DandY

First, I think I was incorrect, I assumed that an IP block was tied to TOS, but you are correct, I cannot find that. So I guess it is a simple DOS attack prevention. 
 
Secondly, I only developed an algorithm because the VJ one was so poor at predicting our dates--it just kept sliding and sliding, and I saw no one else posting that had tried to "build a better mousetrap". So I grabbed Fiddler to capture the request and response to USCIS site so I could mimic that and get my snapshots, then I wrote an algorithm. Mostly out of intellectual curiosity. I decided to share my results here because I thought others might be interested. I posted this without warranty expressed or implied and only defended the numbers in advocating that some thought and logic had gone into the algorithm (and I have given conceptual level info for the algorithm). I am not trying for a job interview. Don't believe me. I don't care.
 
Since I was incorrect in my assumption that others would be interested and it seems this is quite unwelcome--I will no longer post the results of my calculations or my opinions on the estimated wait time. Hopefully this will ease tensions in the forum. Everyone believe what they want and lets call it good.
 
Since you asked Naes, I'll answer your question quickly, then I will speak of it no more.
 
I'm not using the term "snapshot" meaning VM snapshot or snapshot vs. full backup, there are other definitions of the term and I'm using it with this definition:
- a record of the contents of a storage location or data file at a given time.
 
Because my algorithm relies heavily on trend information, I need months worth of data. So I have to have multiple snapshots for the same data set to build average wait and trend information along with calculating the backlog size and average worked per week. Because I need data across over 4 months, I am not pulling full month data sets like you. I'm a pulling data sets for smaller ranges of dates over a longer period and extrapolating. This lets me calculate not just the average wait time across the data, but the trend line which I use for predicting.
 
The downside of this is the until the trend line changes direction, I keep projecting out based upon the current trend line weighted primarily by backlog size and the trend line of average batches worked per week. That is why I was overlaying it with the historical trend line assuming the the "down turn" would happen in about the same place chronologically. This was a false assumption and led to overly optimistic predictions (like my algorithm was predicting high 150s for us 2 months ago because historically it is usually getting shorter in September and October, but that didn't happen this year so I adjusted the algorithm to almost completely ignore historical trend lines--I'm giving that only 10% of the weight it had in my initial algorithm). One other interesting adjustment to make might be calculating the standard deviation as it seem they work about 75% or so of a month (I have not calculated, SWAG for illustration purposed) then move to the next month, thus the average wait time is skewed a little high, given this pattern. 
 
So there is the conceptual level of how it works. If you want to write you own with my approach or develop a different approach have at it, I'd love to see your results. But I'm not discussing this anymore.

First, I think I was incorrect, I assumed that an IP block was tied to TOS, but you are correct, I cannot find that. So I guess it is a simple DOS attack prevention. 
 
Secondly, I only developed an algorithm because the VJ one was so poor at predicting our dates--it just kept sliding and sliding, and I saw no one else posting that had tried to "build a better mousetrap". So I grabbed Fiddler to capture the request and response to USCIS site so I could mimic that and get my snapshots, then I wrote an algorithm. Mostly out of intellectual curiosity. I decided to share my results here because I thought others might be interested. I posted this without warranty expressed or implied and only defended the numbers in advocating that some thought and logic had gone into the algorithm (and I have given conceptual level info for the algorithm). I am not trying for a job interview. Don't believe me. I don't care.
 
Since I was incorrect in my assumption that others would be interested and it seems this is quite unwelcome--I will no longer post the results of my calculations or my opinions on the estimated wait time. Hopefully this will ease tensions in the forum. Everyone believe what they want and lets call it good.
 
Since you asked Naes, I'll answer your question quickly, then I will speak of it no more.
 
I'm not using the term "snapshot" meaning VM snapshot or snapshot vs. full backup, there are other definitions of the term and I'm using it with this definition:
- a record of the contents of a storage location or data file at a given time.
 
Because my algorithm relies heavily on trend information, I need months worth of data. So I have to have multiple snapshots for the same data set to build average wait and trend information along with calculating the backlog size and average worked per week. Because I need data across over 4 months, I am not pulling full month data sets like you. I'm a pulling data sets for smaller ranges of dates over a longer period and extrapolating. This lets me calculate not just the average wait time across the data, but the trend line which I use for predicting.
 
