FIRST in Michigan: 2015 Mid-Season Review

[FiM Informer]

After an exhausting 9 events, 860 matches, and 307 teams played, the Michigan district season is only half over. The game has changed a little bit from build season projections to now, which I’ll discuss in detail, topic by topic. For a bonus at the end, a midseason power ranking of Michigan teams by Michigan experts. Michigan. Let’s get started.

Importance of Auton
For the average mid-level team, auton hasn’t been at all a factor. For elites, it has been a serious differentiator from other elites without the same capabilities. Auton may not be as important as in years past, but 20 points is plenty worth shaking a stick at. A number of teams have pulled off these 20-point autons with various consistencies (4003, 67, 2337, 70, 5046, 107, obviously 2054 with their 28-pointer). While 2337 and 4003 look the most consistent so far, none look 100% yet.

A couple other interesting autons you may have missed: 1322 scores an 8-pointer by dragging all staging zone RCs into the auto zone. Reminds me a bit of 2145’s colored disc scoring in 2013: not all that useful, but cool nonetheless. Another interesting auton is 3452’s two GREY tote auto. 3452 is a double-stacker that gets a head start by grabbing a couple greys during the autonomous period.

Importance of Litter
Meanwhile, litter has been seriously low-hanging fruit (per Tom Nader). Any team has the capability to have a big-time human player. Two Michigan teams have taken this the farthest: 5167 and 2767. Each averaged about 20 more points of litter per match than the average teams at their events. That’s right, the elitest of human players were at the same level as an elite auton (what is this, 2009?). Litter is also a great form of defense, as it can cause all sorts of problem for robots in a game that necessitates precision. Just take care to get those noodles to the other side of the field, or you might be playing D on yourself.

So far, it’s unlikely teams have been heavily scouting or selecting alliance members based on litter, but not at least considering it would be a big mistake. Even the RC-scoringest of alliances can get 10-20 points of litter by throwing, so the litter effect is here to stay.

Importance of Coopertition
Coopertition isn’t quite at the level of importance it had in 2012, but it’s close. In 2012, it was commonplace for both midlevel and elite teams alike to have more coop points than actual qualification points. This year, only 5 teams so far have had more coop points than other points scored (4 at Standish and 1 more at Gull Lake). Only 2 of those teams made the playoffs, and only 1 advanced past the quarterfinals. My point: coop hasn’t overshadowed regular scoring all that much.

However, teams have been plenty successful with coop. Just ask 217, 1250, or 4381, who each had successful coop stacks in 9 of their 12 matches. These teams were able to integrate coop into successful non-coop qualification strategy.

The importance of coopertition success depends on how much it benefits the teams involved. As robots improve and scores increase, 40 points will decrease in value, even to the point that it’s no longer worth pursuing for elite teams. However, this isn’t the only factor. When robots improve, the risk involved decreases, so we’ll continue to see lots of coop stacks as lower level teams reach levels of risk low enough to make coop worth doing. At MSC, coop will often be left to a 3rd alliance partner to give them a task worth doing.

Independent stacker-cappers vs One-dimensional
Two models of match strategy have developed so far: one in which teams work independently to create their own capped stacks and a second in which teams work together, typically with one team only stacking totes and the other focusing more on capping. In playoff matches, it has been demonstrated that either model can be successful. Typically a jack-of-all-trades will be at a speed disadvantage, but in this case the coordination requiring for partners to work together on stacks counteracts any raw ability advantage.

Independent robots have had more success so far, and to some degree this is true simply because those independent robots have been miles ahead of their competition, but the full truth goes farther than that. As the robots improve, the game gets faster and coordination between partners gets harder. Further, the gap between “great” and “elite” teams in stacking or capping ability will narrow as teams approach skill ceilings. That’s why MSC will be won by an alliance featuring at least one robot working entirely independently. My guess is that the 2nd robot will collaborate with the 3rd bot to make best use of the 3rd bot’s abilities.

The Playoff game and Qualification games have rarely been so far apart and it goes beyond Coop. Success in the qualification rounds (at any event short of MSC) will absolutely require the ability to play independently. This is due to the low average level of play at Michigan districts. Teams that need their partners to help create capped stacks will have occasional poor matches that pull them far down the rankings. Although it could have gone much worse, one example match of qual alliance partners with the same specialties occurred at Woodhaven this past weekend when 27 and 3322 (combined auton RC grabbing capability of 6) paired up: www.youtube.com/watch?v=5-YzPhWj_WU

Ramps
Using a ramp to aid HP stacking is certainly not necessary, but the best landfill teams are both rarer and have higher potentials than HP loaders without ramps. With two elite bots working out of the same areas unlikely to want to pair up, the top HP bots will want to differentiate from each other. Ramps help to do that because they improve both typical performance (reducing tote chute failures) and top potential performance (pre-develop stacks to avoid being HP-speed-limited). The greatest benefit will be for teams like 548 who rely on the tote chute to provide upright totes, although many of those teams could get partial benefits by simply leaving a tote at the chute to serve as a make-shift ramp (like 2834, 1076, and many others).

