Log in

View Full Version : Results and Reactions from STAMP at UCF


buss
11-03-2006, 22:13
So at UCF I was running STAMP (http://stampscouting.org) , as well as an open access point so any team could look at our data (I commented out isUserLoggedIn() at the top of the rankings.php file).

I don't know how many people used our data, but my team (SPAM 180) found it to be extremely valuable. We used the statistics generated from STAMP to pick our alliance partners (who were not ranked very high) and we ended up winning the regional! Yes, I have proof that STAMP works! We made a list of the best fit for the robots we needed. Being able to sort by a certain column in rankings is the most valuable thing to me. We were able to pick two teams that got on the ramp nearly every time and then we sorted by balls scored.

The way we ran it was like so: I set up a debian linux server and installed stamp dev. v 1.3 (not public yet). Everything was running smoothly and the AP was open so like I said earlier, anyone could view our data. This server was put off in a corner on the second floor to blanket the stands in wifi, I also boosted the power from 21mw to 251mw via the Alchemy firmware for the Linksys WRT54G. Team 180 SPAM has always had a huge scouting team and so we did a bit of a hybrid of computer and paper scouting. I found that this system worked the best. We had a scouting paper for each team kept in a labeled folder which was distributed just prior to the match to one of two scouting teams (they work in shifts and don't overlap, that way they get about one and a half hours to watch and relax). After each match, the scouts would hand the cards back to me and I would input the data into STAMP. It took me about the time to put the data in that we had between matches. After putting the data in, I would put the folders back into a tote for easy access for their next match. After this, if you don't have a helper to do it for you, make sure that you have the next three or four match scouting cards queued up for quick distribution. If everything goes as planned, you'll be like me and not have any downtime for about five hours :P.

Ok, so the benefits:
The statistics are incredible. Check out http://stamp.spamrobotics.com and just click the rankings page in the menu. There are a few bugs and Ranking Points and Record aren't working due to the server transfer (yeah...that's odd, I know). Try sorting by different columns. This data was completely freely available to any team who wanted it. I had a (small) sign hung up advertising the data, but I don't think many people saw it. We used the data at our scouting meetings to analyze our competition and get some idea of what we were going to face. My team mentors loved having all this data.

And the drawbacks:
The scouting input interface needs to be reworked. As JoeChrz said in this thread (http://www.chiefdelphi.com/forums/showthread.php?t=43174&page=2#post468352), the + and - buttons aren't really of any value. Scouts can just not focus on a computer and the match at the same time. A piece of paper is much better for counting because you can just make tick marks and you have tactile feedback of where you are making the marks. Ac computer is just not good for this task. The match predictor wasn't as useful as I hoped for, it turned out to be pretty inaccurate, but it gave us some laughs (we predicted a loss of 67 to 7 against us :P). It is not easy to set up if you've never done anything like this before. We can tell ourselves "Oh, you just need to install this, and that, etc" but the truth is, for people who don't breathe this stuff, it is hard. This is a problem with any application that requires some sort of web server, though. I have a few ideas on how to solve this, but it won't be solved next year for version 2007 unless I have some sort of breakthrough and devour some information.

So what's my overall opinion?
I may be biased (I am a developer, after all), but I love it. It actually performed a bit worse than I had hoped, but that just means it will get better. The absolute truth is that it helped us pick the best teams for our alliance, which enabled us to win the regional. By the way, thank you team 79, Captain Krunch (http://www.krunch79.com/) and team 1649 (sorry, I couldn't find your website or your team name!).

Rombus
12-03-2006, 01:20
im actually a little disappointed we were not able to get STAMP up and running at buckeye. It was one of those really unplanned 3 days ahead of time things i wanted to setup quick when we got there. Unfotunitally, we had a deluge of robot controller issues that took almost all of my time thursday so i had no chance to get the wireless AP up and running, nor install stamp on the desktop we had.

Ethulin
12-03-2006, 02:34
So at UCF I was running STAMP (http://stampscouting.org) , as well as an open access point so any team could look at our data (I commented out isUserLoggedIn() at the top of the rankings.php file).

I don't know how many people used our data, but my team (SPAM 180) found it to be extremely valuable. We used the statistics generated from STAMP to pick our alliance partners (who were not ranked very high) and we ended up winning the regional! Yes, I have proof that STAMP works! We made a list of the best fit for the robots we needed. Being able to sort by a certain column in rankings is the most valuable thing to me. We were able to pick two teams that got on the ramp nearly every time and then we sorted by balls scored.

The way we ran it was like so: I set up a debian linux server and installed stamp dev. v 1.3 (not public yet). Everything was running smoothly and the AP was open so like I said earlier, anyone could view our data. This server was put off in a corner on the second floor to blanket the stands in wifi, I also boosted the power from 21mw to 251mw via the Alchemy firmware for the Linksys WRT54G. Team 180 SPAM has always had a huge scouting team and so we did a bit of a hybrid of computer and paper scouting. I found that this system worked the best. We had a scouting paper for each team kept in a labeled folder which was distributed just prior to the match to one of two scouting teams (they work in shifts and don't overlap, that way they get about one and a half hours to watch and relax). After each match, the scouts would hand the cards back to me and I would input the data into STAMP. It took me about the time to put the data in that we had between matches. After putting the data in, I would put the folders back into a tote for easy access for their next match. After this, if you don't have a helper to do it for you, make sure that you have the next three or four match scouting cards queued up for quick distribution. If everything goes as planned, you'll be like me and not have any downtime for about five hours :P.

