Data Entry Form for Steamworks

[Tod Traughber]

My name is Tod Traughber and I am the Public Relations mentor for 3937 Breakaway. I am sharing a link to the prototype for our scouting data-entry form. We would like to open-source our ideas and create streams of collaboration regarding scouting concepts for this year's game.

https://drive.google.com/open?id=0By...05QWjVQUDZqams

Although there are available packages for doing this, we create a Microsoft Access database for data-entry and analysis mainly to teach students concepts and procedures in the use of a database.

I welcome input, constructive criticism, and dialogue regarding our entry form and further questions about what questions to ask of the data and how best to use it to maximize tournament play.

[Rangel]

I like the subjective scouting in terms of speed and accuracy opinions. I definitely think more opinion based scouting is going to be needed than in recent years. Few things I might add would be fuel intake speed, gear placement speed and retrieval station speed(how fast can a robot line up and load up on game pieces).

For ball scoring, we plan to mostly just figure out how many balls each robot can hold and use fire speed and accuracy to measure ball scoring ability. We think it will be too unreliable to count how many balls each robot scores.

For making pick-lists, subjective averages are probably going to be our baseline but final sorting of robots are likely going to be going over each robot individually.

[NShep98]

Some additions I might make are approximate time, i.e. cycle times, how long it takes to place a gear or shoot, etc.

Also for fuel intake add the human player station. I'm not quite sure what "South" is referring to in the auto section.

Good stuff overall!

[MARS_James]

Can I ask what South is in autonomous?

[Tod Traughber]

Thx - North/South refers to the hopper position on the field. North is the hopper closest to the scorers table and South is the hopper closest to the audience position. Yes, I know "The enemy's gate is down", but it helps us reference the attempted hopper without adding another numbered system (we already have peg number and shooting zone number).

We want to know if a robot has the code to adjust where they can fuel up in auto as different starting positions should be able to more easily access the north hopper or south hopper based upon location. If they are coded to only fuel up at a particular hopper, we can adjust and go to the other to maximize ball intake for the alliance.

[NShep98]

Originally Posted by Tod Traughber

Thx - North/South refers to the hopper position on the field. North is the hopper closest to the scorers table and South is the hopper closest to the audience position. Yes, I know "The enemy's gate is down", but it helps us reference the attempted hopper without adding another numbered system (we already have peg number and shooting zone number).

We want to know if a robot has the code to adjust where they can fuel up in auto as different starting positions should be able to more easily access the north hopper or south hopper based upon location. If they are coded to only fuel up at a particular hopper, we can adjust and go to the other to maximize ball intake for the alliance.

Did you mean to include "North" then as well? As of your above post there is only a "South" option and no label for that data field.

[Tod Traughber]

Ball capacity * percentage accuracy is an interesting way to determine shot total, but extremely difficult to measure on a per/cycle basis. In autonomous, maybe not a hard, but how are you going to measure capacity if you have a gear-centric bot who only shoots every third gear cycle in teleop? We figure only the elite teams are going to shoot at a rate beyond a fair guestimate, though of course that remains to be seen.

[Tod Traughber]

Oops yes - there is a North adjacent to the South

[Tod Traughber]

Originally Posted by Tod Traughber

Oops yes - there is a North adjacent to the South

This has been adjusted on the original.

[Rangel]

Originally Posted by Tod Traughber

Ball capacity * percentage accuracy is an interesting way to determine shot total, but extremely difficult to measure on a per/cycle basis. In autonomous, maybe not a hard, but how are you going to measure capacity if you have a gear-centric bot who only shoots every third gear cycle in teleop? We figure only the elite teams are going to shoot at a rate beyond a fair guestimate, though of course that remains to be seen.

You won't. Ball capacity will be tricky. Our plan is to use a combination of pit scouting and visual guessing on the field(bin of fuel is roughly ~50 fuel.).

But my question is, is it more important to figure out exactly what each robot did in every match or what their actual capabilities are? For example, a robot that does mostly gears but ball scoring when able to is going to have a very low average for ball scoring, whereas a team that only does ball scoring might have double or triple the average despite being a far worse ball scorer. For us at least, we will be caring more about how good robots are at each task rather than how many times they do it in a match. Of course I'm sure match data can be useful too. Especially for determining what the robot will probably try to do. I would just avoid arguments of this robot scores this many balls on average vs this robot when discussing who is the best at any particular task.

[Skyehawk]

There is sure a lot going on here. I understand you want to cover all the bases, but you need to to keep an eye on extraneous information. Now maybe I'm dead nuts wrong, maybe you can gather, tabulate, and analyze all that data, taking it all into account. But I've found there are only a few categories one can evaluate at a time. That being said this sheet is built for data acquisition.

