in returning to a 70% vs. 80% split this actually means we’re raising the enfranchisement standards back up for Level 2.
The above might cover everything we’ve been asked about so far, but if it turns out there’s something else to add, we’ll update this post accordingly. Should you have additional questions, you’re still welcome to e-mail support@eyewire.org. We hope the new system does make more sense, however, and we promise that it’s going to get even better before the end of the year. Thank you very much for your patience as we adapt and refine accuracy on EyeWire!