The downside of this is the until the trend line changes direction, I keep projecting out based upon the current trend line weighted primarily by backlog size and the trend line of average batches worked per week. That is why I was overlaying it with the historical trend line assuming the the "down turn" would happen in about the same place chronologically. This was a false assumption and led to overly optimistic predictions (like my algorithm was predicting high 150s for us 2 months ago because historically it is usually getting shorter in September and October, but that didn't happen this year so I adjusted the algorithm to almost completely ignore historical trend lines--I'm giving that only 10% of the weight it had in my initial algorithm). One other interesting adjustment to make might be calculating the standard deviation as it seem they work about 75% or so of a month (I have not calculated, SWAG for illustration purposed) then move to the next month, thus the average wait time is skewed a little high, given this pattern. 
 
So there is the conceptual level of how it works. If you want to write you own with my approach or develop a different approach have at it, I'd love to see your results. But I'm not discussing this anymore.

Quite a presumptuous and pejorative response especially not knowing the qualifications, background, specialty, or algorithms of anyone posting. To be sure beyond a shadow of a doubt that you would develop a better algorithm even lacking this information is quite incredulous.
 
However, in spite of that, I certainly welcome a better algorithm than mine. Mine is now based more on trend line, backlog, and rate of adjudication than my earlier version (which was more skewed to overlaying current trend line on historical trend line to predict). The current version I have now is predicting early May pretty well--but please do post your approach and result as I would be quite interested to see different approaches and their results.

Quite a presumptuous and pejorative response especially not knowing the qualifications, background, specialty, or algorithms of anyone posting. To be sure beyond a shadow of a doubt that you would develop a better algorithm even lacking this information is quite incredulous.
 
However, in spite of that, I certainly welcome a better algorithm than mine. Mine is now based more on trend line, backlog, and rate of adjudication than my earlier version (which was more skewed to overlaying current trend line on historical trend line to predict). The current version I have now is predicting early May pretty well--but please do post your approach and result as I would be quite interested to see different approaches and their results.

That should have been followed by a mic drop!  
 
I'm not sure why the math is so controversial. I've shared the approach to my algorithm and I've never had the approach challenged, just the result. And the challenges never have a better approach advocated, just anecdotal evidence that a case here and a case there demonstrate something. They are outliers. Outliers only demonstrate that there are outliers. I have never argued there are outliers, I've argued against inferring averages from the outliers.
 
 But, if this timeline stuff is too sensitive of a topic, I'm happy to declare a moratorium on timeline discussions. There is no need to get worked up.    

It is not as helpful as you might think because there does not seem to be a dedicated pool of people working by petition type. You can look at the posted numbers of how many they work in a day and simply divide by 28-30 (depending upon how optimistic you are) and they will give you the number of FTE (full time equivalents) that worked cases that day. I have not seen it exceed 6 FTE by much (in other words, they rarely process more than 180 in a day) and many days are 1 FTE or less of work.
 
I have to run the numbers again (or someone may have them), but IIRC they are averaging less than 100 per day over the last month. And there are currently almost 5000 petitions in the queue before any of us (although they only work about 75% before getting into a new month, so you could argue that there are about 3000 in the queue before us, but still that math is not good as that is about 6 or 7 weeks worth of work at least). That means we hit thanksgiving.
 
Just wait for the RFE floodgates to open within the next 30 days.

I take a day or so off an I come back to this Kumbaya moment. Does this mean we are all aligned now and our good friend The Squirrel has come to grips with reality?  

I view the finish line as POE. John might line jump on NOA2, but he has additional logistical issues due to kids and Filipino policy. So while I wish John the best, line jumping in USCIS is unlikely to get him to the finish line before many of us.

One thing is true, we all want the same thing. 
 
My only point is to avoid false expectations. I know personally, that I get very let down when I expect something that does not happen. And I've I've said, 160 day AVERAGE to set your expectations seems like you are setting yourself up for disappointment. VJ says I'm 157-160, I know if I trust that I'll just be disappointed.

One thing is true, we all want the same thing. 
 
