Chief Delphi
Page 4 of 7 < 123 4 567 >
Show 100 post(s) from this thread on one page

Chief Delphi (http://www.chiefdelphi.com/forums/index.php)
-   Rules/Strategy (http://www.chiefdelphi.com/forums/forumdisplay.php?f=6)
-   -   Strategy Sub-Team (http://www.chiefdelphi.com/forums/showthread.php?t=137385)

EricH 14-06-2015 23:00
Re: Strategy Sub-Team
 
Quote:
Originally Posted by artK (Post 1486810)
Why not just not get data points that you can't verify in the pit? It saves time and reduces the amount of noise in your data.
That's kind of the point of the "detector". If you aren't at least looking in the pits, it's kind of odd... but it's not that hard to find a team that is using favorable numbers when you do.

For example, let's say your pit scouts talk to a team about their stacking. And they say that they can do two 6-stacks, uncapped. Sounds reasonable, right? I can see teams going for that this year. But what YOU don't know--necessarily--is that that's their design/non-interference/non-spec (and, in non--RR games, non-defense) field number. The real number, as your scouts find by observation, is that they're actually doing 1 6 stack, or maybe 3 3-stacks, or one 4-stack and one 2-stack.

Basically, it's not the data points that you can't verify that you get from the pit that are the problem--most teams can verify anything they get from the pit scouts. It's that many teams will be optimistic--I don't accuse them of lying outright--but this is a physics problem. (Read: Equations (gameplay) work out nicely in the theoretical frictionless vacuum world, not quite so well in reality.) Then the verification shows that optimism to be misplaced.


My opinion on pit scouting is actually a bit different than most people's. My personal opinion is that it's NOT about the robot when you're scouting in the pit. The only thing you do with the robot is to take a picture (remind the field scouts what this robot looks like), and maybe note what work is being done. The TEAM is far more important, in terms of getting to know them. Let the match scouts collect the data on whether the team works on the field: if two teams work well together and are good friends, and both have decent robots, look out in eliminations if they're on the same alliance.

IronicDeadBird 15-06-2015 10:33
Re: Strategy Sub-Team
 
Quote:
Originally Posted by evanperryg (Post 1486479)
Robot ideas/strategic design ideas can come from anyone, but actual match strategy is done exclusively by the strategy team members. For the sake of simplicity, I will consider strategic design apart from event strategy. Our strategy "subteam" is students from any tehnical/business subteam that is interested in strategy, understands it well, and can communicate effectively in a scouting meeting. Scouters are any student not in the pit, who aren't presenting chairman's, who have been trained on our scouting app. Here's our hierarchy for the scouting team:
Code:
Lead Strategist
        |-------------------------------------------------------------------|
Lead Scout                                                                  |
        |------------------------------------------|                        |
Match Scouters (~20ish)      Qualitative Scouters(3-5)          Pit scouters (1-2)
Lead strategist: Develops match strategy with the drive team. Collects data aggregated by the lead scout, and processes it with Tableau and MS Access. Manages pit scouters, collects information as they see necessary for developing strategy and picklists.Assembles the shift schedule for match scouting before each event.

Lead scout: Develops the match scouting interface, i.e. scouting app or paper sheets. Aggregates data from qualitative scouters and match scouters, to be collected by the lead strategist. Makes sure match scouters and qualitative scouters are focused, comfortable and are attending their shifts.

Match scouters: collect qualitative data using the scouting app. Work 2-3 2 hour shifts per event. Makes sure that the lead scout has collected their data before leaving their shift.

Qualitative scouters: dont have a schedule, but scout for 75-100% of matches out of their own enthusiasm. Collect qualitative data from matches, including notes about what a team did, why they did it, speculation on a team's strategic potential, etc. Extremely important in scouting meetings, as these are the students who see the most matches, and can provide excellent feedback on almost any team.

Pit scouters: collect general technical information on every team at the beginning of the event. They also do not have a predefined schedule, but are called upon by the lead strategist when they are needed. They will be frequently called upon Saturday morning to get last-minute, very specific info needed to make the picklist as strong as possible. Our picklist meetings include the lead mentor, drive team, lead strategist, lead scout, and qualitative scouters.
Ah that makes sense the strategy activity is during comp I take it.

artK 15-06-2015 17:02
Re: Strategy Sub-Team
 
Quote:
Originally Posted by EricH (Post 1486812)
That's kind of the point of the "detector". If you aren't at least looking in the pits, it's kind of odd... but it's not that hard to find a team that is using favorable numbers when you do.

My opinion on pit scouting is actually a bit different than most people's. My personal opinion is that it's NOT about the robot when you're scouting in the pit. The only thing you do with the robot is to take a picture (remind the field scouts what this robot looks like), and maybe note what work is being done. The TEAM is far more important, in terms of getting to know them. Let the match scouts collect the data on whether the team works on the field: if two teams work well together and are good friends, and both have decent robots, look out in eliminations if they're on the same alliance.
I'm not suggesting looking at the pits at all. My opinion about pit scouting is: get a good picture, get a look at the drivebase (just to get an idea of what there driving might look like), record important dimensions (can the robot fit under a pyramid, does it drive long or wide when balancing). It got all the information we could want that match scouts usually can't get by watching a match, without any risk of memory biases being introduced into your data, and was relatively simple (a championship divisions worth of pit scouting could be kept on like 4 sides of paper).

jajabinx124 15-06-2015 17:18
Re: Strategy Sub-Team
 
Quote:
Originally Posted by artK (Post 1486870)
It got all the information we could want that match scouts usually can't get by watching a match
The definition of pit scouting in my book is asking teams questions about their robot that can't be answered through match scouting. Limit pit scouting to data your actually going to use(anything that can't be answered through match scouting) or in other words like you said, keep it simple(actual questions can be complex, but not redundant to what your match scouters are observing). Don't ask teams questions you can answer through match scouting because It'll be a waste of time and It'll be a pain on Friday night going through redundant pit scouting data.

Citrus Dad 15-06-2015 18:30
Re: Strategy Sub-Team
 
Quote:
Originally Posted by jajabinx124 (Post 1486877)
The definition of pit scouting in my book is asking teams questions about their robot that can't be answered through match scouting. Limit pit scouting to data your actually going to use(anything that can't be answered through match scouting) or in other words like you said, keep it simple(actual questions can be complex, but not redundant to what your match scouters are observing). Don't ask teams questions you can answer through match scouting because It'll be a waste of time and It'll be a pain on Friday night going through redundant pit scouting data.
I agree. The last two years we have done extensive pit scouting on a couple of elements that couldn't be discerned from the stands. Last year it was pass through time and reliability because many teams weren't doing this enough to get a good measure (and we walked around with a watch and clipboard). And this year of course we were looking for the ability to add cangrabbers. Perhaps the biggest plus was being able to get to know teams about a specific issue. We both got to know how amenable they were and also just simply connected. We now intentionally try to come up with a specific pit scouting task like this.

EricH 15-06-2015 19:56
Re: Strategy Sub-Team
 
Quote:
Originally Posted by artK (Post 1486870)
I'm not suggesting looking at the pits at all.
"Why not just not get data points that you can't verify in the pit?" Guess you're a stereotypical English-challenged engineer; that's a multi-negative and an unclear prepositional reference in the same sentence. What I initially read was something more like: "Why not just get data points in the pit that you can verify?" If I could rewrite the initial statement to what you meant: "Why not only get data when you're pit scouting that you can't verify on the field?"

And if that is what you meant, then my response is: What data do you mean?

Just about any key data point that you could pick up in the pit--like the ones you mentioned--can also be picked up on the field, given time. The biggest difference is that if you get it in the pit area, you might get it faster (given that teams don't always play all their cards on the field at any one time).

That being said, I'm sure there are exceptions to that: inter-team dynamics, and what improvements they're planning on (or can be persuaded to adopt), are two of the ones that come to mind. Oh, and the all-important one for a top seed: What the odds of them accepting your selection are...

Michael Hill 15-06-2015 20:03
Re: Strategy Sub-Team
 
Quote:
Originally Posted by Citrus Dad (Post 1486888)
I agree. The last two years we have done extensive pit scouting on a couple of elements that couldn't be discerned from the stands. Last year it was pass through time and reliability because many teams weren't doing this enough to get a good measure (and we walked around with a watch and clipboard). And this year of course we were looking for the ability to add cangrabbers. Perhaps the biggest plus was being able to get to know teams about a specific issue. We both got to know how amenable they were and also just simply connected. We now intentionally try to come up with a specific pit scouting task like this.
Ya, we had the weirdest pit scouting questions last year (out of pit scouting questions we've been asked). They were mostly about drive team experience. For example, we asked how much experience the drivers had, how much practice time they had with that particular robot and drivetrain. Took note whether or not the coach was a student or mentor and how many much experience they had. A lot of that is valuable information. For example, our drive team is 100% students. However, when discussing strategies, there were numerous occasions where an adult coach would "bully" the other two student coaches into playing their strategy. If we noted that their coach is a mentor, we also send a mentor to the strategy sessions; not to really participate, but to maintain a presence. We've found that if another mentor is present, those problems tend to go away. We also ask questions about what the intent of the robot (specifically worded to extract intent of the design). That way, we are able to juxtapose that with how they are doing (and if they are performing to expectations).

artK 15-06-2015 23:29
Re: Strategy Sub-Team
 
Quote:
Originally Posted by EricH (Post 1486890)
Guess you're a stereotypical English-challenged engineer;
Yeah. I know I am not a good writer (I usually revise my posts a fair bit to compensate for this, like five minutes just writing this sentence out).

Quote:
And if that is what you meant, then my response is: What data do you mean?

Just about any key data point that you could pick up in the pit--like the ones you mentioned--can also be picked up on the field, given time. The biggest difference is that if you get it in the pit area, you might get it faster (given that teams don't always play all their cards on the field at any one time).

