Discrepancy at Chesapeake, Israel, Waterloo?

[Killraine]

I looked over the data from all of the regionals of the first 4 weeks after getting back from Chesapeake this weekend. At Chesapeake, there seemed to be an occasional problem where the middle blue alliance station would not establish communication with the robot. I was not surprised to see the data reflected this, with the Red alliance winning 61.5% of the qualification matches. I was surprised to see similar results at both the Israel and Waterloo regionals. Did you guys have a similar problem at your regionals?

Here is the data:

Regional                     Blue Win % Red Win % Difference 
Israel Regional	                37.97%	62.03%	12.03%
Chesapeake Regional	        38.46%	61.54%	11.54%
Waterloo Regional     	        61.36%	38.64%	11.36%
Pittsburgh Regional	        58.06%	41.94%	8.06%
Washington DC Regional	        42.11%	57.89%	7.89%
Finger Lakes Regional	        57.58%	42.42%	7.58%
NASA VCU Regional	        57.14%	42.86%	7.14%
Greater Kansas City Regional   56.94%	43.06%	6.94%
MI - Traverse City	        56.58%	43.42%	6.58%
Peachtree Regional	        56.34%	43.66%	6.34%
MI - Cass Tech	                56.25%	43.75%	6.25%
Philadelphia Regional	        56.25%	43.75%	6.25%
Wisconsin Regional	        54.93%	45.07%	4.93%
Midwest Regional	        45.07%	54.93%	4.93%
Bayou Regional	                45.61%	54.39%	4.39%
Los Angeles Regional	        54.29%	45.71%	4.29%
Oregon Regional	                53.97%	46.03%	3.97%
Boilermaker Regional	        53.85%	46.15%	3.85%
Arizona Regional	        46.97%	53.03%	3.03%
St. Louis Regional	        46.97%	53.03%	3.03%
Silicon Valley Regional	        47.62%	52.38%	2.38%
BAE Systems Regional	        52.38%	47.62%	2.38%
Oklahoma City Regional	        47.76%	52.24%	2.24%
New York City Regional	        48.05%	51.95%	1.95%
New Jersey Regional	        48.61%	51.39%	1.39%
MI - Detroit	                48.75%	51.25%	1.25%
MI - Kettering University	51.25%	48.75%	1.25%
MI - Lansing	                51.25%	48.75%	1.25%
Dallas Regional	                49.25%	50.75%	0.75%
Buckeye Regional	        50.63%	49.37%	0.63%
Boston Regional	                50.00%	50.00%	0.00%
San Diego Regional	        50.00%	50.00%	0.00%
Florida Regional	        50.00%	50.00%	0.00%

The difference column is merely how far off the results are from a 50/50 split (Obtained by abs(50 - Blue Win %)). You can see it jumps from 8% at Pittsburgh to 11% at Waterloo. Everything up to Pittsburgh seems to go up fairly linearly.

[Alex Cormier]

Very interesting information. But I am not really liking the "error" name for the difference. How about use the word "difference"? sounds better.

[Killraine]

Agreed. Error makes it sound like Percent Error, which would be a whole different thing entirely. I edited the post.

[Tom Line]

Putting your %'s into minitab and doing a simple capability sixpack, the distribution has a P value of .59, which strongly suggests you have a normal distribution.

Looking at the tails of the normal curve, only a couple values fall just outside that range, and you have both a long term and short term capability of well above 2. There is a bimodal distribution, but this sample size of regionals is somewhat limited.

With this limited sample size, my answer currently would be "It's well within expected limits".

Yeah, I know, it doesn't answer your question really. But everything points to normal statistical variation in Team winning percentage.

[Jonathan Norris]

Waterloo only had 34 qualification matches... So I am not surprised it is a bit off.

[Alex Cormier]

Originally Posted by Jonathan Norris

Waterloo only had 34 qualification matches... So I am not surprised it is a bit off.

44 Qualification matches.

[Killraine]

Originally Posted by Jonathan Norris

Waterloo only had 34 qualification matches... So I am not surprised it is a bit off.

Well, that addresses one of them. Does anyone have any details on Israel?