Ok, so the benefits:
The statistics are incredible. Check out http://stamp.spamrobotics.com and just click the rankings page in the menu. There are a few bugs and Ranking Points and Record aren't working due to the server transfer (yeah...that's odd, I know). Try sorting by different columns. This data was completely freely available to any team who wanted it. I had a (small) sign hung up advertising the data, but I don't think many people saw it. We used the data at our scouting meetings to analyze our competition and get some idea of what we were going to face. My team mentors loved having all this data.

And the drawbacks:
The scouting input interface needs to be reworked. As JoeChrz said in this thread (http://www.chiefdelphi.com/forums/showthread.php?t=43174&page=2#post468352), the + and - buttons aren't really of any value. Scouts can just not focus on a computer and the match at the same time. A piece of paper is much better for counting because you can just make tick marks and you have tactile feedback of where you are making the marks. Ac computer is just not good for this task. The match predictor wasn't as useful as I hoped for, it turned out to be pretty inaccurate, but it gave us some laughs (we predicted a loss of 67 to 7 against us :P). It is not easy to set up if you've never done anything like this before. We can tell ourselves "Oh, you just need to install this, and that, etc" but the truth is, for people who don't breathe this stuff, it is hard. This is a problem with any application that requires some sort of web server, though. I have a few ideas on how to solve this, but it won't be solved next year for version 2007 unless I have some sort of breakthrough and devour some information.

So what's my overall opinion?
I may be biased (I am a developer, after all), but I love it. It actually performed a bit worse than I had hoped, but that just means it will get better. The absolute truth is that it helped us pick the best teams for our alliance, which enabled us to win the regional. By the way, thank you team 79, Captain Krunch (http://www.krunch79.com/) and team 1649 (sorry, I couldn't find your website or your team name!).


Ah, so many good points, I agree with all of them!

Many will be re-worked for v1.3 .

Match predictor was odd, thats all I have to say for that :).

Gary Dillard
12-03-2006, 11:16
And the drawbacks:
....The match predictor wasn't as useful as I hoped for, it turned out to be pretty inaccurate, but it gave us some laughs (we predicted a loss of 67 to 7 against us :P).

Actually Steve, the match predictor was (unfortunately) quite accurate. We were 6-0 after the first day, that night you predicted we would lose the first match by a small margin (which we did), win the second match by a small margin (which we did), and lose the last match by a huge margin (which we did; it was 82-8, not 67-7). We had a good laugh about your predictions but the joke was on us. I'm still curious how your "luck algorithm" works. :D

This program was invaluable to us for selecting our alliance and coming up with strategies that allowed us to win the regional. We did not have the best robot there, but we were able to develop an alliance and strategies for every match with the detailed information, especially the percentages of shots made, times on the ramp, and autonomous mode success. Thanks for all your hard work on this program - I strongly encourage other teams to use it.

Billfred
12-03-2006, 11:31
I haven't had a chance to put STAMP to the test (haven't been to a regional where I've had a dog in the fight yet), but if what Gary's saying is true (and I have no reason to doubt him), I'll be very interested in giving it a spin. Is SPAM bringing STAMP to Palmetto?

(By the way, I saw the sign while scanning the mass of banners around the balcony at UCF. Plenty readable for me, although the SSID could stand to be more visible. Good job for short notice.)

Warren Boudreau
13-03-2006, 12:26
I haven't had a chance to put STAMP to the test (haven't been to a regional where I've had a dog in the fight yet), but if what Gary's saying is true (and I have no reason to doubt him), I'll be very interested in giving it a spin. Is SPAM bringing STAMP to Palmetto?

(By the way, I saw the sign while scanning the mass of banners around the balcony at UCF. Plenty readable for me, although the SSID could stand to be more visible. Good job for short notice.)

When I last talked to Steve on Saturday, he wasn't planning on going to Palmetto. Classes get in the way.

Hopefully, y'all can find someone to carry the STAMP banner up to Palmetto.

It's only one datapoint, but so far it looks very promising.

As for the match predictor, it is hard to predict luck and strategy variations on such a small population of data.

Ethulin
13-03-2006, 18:01
When I last talked to Steve on Saturday, he wasn't planning on going to Palmetto. Classes get in the way.

Hopefully, y'all can find someone to carry the STAMP banner up to Palmetto.

It's only one datapoint, but so far it looks very promising.

As for the match predictor, it is hard to predict luck and strategy variations on such a small population of data.

I am perfecting the algorithim but it will never be close to perfect. I am hoping to make it output a spread for the data, but I doubt if it is within 10pts it can be relied on at all.