488 XBot PNW Statistical analysis, and results from friday matches

[Fred Sayre]

Team 488's hard working scouts collected all of the match data from Friday which we used for strategy and alliance selection on Saturday. We decided full scouting on Saturday wasn't worth the effort, so unfortunately the data doesn't perfectly reflect a team's overall performance, but is interesting and a pretty accurate method of ranking.

It would have been difficult to evaluate exactly how many points a ringer robot would score in a match because the points changed so quickly with alliance members forming rows etc, so we only recorded number of ringers scored. To turn these numbers into a comparison of actual scoring potential we used two different statistics, average base, and average max score. The base score consisted of lifting points, plus two points for eachringer. This would bias the overall score in favor of lifters, so we also looked at average max score which assumed that all of the ringers the robot placed were in a row. This is an interesting statistic as it can favor ringers more, but also not really describe the ability of alliances to jointly create rows. Overall with defense and alliance organization we felt that the average max score was a pretty balanced indicator of score.

That being said, our data showed some interesting things.

Here is the list of teams and their corresponding max score per Friday data. Out of the top spots 2, 3, 6, and 8 were all held by teams that primarily lifted.

Rank Team AvgMaxScore
1 1540 37.5 (seed 2)
2 997 37 (seed 4)
3 488 28.5 (seed 6)
4 272 18.33333333
5 1571 18
6 1891 16.8
7 948 13 (seed 1)
8 1280 13
9 2149 11
10 1318 9.4
11 1595 8.2
12 957 7.2
13 1778 7
14 1823 7
15 1425 6.333333333
16 1510 6.333333333
17 2142 6
18 1087 4.8
19 956 4.4
20 360 4.333333333
21 2249 4
22 2222 3
23 2148 2.666666667
24 2046 2.6 (seed 8)

Notice seeds 3, 5, and 7 are not in the top 24 for points scored, something I heavily attribute to (my personal opinion) horrible matching algorithm.

If you sum the max scores for the 8 alliances, you get a pretty good predictor of the finals performances, with the statistical upset of 1294, 1425, and 2122 winning against 360, 2046, and 2149. They even went on to beat the regional winners 997, 272, and 1087 in a very close 62-61 point game before they came back to advance to the finals.

Rank SumMaxScore Aliance-
1 60.13 997 272 1087
2 57.5 1778 1540 948
3 46.1 1318 1595 488
4 25.53 1013 1510 1571
5 18 1359 1983 1891
6 17.93 360 2046 2149
7 15.7 1887 114 1280
8 8.73 1294 1425 2122

Top teams by average ringers placed per match:

Rank Team AvgRingersPerMatch
1 1540 3.83
2 272 3.33
3 1571 2.8
4 948 2.6
5 1595 2.6
6 1318 2.2
7 360 2
8 1425 2
9 1510 1.66
10 956 1.4

Top teams by average points from lifting:

Rank Team AvgLiftPoints
1 997 36
2 488 27.5
3 1891 15
4 1280 12
5 2149 10
6 1778 6
7 1823 6
8 957 6
9 2142 5
10 948 3


Overall I think lifting, defense, and agility became very important aspects of the game. In our final matches some of the best ringers at PNW were playing primarily defense, and the ability robots have to quickly approach and climb ramps was key to those lifting points. Many matches with very good ringer robots ended up with one or two ringers on the rack and a brutal fight for lifting points at the end.

I came to Portland with some expectations about the game, and was overall pleasantly surprised. I was expecting low ringer scores and lifting to dominate, but the difficulty of driving and lifting provided for some high scores through a great variety of strategies, but also low scoring and incredibly close and suspenseful games played by some of the best teams at the event. The combination of lifting, forming long rows, and placing spoilers make the ability for 10 seconds at the end of the game to change everything, and forced the drivers and alliances to work together to make some very quick and important decisions.

If anyone has questions, or is looking for any specific data let me know. We will be attending San Jose and will hopefully be collecting the same scouting data.

[pakratt1991]

Thank you for this wonderful information...

This looks like a great system supported by a really dedicated scouting team..
we have something similar but it reads out different data. hopefully we worked the bugs out at PNW and can make an excellent scouting report from U.C. Davis...

Although our scouting spreadsheet does have a match predictor and predicted the winners of every finals match up... it also predicts score and last year was much easier to predict final score then this year...

do you guys use a spreadsheet for this or some other type of data collection program?