For 2017 (before the game was even revealed), I was planning on trying to move towards a more qualitative methodology, for the express purpose of more eyes-on-field time. As long as you have a few unique bases of comparison the desired result of ranking can easily be accomplished. The trick is to know what you are looking for and focus any additional scouting resources on teams that for those criteria. Qualitative to get the big picture, quantitative for close evaluation.

I am glad to see you have a quasi-qualitative implementation for shooter accuracy, I would break it down into; 0-33%,33-66%,66-95%,95+%, but that's just me. The other qualitative fields are also well done, (with too many options your accuracy may go up but precision goes down, with fewer options the data accuracy decreases but the precision rises), I have found 4-5 options in qualitative to be the 'sweet-spot'.

My big takeaway from this sheet is; good organization, it's logical, and complete. It includes some qualitative data points where appropriate. My suggestions are: watch for extraneous information, no sense in collecting it if you are not using it, redefine/adjust on your qualitative distributions to catch the outliers more efficiently, and cut down on the large black areas. your printer won't like it and it will be harder to read.

Overall excellent sheet for logging everything a team is doing over the entire match. Good job.

EDIT: Human error is a huge factor, you will always want to mitigate it with your alliance partners. It may be worth your time to force the user to evaluate a team's human components.

[Tod Traughber]

Originally Posted by Rangel

But my question is, is it more important to figure out exactly what each robot did in every match or what their actual capabilities are? For example, a robot that does mostly gears but ball scoring when able to is going to have a very low average for ball scoring, whereas a team that only does ball scoring might have double or triple the average despite being a far worse ball scorer. For us at least, we will be caring more about how good robots are at each task rather than how many times they do it in a match. Of course I'm sure match data can be useful too. Especially for determining what the robot will probably try to do. I would just avoid arguments of this robot scores this many balls on average vs this robot when discussing who is the best at any particular task.

Understood. I am currently ABD in my quest for my doctoral degree and have struggled with qualitative research and writing. In my quest to be as quantitatively black and white as possible (which is impossible with this game), I am seeking snapshots of performance in order to support my allies and strategically weaken my opponents. Part of the process is minimizing human differences of opinion. We run waves of 18 students in scouting at a tournament with various levels of technical understanding and honestly - interest. Keeping it as yes/no as possible helps focus the picture for the scout team leaders when its 1:00 am Saturday morning and we are punch-drunk trying to determine possible alliance selections.

[Fields]

Originally Posted by NShep98

Some additions I might make are approximate time, i.e. cycle times, how long it takes to place a gear or shoot, etc.

Also for fuel intake add the human player station. I'm not quite sure what "South" is referring to in the auto section.

Good stuff overall!

Probably won't need to do cycle times (amongst the many other items you're watching for)
If they only get 2 cycles of fuel and 1 gear, you can get a good idea of cycle time.

Would want to add notes afterward like they played mostly defense or two robots tried playing def against them the whole time.

[Skyehawk]

Originally Posted by Fields

Probably won't need to do cycle times (amongst the many other items you're watching for)
If they only get 2 cycles of fuel and 1 gear, you can get a good idea of cycle time.

Would want to add notes afterward like they played mostly defense or two robots tried playing def against them the whole time.

Hold the Phone!
Cycle times are indeed important, at a high level of play it is important to know that a team can consistently do 3 consistent fuel cycle at 20 seconds apiece, a gear at 30 seconds, and the rest of the time is spent on 'other', you may be able to squeeze out another gear cycle. It is this reason that averages can be very misleading. When in elims the distribution of resources in an efficient manner is super important, the balance of absolute and competitive advantages dictates the scoring potential more than you may realize...

[Fields]

Originally Posted by Rangel

You won't. Ball capacity will be tricky. Our plan is to use a combination of pit scouting and visual guessing on the field(bin of fuel is roughly ~50 fuel.).
.

Not sure why anyone would consider capacity, speed or accuracy separately at all. If someone can pick up 9 fuel and buck-shot them high 100% with a 5s cycle time or another can scoop 54 and drop in the low goal with 10s cycle the end result is the same. Again same as someone scooping 27 and scores high 33% in 5s.

Main thing I would look for there is end kPa score contribution. Don't get me wrong, that will be tricky in itself (unless you have a counting app you can press), but it does combine accuracy, target, capacity and cycle into one.

Same with gear placements. Total gear placements with a note of did they specifically stop after rotor 2 or 4 (like if they're getting no help from the alliance and shift their focus to fuel) and if they are dropping a bunch in the way.

And of course climb