Also, @ the data from minitab: Are you sure its not looking at it on too large of a scale? The data is clearly pretty linear up until the 3% jump.

[Tom Line]

Any statistician will tell you that statistics never conclude anything - they are merely interpreted to mean something. You can, generally, argue any point you like if you know how to twist them well enough.

However, in this case, I think it's pretty fair to say you've simply found the tails of the curve. Someone's got to have the highest and lowest scores, after all.

[vivek16]

I think it's just coincidence, it's not like one alliance color has more at every single regional. The discrepancies happen.



-Vivek

[Nuttyman54]

Originally Posted by vivek16

I think it's just coincidence, it's not like one alliance color has more at every single regional. The discrepancies happen.



-Vivek

It's not always coincidence. In the case of Chesapeake, the middle blue station appeared to have chronic problems with dead robots. As we all know, a dead robot is pretty tough to overcome, and is possibly responsible for the discrepancy we've seen at Chesapeake. I don't think it's possible at this point, but I'd like to see the stats on how many matches had a disabled robot at the Blue 2 station, and how many of those matches blue lost.

[Seat Ninja]

My team was at Chesapeake. We were mostly on the blue alliance. As you can see below.

[rcflyer620]

I was an FTAA at Chesapeake and can tell you the most of the problems with "dead" robots were operator error. The most common was robots going on the field with the cRio un-plugged from the radio. The FTA had made it clear that once the gates were up you could not correct things like that on the field. We even announced it in the pits to no avail.
Unfortunately I also mentor team 2199 who lost a very good chance at being regional champion due to the robot in Blue-2 de-linking after autonomous. The weird thing is that it exhibited the same behavior in the previous match but in Blue 3 lending credence to the argument that it was a robot problem.
One thing I found that was fairly consistent was the flat black ethernet cable provided in the kit was problematic. teams that had swapped it out for real twisted pair ahd almost no problems with communication.

[JohnBoucher]

Originally Posted by rcflyer620

The most common was robots going on the field with the cRio un-plugged from the radio. The FTA had made it clear that once the gates were up you could not correct things like that on the field. We even announced it in the pits to no avail.

Really? How many robots were not able to compete because of this?

[Kingofl337]

Originally Posted by rcflyer620

I was an FTAA at Chesapeake and can tell you the most of the problems with "dead" robots were operator error. The most common was robots going on the field with the cRio un-plugged from the radio. The FTA had made it clear that once the gates were up you could not correct things like that on the field. We even announced it in the pits to no avail.
Unfortunately I also mentor team 2199 who lost a very good chance at being regional champion due to the robot in Blue-2 de-linking after autonomous. The weird thing is that it exhibited the same behavior in the previous match but in Blue 3 lending credence to the argument that it was a robot problem.
One thing I found that was fairly consistent was the flat black ethernet cable provided in the kit was problematic. teams that had swapped it out for real twisted pair ahd almost no problems with communication.

Yes, it was funny how blue had consistent problems at Chesapeake. So, bad in fact that before the finals two members from team 190 at different times said "I hope you don't end up on the blue alliance". Team 134 ran consistently though qualifications and eliminations never changing their code or robot configuration. We were forced to change to the blue side of the field. On our first finals match, two robot crashed and disabled each other. They disconnected from the field so the FTA claimed it was a power issue. Then on the next match 134 disabled in the middle of the match for no apparent reason and was still connected to the field. All attempts to convince the FTA that field my have been the issue were ignored. In the third match with no changes other than rebooting the robot and swapping driver station loactions with us team 134 ran fine.

Really? How many robots were not able to compete because of this?

I know of at least 6 times where the FTA took the time to point out the mistake to the team(s) and told the scoring table to disable the robot.

[dk5sm5luigi]

Originally Posted by JohnBoucher

Really? How many robots were not able to compete because of this?

I would have to say that I saw that happen at least 7 times. The FTA would bring the student out onto the field point out the issue and then disable them before the match started.

On another note do these stats include eliminations? If so then the data won't be accurate because in eliminations the higher seed is in red and in theory more likely to win.