Statistics for week 3

[Bongle]

This week (week 2) (week 1)
Average red score: 29.44 () (23.8)
Average blue score: 29.53 ()(25.4)
Average score: 32.05 (29.493) (24.6)
Average score when neither team scored zero: 33.2 (30.41) ()
NUMBER of matches where at least one team scored zero: 26 (30) (35)
# of red wins: 184 (257)
# of blue wins: 183 (254)
# of ties: 3 (8)
Median score: 28 (27)

Average win margin: 23.8 (24.34) (18.2)
Average win margin when both teams were nonzero: 23.1 (24.06)
Average winning score: 44.1 (42.0) (33.7)
Average losing score: 20.1 (17.2) (15.4)

Effect of multiple regional attendance
Definition: Robot-regional: A robot-regional is a measurement of how many robots on an alliance have attended previous regionals. If an alliance has three robots on it that have attended 2 previous regionals each, then the alliance is a 6 robot-regional (RR) alliance and is probably the very definition of unstoppable. I measured the win rate based on the robot-regional difference between the two competing alliances. If you had a 4 RR alliance competing against a 1 RR alliance, then the win would be recorded as a 3RR win (since the difference between the RR count of each alliance is 3). Luckily, there was only one game with a 3RR difference this week.

Win rate for 0RR difference: 50%
Win rate for 1RR difference: 72%
Win rate for 2RR difference: 70%
Win rate for 3RR difference: 100% (only 1 game, so just ignore this)
More to come...

[Bongle]

Stats in Aggregate
These include ALL the matches played in all the regionals so far. The previous post was just the qualifying matches for the week that just finished.

Definition: Average alliance number: If you have an alliance with robot numbers b1,b2, and b3, then the average alliance number if (b1+b2+b3)/3. That is, it is the average of the 3 robot numbers that make it up.

teamNumVersusScoreAggr.png - the average alliance number versus the score they achieved. As you can see, this correlation is holding true even with 1200 matches. Teams with lower numbers tend to score more, because they tend to be more experienced.

differentialAggr.png - the win rate versus the difference in two competing alliances' average team numbers. If you have an alliance with (40,50,60) playing against an alliance with (1400,1500,1600), then the DIFFERENCE (as measured here) is 1450. So this looks at the win rate for the set of all matches with a given difference. If you see a match where the average alliance numbers of the competing teams differ by 1000, then you can look on this chart and see that there is a 35% chance the team with the higher average number will win. Hehe, they're almost doomed.

prevnextAggr.png - the score achieved in a team's (n+1)th match given their performance in their nth match. From this, we can tell that if a team scored 50 in one match, then their next match score has a 50% chance of being between about 18 and 45. See my week 2 thread for more details on how to read a boxplot, and more details on how to read this particular chart.

matchNumberAggr.png - the score achieved versus the match number that it was scored it. Notice that there really isn't that much of a trend upwards, which kinda surprised me. You'd think that there would be a strong tendency upwards as teams figure things out or upgrade their robot and tactics.

[thoughtful]

WOW,

Amazing, I find the stats very interesting. They goto show that experience does matter; however, determined teams can always break the mold.

I hope that you keep posting these stats as weeks go on.

Thankyou!

[DCA Fan]

Interesting to see how the Match Number/Score has panned out. Perhaps we're beginning to see that robots have achieved the best that they can, and aside from the necessary repairs, have stopped major upgrades.

[Bongle]

Originally Posted by DCA Fan

Interesting to see how the Match Number/Score has panned out. Perhaps we're beginning to see that robots have achieved the best that they can, and aside from the necessary repairs, have stopped major upgrades.

That surprised me as well. Keep in mind though: 72% of teams only attend a single regional. So at any given regional, around 3/4 of the teams there WILL be upgrading and learning like mad, so you'd expect that to show in the graph. Also keep in mind there is a decent climb from the mid-20s to low 30s if you look at the averages. It's not very visible because the graph is kinda crushed vertically.

[sciencenerd]

Originally Posted by Bongle

That surprised me as well. Keep in mind though: 72% of teams only attend a single regional. So at any given regional, around 3/4 of the teams there WILL be upgrading and learning like mad, so you'd expect that to show in the graph.

Not necessarily. A mentor on our team crunched some numbers and figured out that of the teams signed up for the Sacramento regional, for example, 60% of them will have attended at least one previously. What I'm saying is that there is obviously a big discrepancy between regionals on the number of veterans attending.

[Rick TYler]

Originally Posted by sciencenerd

Not necessarily. A mentor on our team crunched some numbers and figured out that of the teams signed up for the Sacramento regional, for example, 60% of them will have attended at least one previously. What I'm saying is that there is obviously a big discrepancy between regionals on the number of veterans attending.

You're right, Kyle. Some regionals attract a lot more veterans. PNW attracts a lot of new teams (it's also a newer regional). The Davis/Sacramento is really close to the weighted average of 1016. (By the way, Kyle is an amazingly talented freshmen who helped me install bumpers on Maverick at PNW. His team roolz.)

You can see more discussion on a related topic in this thread: http://www.chiefdelphi.com/forums/sh...threadid=45629

I'm sorry I won't see IRS at Davis. Good luck! By amazingly good scheduling, I will be visiting the SoCal Regional next Saturday, and I might have an excuse to be in Atlanta on business the weekend of the Championships...