That being said, I'm sure there are exceptions to that: inter-team dynamics, and what improvements they're planning on (or can be persuaded to adopt), are two of the ones that come to mind.
The rationale I have for pit scouting being really simple is because pit scouting usually happens during practice day, and a lot can change between then and drafting. But some things, like a drivebase or height, usually don't change between those points, or are changed for a match for some specific strategy.

With respect for inter-team dynamics and improvements, most of that analysis can be done informally by the drive team and pit crew in preparation for matches, especially since they'll be the ones working directly with them during matches.

GeeTwo 16-06-2015 00:33
Re: Strategy Sub-Team
 
Quote:
Originally Posted by artK (Post 1486918)
The rationale I have for pit scouting being really simple is because pit scouting usually happens during practice day, and a lot can change between then and drafting. But some things, like a drivebase or height, usually don't change between those points, or are changed for a match for some specific strategy.
Pit scouting is also essential for strategy in the first few rounds, especially if a robot didn't play many practice rounds, or calibrated and tested during practice rounds rather than competing. If you base your strategy on an assumption that your adversaries can do 90% of what pit scouts report, and your allies can do 50% of what pit scouts report, you'll usually do better than if you start from scratch.
It may also help the field scouts select which robot each will watch. For example, if Bill and Carol watch most of the landfill 'bots and Mary, Frank, and Ted watch most of the feeder station 'bots, results are likely to be more consistent than if assignments are random.

gblake 16-06-2015 00:56
Re: Strategy Sub-Team
 
I'll bet EricH and ArtK would quickly meet in the middle if they both agreed scouts should collect data that drivers, etc. want and will actually use; and that scouts shouldn't waste time collecting any other data (unless they have time to waste); and that scouts should get the needed data from wherever it is available, regardless of where that is.

If you need data from the pits, go get it. If you don't, don't do scouting there.

People shouldn't have the job of pit scouting, they should have the job of getting needed data; if you know what I mean.

Kevin Leonard 16-06-2015 08:05
Re: Strategy Sub-Team
 
Quote:
Originally Posted by gblake (Post 1486924)
I'll bet EricH and ArtK would quickly meet in the middle if they both agreed scouts should collect data that drivers, etc. want and will actually use; and that scouts shouldn't waste time collecting any other data (unless they have time to waste); and that scouts should get the needed data from wherever it is available, regardless of where that is.

If you need data from the pits, go get it. If you don't, don't do scouting there.

People shouldn't have the job of pit scouting, they should have the job of getting needed data; if you know what I mean.
I agree with you for the most part, but sometimes its difficult to tell at the beginning of the season exactly what data is going to be the most important.

For example, in 2014 Team 20 thought that the amount of time the robot possessed the ball was going to be extremely important, and that our second pick would be largely based on that. As it turned out, that was completely unimportant, and our match scouts did far more work than necessary getting that information, and things like "What drivebase does Team X have?" and "How smart are their drivers and how well do we work with them?", both of which rely at least partially on pit scouting.

However, assuming you have the manpower to make it happen (not something Team 20 lacks), it's always better to have more data than not enough.

If you don't have the manpower to make it happen, either team up with another team so you gain the manpower, or find a way to narrow the data collection to only what is essential.

GeeTwo 16-06-2015 12:45
Re: Strategy Sub-Team
 
Quote:
Originally Posted by gblake (Post 1486924)
I'll bet EricH and ArtK would quickly meet in the middle if they both agreed scouts should collect data that drivers, etc. want and will actually use; and that scouts shouldn't waste time collecting any other data (unless they have time to waste); and that scouts should get the needed data from wherever it is available, regardless of where that is.
Yes, I concur that this thread has reached the point of (verbally) violent agreement.

IronicDeadBird 16-06-2015 13:15
Re: Strategy Sub-Team
 
*INCOMING CONTROVERSIAL STATEMENT*

I'm truly surprised at how intensive some of you guys are with scouting during competition. I mean to be fair this year was the first time I was faced with having to make a pick list. But scouting didn't effect match to match strategy as much as Q'ing up did...

connor.worley 16-06-2015 13:20
Re: Strategy Sub-Team
 
Quote:
Originally Posted by IronicDeadBird (Post 1486968)
*INCOMING CONTROVERSIAL STATEMENT*

I'm truly surprised at how intensive some of you guys are with scouting during competition. I mean to be fair this year was the first time I was faced with having to make a pick list. But scouting didn't effect match to match strategy as much as Q'ing up did...
2nd round picks win worlds...

IronicDeadBird 16-06-2015 13:24
Re: Strategy Sub-Team
 
Quote:
Originally Posted by connor.worley (Post 1486969)
2nd round picks win worlds...
Well that and caffeine...


All times are GMT -5. The time now is 15:47.
Page 4 of 7 < 123 4 567 >
Show 100 post(s) from this thread on one page

Powered by vBulletin® Version 3.6.4
Copyright ©2000 - 2017, Jelsoft Enterprises Ltd.
Copyright © Chief Delphi