• Please take a moment and update your account profile. If you have an updated account profile with basic information on why you are on Air Warriors it will help other people respond to your posts. How do you update your profile you ask?

    Go here:

    Edit Account Details and Profile

23MAY2022 SNA/SNFO BOARD

S

smpl_dude

Well-Known Member
Looks like we got some more entries on the google sheet in that time though, so people are starting to trickle in. I managed to get all my paperwork done so now all I should have left to do for my application is go to MEPS.
 
  • Like
Reactions: Fib
D

Dboom85

Banned
I was thinking people were pissed off about the previous board and others were keeping their head down studying and work hard for their next submission
 
C

carviator

SNA
smpl_dude said:
Looks like we got some more entries on the google sheet in that time though, so people are starting to trickle in. I managed to get all my paperwork done so now all I should have left to do for my application is go to MEPS.
Click to expand...
Do you know if active duty people, already did meps to enlist, have to redo it?
 
Maze_soba

Maze_soba

Well-Known Member
Hey y’all, I had too much time on my hands so I tried making an ML model of predicting SNA selection this afternoon.

It currently predicts Y/N with a 93% accuracy and I can try your scores through it if you want.

A couple of huge caveats here with the data:
  • The quality of training data is pretty shit. I could only find 3ish spreadsheets easily on the AW forums, and was too lazy to go and find others.
  • The “waivers” column seems to be a relatively recent addition, so it doesn’t really exist in the previous data. That said, accuracy didn’t improve when I discarded
  • There is a pretty big bias towards “yes” because only the die-hards pilot wannabes post on this forum/spreadsheets and those that do are generally pretty likely to get yes’s. There is a surprising lack of prorec-N’s in the sheets I looked at.
  • The selection yes/no is purely based off of SNA applicants. I didn’t include NFO because that was too much work and it seems you’re pretty likely to get in if you select NFO as your primary.
  • I only had 150ish rows of data from the sheets I found, so I duplicated it on the assumption that the spread of applicant stats is relatively similar across boards, in order to get more training data. So, the existing biases are doubled.
  • Didn’t include college major b/c I didnt' want to deal with it.
  • This is strictly from the spreadsheet data, so it’s not reflecting the “whole person” concept with stuff that can’t be quantified. <- biggest caveat
So with those rather large caveats out of the way, here are some observations:
  • Age IS unsurprising a big factor. Curiously, there’s a dip between 27-29, but I think that’s mostly a dataset issue.
  • PFAR is the next best predictor
  • OAR and GPA are weighted pretty similarly, as is AQR
  • Interestingly, the FOFAR is a better predictor than Flight experience - meaning it can be said with pretty good confidence that flight exp doesn’t really matter.
  • Prior service and sex are also negligible.
Next steps: if any of you guys and gals have extra time, it would be interesting to build out a better dataset together. That means finding the past board spreadsheets, cleaning it, and putting it into this model to train it better. That would yield better results, and I might be able to make a webapp so ppl can look up their own scores.

Using the model, I tried predicting my own prob of prorec-Y and got a Yes with 96%. LMK if you want me to try putting your scores into it and seeing what comes out. Caveats blah blah blah.

The model itself:

34200

Trying my own scores
34201
 
C

carviator

SNA
kuni_fg said:
Hey y’all, I had too much time on my hands so I tried making an ML model of predicting SNA selection this afternoon.

It currently predicts Y/N with a 93% accuracy and I can try your scores through it if you want.

A couple of huge caveats here with the data:
  • The quality of training data is pretty shit. I could only find 3ish spreadsheets easily on the AW forums, and was too lazy to go and find others.
  • The “waivers” column seems to be a relatively recent addition, so it doesn’t really exist in the previous data. That said, accuracy didn’t improve when I discarded
  • There is a pretty big bias towards “yes” because only the die-hards pilot wannabes post on this forum/spreadsheets and those that do are generally pretty likely to get yes’s. There is a surprising lack of prorec-N’s in the sheets I looked at.
  • The selection yes/no is purely based off of SNA applicants. I didn’t include NFO because that was too much work and it seems you’re pretty likely to get in if you select NFO as your primary.
  • I only had 150ish rows of data from the sheets I found, so I duplicated it on the assumption that the spread of applicant stats is relatively similar across boards, in order to get more training data. So, the existing biases are doubled.
  • Didn’t include college major b/c I didnt' want to deal with it.
  • This is strictly from the spreadsheet data, so it’s not reflecting the “whole person” concept with stuff that can’t be quantified. <- biggest caveat
So with those rather large caveats out of the way, here are some observations:
  • Age IS unsurprising a big factor. Curiously, there’s a dip between 27-29, but I think that’s mostly a dataset issue.
  • PFAR is the next best predictor
  • OAR and GPA are weighted pretty similarly, as is AQR
  • Interestingly, the FOFAR is a better predictor than Flight experience - meaning it can be said with pretty good confidence that flight exp doesn’t really matter.
  • Prior service and sex are also negligible.
Next steps: if any of you guys and gals have extra time, it would be interesting to build out a better dataset together. That means finding the past board spreadsheets, cleaning it, and putting it into this model to train it better. That would yield better results, and I might be able to make a webapp so ppl can look up their own scores.

Using the model, I tried predicting my own prob of prorec-Y and got a Yes with 96%. LMK if you want me to try putting your scores into it and seeing what comes out. Caveats blah blah blah.

The model itself:

View attachment 34200

Trying my own scores
View attachment 34201
Click to expand...
I'd definitely be interested in testing out my scores! They're underneath my posts. My low GPA is what worries most. Do you need more info?
 
You must log in or register to reply here.
Top