[Fred Sayre]

Originally Posted by pakratt1991

Thank you for this wonderful information...

This looks like a great system supported by a really dedicated scouting team..
we have something similar but it reads out different data. hopefully we worked the bugs out at PNW and can make an excellent scouting report from U.C. Davis...

Although our scouting spreadsheet does have a match predictor and predicted the winners of every finals match up... it also predicts score and last year was much easier to predict final score then this year...

do you guys use a spreadsheet for this or some other type of data collection program?

We use access for the database side of things. We created a form that mimics paper scouting sheets that we have for 6 scouts to write down the data, and pass it to someone doing data entry. It is pretty easy to write a query to give you match by match data, and ranking averages for each team.

Congratulations on your PNW win. I was hoping we would get to play you guys since our designs were so comparable. Depending on how we do in San Jose, maybe we will see you in Atlanta

[Guy Davidson]

We'll be running a slightly different scouting system at SVR, but I'd love to compare data in some point. I wanna see how practical our data is compared to yours, as I might discover we're doing too much work.

-Guy Davidson

[Greg Marra]

How about the correlation between "lifting points" and seed or "ringing points" and seed? I would be interested to see what wins more qualifiers, as it seems like most teams were putting up very few rings early on in Week 1.

[Guy Davidson]

Since I already ran some FIRST related stats this week, I'm willing to run some more. I'll do it, given the data.

-Guy

[ChrisXPPro]

i have to say, these stats dont seem every correct, we had mininum of 2 rings each round, when not being tackeled by other teams, we had a max of 7 1 round, but we usely got 4 up! so i have to say that these arent every correct

[FunkyRatDemon]

I like how we were left off this list...
You guys did great at PNW

[Jimbo5051]

This is for the St Louis Regional isn't it?

[FunkyRatDemon]

Originally Posted by Jimbo5051

This is for the St Louis Regional isn't it?

Pacific North West

[Nate Edwards]

Thanks for this data, I can show my team that they did do a good job... Showing my team that they were the 12th highest scoring team on average will boost their confidence... Finishing in 50th place is just not where we should have ended up. We faced the eventual Regional Champs 8 times (12 including practice) and won twice (more including practice) We also had lots of radio issues. Matt from IFI was a big help. He should be thanked for all his hard work

[ericand]

Which average are you using (mean, median, mode) for ringers/match?
1425 had 2 matches where the gripper didn't work so for those it scored 0, Thus the average ranking failed to predict performance in the championship (were we had things working).

I would think looking at the numbers using a median number of ringers scored might be interesting.

[Fred Sayre]

Originally Posted by FunkyRatDemon

I like how we were left off this list...
You guys did great at PNW

You guys did some great defense. You were definately high on our list of teams to pick, as we knew how heavy the defensive game was going to get in the finals. (and we were positive that drivetrain would have zero problems with our lifts )

Originally Posted by ChrisXPPro

i have to say, these stats dont seem every correct, we had mininum of 2 rings each round, when not being tackeled by other teams, we had a max of 7 1 round, but we usely got 4 up! so i have to say that these arent every correct

Originally Posted by ericand

Which average are you using (mean, median, mode) for ringers/match?
1425 had 2 matches where the gripper didn't work so for those it scored 0, Thus the average ranking failed to predict performance in the championship (were we had things working).

I would think looking at the numbers using a median number of ringers scored might be interesting.

I prefaced this by saying that the data was from friday only. I know teams did better on saturday, but so did most people.

Our data showed for 1425:
Match Ringers
3 2
12 3
20 0
29 3
37 0
46 4

Which, averages (mean) to 2 ringers per match. Sure, the data is not perfect (just friday), and maybe it was less of an accurate indicator for 1425 then other teams (you guys did outperform the alliance max score predictions versus 360, 2046, and 2149).

[FunkyRatDemon]

Thanks, you guys did a great job and we appreciated the competition
ya, we're basically unmovable w/ our tracks, too bad our grabber didn't work properly

[jeffyman1011]

Thanks a bunch for the ranking system, i am from team 1891 and yes our team thought that we did well was a good feeling. But having data proof from another team is pretty invigorating.

Toodles, and can't wait for next year. Oh and who has started working on designs for next years robot: