I’ve been downloading the match results from the RobotEvents API. In looking at the results of elimination rounds, one would typically expect the red team to win, as it is generally speaking the higher seed.
Something interesting in this season’s results may be emerging. Granted that COVID has dramatically reduced the number of elimination matches to analyze, I do think there may be something here.
ElimRound
R16 Seed
Red Wins
Blue Wins
Ties
Red WinRate
R16
1v16
4
0
0
100%
R16
2v15
4
0
0
100%
R16
3v14
4
0
0
100%
R16
4v13
4
0
0
100%
R16
5v12
5
1
0
83%
R16
6v11
5
2
0
71%
R16
7v10
7
3
0
70%
R16
8v9
6
8
0
43%
QF
1v8
22
0
0
100%
QF
2v7
23
2
0
92%
QF
3v6
26
6
0
81%
QF
4v5
15
20
2
43%
SF
1v4
34
4
1
89%
SF
2v3
31
8
0
79%
F
1v2
32
11
2
74%
Of particular note are the win rates for Red seeds 8 against Blue seed 9 and Red seed 4 against Blue seed 5. Both are significantly below 50%, whereas in prior years 8v9 and 4v5 would have a roughly 55% win rate. The win rates for all other seeding matchups are pretty close to what they have been in year’s past, except for these 2 matchups. Aside from not having many of these matches, it is possible this may be a fluke.
However, given that the rest of the pairings align with historical norms, I think that this may be attributed to a significant change in the way teams are ranked this year. This may be a result in the WP teams can get for performing the “home row” during autonomous, where teams in the middle of the “ranking pack” are boosted by having an extra Win Point that provides a bonus during Qualification rankings (e.g. even if they do not win the Auton bonus, they can still acheive a WP), but serves no real purpose during Elimination rounds (where WP don’t come into play).
I’m not saying anything needs to change, just that this could be an interesting knock-on effect of this years Auton WP rule.
I would argue that the 8v9 is still too small a sample to make any conclusions and one would expect those to be extremely close. And since the 4v5 has a much larger sample size, I don’t think those results are surprising, a little out of the norm, but not out of the realm of possibility.
The 4th seeded team and 5th seeded team are basically at the same level of talent, so it’s entirely reasonable for the win rate to hover at anywhere between 40% and 60%. The difference between the 4th and 5th seed is pretty much just whoever wins statistical tiebreakers, so I would not say it’s out of the norm per se.
While I agree and grant that the sample size is potentially still too small, the fact that the rest of the seeds (especially past QF) track close to historical (I’ve posted similar for prior years, for example: Tower Takeover Competition - Statistical Analysis - #24 by squirrelman777), I am inclined to think there’s something unusual about this year. Clearly, one could flip a penny 10 times and get all heads.
In both the 8v9 and 4v5, the historical win rate over the last 3 years is 54% for both (and very low standard deviations, we’re talking 54,53,54 over the last 3 years, not 45,60,55). Again, it’s what you’d expect - these are generally closely matched teams, with the lower seed generally slightly better. To see that win rate flip to 43% - more than a 10 point swing!- does seem to have an explanation other that statistical anomaly.
One potential explanation could be the new home row bonus during the autonomous period. This allows teams to pretty much be mismatched depending on their performance in the autonomous period, because you can have a 2-4-0 team placing highly in a tournament (assuming they get home row bonus in all 6 matches) while a 4-2-0 team will be behind them in the standings (assuming they never get the home row bonus). The home row bonus rewards teams for having a consistent autonomous routine to the point where the actual ability to win matches becomes less of a factor. Of course, that’s not to say that it’s a game flaw, however the ability to directly affect your WP without winning or tying is probably what led to this change. Teams who succeed during the autonomous period this year are not necessarily efficient at winning matches, and teams who win matches are not necessarily efficient at completing the home row during the autonomous period. The latter could explain why the 5th/9th seed have been relatively dominant throughout this season.
It does for the most part, but the main point of contention here is the statistics during the 8v9 and 4v5 matches. As you said, the red alliance ranks higher than the blue alliance, so it is interesting that the 8th seed and 4th seed (red alliance) are actually winning at a rate less than 50%, when in previous years the win rate for red alliance has settled around 54%.
You have to also account that it is during a pandemic not many teams can compete and from what I have seen many of the teams competing have much less skill increasing the skill gap between teams.
The 4th and 8th seed are the red alliance in this case, so if they are one seed ahead of the blue alliance, they should be winning at a rate above 50% because as you said, the skill gap is larger, but their win rates are at 43%, so the skill gap doesn’t explain why the lower seeded team is winning more often against the higher seeded team during a 4v5 or 8v9 situation, because from what you’re saying, the red alliance should be easily beating blue alliance, however the win rate for red alliance in 4th vs. 5th and 8th vs. 9th is 43%.
I don’t think the question is why are lower seeds beating higher seeds I believe the point of focus is why are the better teams in elims ranked the way they are. The point was made that during elims home row bonus does not matter, so I am wondering if home row bonus is artificially boosting certain teams in the rankings. These teams may have a good auto but fall short in actually winning matches with very similarly skilled opponents.