My only point is to avoid false expectations. I know personally, that I get very let down when I expect something that does not happen. And I've I've said, 160 day AVERAGE to set your expectations seems like you are setting yourself up for disappointment. VJ says I'm 157-160, I know if I trust that I'll just be disappointed.

One thing is true, we all want the same thing. 
 
My only point is to avoid false expectations. I know personally, that I get very let down when I expect something that does not happen. And I've I've said, 160 day AVERAGE to set your expectations seems like you are setting yourself up for disappointment. VJ says I'm 157-160, I know if I trust that I'll just be disappointed.

This is very true. It really is not worth it to calculate how many are actually active. Maybe it is only 40% of them, so The Squirrel's 3% is right!

I thought we explained that your perception of the functionality of the timeline was incorrect so there was no reason not to fill it out. Failure to update will not negatively impact any statistics here for others.

Realistic analysis of the data is called an "orgy of pessimism", huh?
 
Seriously, you want to run over and tell all the April and May filers that are above 160 that they are mistaken and they have gotten NOA2 already?
 
Lastly, trust me, I'm pro-government, but a business it is not. If it were, this would not happen as there are numerous ways to make this process more efficient and someone would have come in with better performance and a lower cost. 
 
The money we have coming up is an insignificant drop in the budget. Delaying a few months is not going to break the bank.
 
You don't have a timeline, when was your NOA1? I want to put an appt on my calendar at 160 days from then.

I here ya! VJ has mine at 157-160 days. No way it will be that fast, but I'd love it it was!!!

@John & Rose - You got a thank you because you are nicer than me and will answer the same question 20 times.

Have you been reading this thread or did you jump in and ask a question without trying to search or review what has been said? There is discussion all over several threads here about timelines.
 
But, since this post gives me an opportunity to opine, what the heck:
 
Several of us have been saying for many weeks that Christmas was not going to happen. For a realistic chance at Christmas, NOA2 needs to happen in the next week, maybe 2. June filers are still an easy month away for anyone but the "line jumpers" that get lucky and get their petitions worked way out of order. 
 
It seems as things go lately, people fall into three categories:
The "Lucky" (in quotes because none of us are lucky)
The Norm
The Exceptionally Unlucky
 
It seems that there is some percentage (I have not calculated, but gut tells me 10-15%) the get worked "early". I call them the line jumpers. Their petition gets worked 1-2 weeks before the bulk of their date range is getting worked. These are the "Lucky" ones.
 
The Exceptionally Unlucky are the ones the take 1-6+ weeks longer than 90% of the people that filed on their date. For instance, the majority of April has been worked, but there are many people in April who are over 172 days of as of this writing and LOTS of May filers creeping up over 165 days.
 
The Norm is where most people are. It seems the majority of people that file on a given date are worked within the same 8-15 day window. It still feels pretty bad, like for mid-April, that window as an average wait of 146-158 days (if you are on the 158 day end, it does not feel "normal"). 
 
I'm starting to think that John's joke that he was the optimist is unfortunately deadly accurate.
 

Not too much activity here today. I'm going to conclude that depression and utter despair has set in. Buck up folks.
 

I wish I or someone would have tracked RFE percentages. I do have a sneaking suspicion there will be an statistically significant increase in RFEs as timelines approach 170-180 days, but I don't have the data to find out if that is true. We could get from VJ data, but it is less accurate since it relies on user updates. If this does happen, this will push many over the 200 day mark.

This has been hypothesized before and as they top 170 days, it is not an unreasonable hypothesis.

This has been hypothesized before and as they top 170 days, it is not an unreasonable hypothesis.

It will pick up I think when they get a ways into June. So I’m going to stick with my prediction that early June filers will have the longest average wait time in at least 4 years. Trend is not good now. But June is lighter than May so I think late June filers will see average wait times 4 days less than early June filers by my math. 

It will pick up I think when they get a ways into June. So I’m going to stick with my prediction that early June filers will have the longest average wait time in at least 4 years. Trend is not good now. But June is lighter than May so I think late June filers will see average wait times 4 days less than early June filers by my math. 

Don’t worry, The Squirrel says we will have soon!  
 
But I agree with you. The trend line is that they are still falling more behind. Lots of people at 168 and more. We are easily over 170 now. 

Sign me up!!!! To whom do I make the check out to?