Play your second event early
You look a lot better, and you have an advantage over 1st event teams. 1023 took special advantage of this by making huge improvements to their robot, while watching regularly elite teams like 3539 and 469 work out robot kinks. In many cases this isn’t a big factor in the playoffs, as the 1st event teams have time to get going by then, but it is a huge factor in qualifications. Teams that come out swinging develop big Qual Avg advantages they’re unlikely to give up. That’s why in week 3, all 3 top seeds and 5 of the 6 winning captains/1st picks were teams that had played before, despite the fact that only about 1/3 of teams playing in week 3 were at their second event.

Accuracy of rankings and Top Seeds 9-0
I’ve never seen event rankings quite as good as they have been this year. While rankings at districts have always been better than typical regionals, there are a few factors in this game that result in better rankings than usual. The first is obvious: averaging a team’s qualification scores is a better indicator of quality than simply comparing a team and its random partners with their random opponents 12 times. Another is the fact that a team’s QS result in a match is dependent only on their 2 partners and not on 2 partners and 3 opponents, eliminating quite a bit of noise. This has been a boon to OPR as well, which makes the “no impact by opponents” assumption within its matrix algebra.

The accuracy of the rankings has been one factor in the success of top seeds in the playoff rounds. When the best robot at a competition gets to have the ability to pick the next best team, it’s unlikely any other alliance will be able to beat them (duh!). Another factor contributing to that win streak: without serious defense in this game, alliances are much more able to put up consistent scores. They can go out with the same gameplan every match instead of needing to outsmart their opponents.

Expect to see top seeds continue to dominate, because, simply put, the 3rd robots on playoff alliances don’t provide all that much when the top two robots can be resource-limited all on their own.

Change in first three weeks, moving forward
While there were significant improvements from Week 1 to Week 2, Week 2 and Week 3 looked pretty similar statistically. Coop, RC, and Tote scores stayed relatively stable. Auton scores decreased, but those scores are highly reliant on which of the few auton-capable teams are playing that week. However, litter scores have consistently increased, from 6.3 points per team per match in Week 1 to 8.6 to 10.4.

I expect there to be bigger changes moving forward. In Week 3, most teams were still playing their first events. That’s not true starting this week, so teams’ robot iterations will begin showing up on the field.

Power Rankings
I conducted a poll of 8 FRC experts from across the state of Michigan. These include mentors from 4 of the current top 10 Michigan teams (in avg event pts per event), while the others represent Hall of Fame, former World Champion, and other generally well-regarded teams.

The voters were asked to rank their top 15 teams. A exponential decay point system was applied to those ballots (#1 vote worth 30 pts, #2 vote was worth 80% of #1, #3 80% of #2, etc.). Only 5 teams were on all 8 ballots, and only 3 were in the top 10 of each ballot. The following are the results of this power ranking.

Team (First place votes) - Points - Commentary

1. 1023 (5) - 222 - While they were good at Southfield, Bedford put their game on another level at Woodhaven. 1023 scored 23 capped 6-stacks in 12 qual matches, and even began adding a 3rd uncapped stack in playoffs. They were in the top 2 of every ballot.
2. 67 (3) - 205 - HOT built a great stacking machine, but their RC abilities need a lot of work. They finally got their 20-pt auton working on Saturday, and I expect them to make even bigger improvements at their second event. 67 was in the top 3 of every ballot.
3. 314 (0) - 99 - While Big MO were overrated at the end of last season (they were worthy of much better than the 14th pick at MSC, but not the copious accolades they got), there’s no such problem this year. 314 outscored 2337 at Kettering despite the latter’s auton capabilities. A little more speed and they could go for a third stack per match.
4. 33 (0) - 89 - The first landfill-focused robot on this list, the Bees have a good intake and an extraordinarily fast elevator (shades of 469 - 2011). Working on that intake’s inconsistencies gives them significant room for improvement.
5. 1025 (0) - 82 - One of the more unique robots in Michigan this year, the IMPIs are admittedly not a team I’d have expected to be in this position. They could put up as many totes as most of the teams on this list, though, and figure to be a more tote-focused stacker at MSC.
6. 2137 (0) - 73 - The literal tape measure auton would’ve seemed to be more interesting than useful, but the reduced time to get that 4th RC proved useful to TORC at Waterford. Their HP stacking is nothing to write home about compared to some of the teams above, but that just means they have room for improvement.
7. 2337 (0) - 69 - The NERDs came out of the gate with a working 3-tote auton that propelled their way to be Kettering champs for the 2nd straight year. They’ll need to put up more than the one capped stack to be a major factor at MSC, but their landfill capabilities are very promising
8. 3604 (0) - 65 - A very capable bot with one of the more unique tote lifts this year, the Goon Squad made few upgrades between Southfield and Woodhaven. Still, their ability to work independently was enough at Woodhaven, where they were the right partner for fellow loner 1023. The ramp certainly improved their HP scoring, and their HP/landfill versatility will be valuable too.
9. 2054 (0) - 49 - One of the more hyped teams from before week 1, the Tech Vikes didn’t disappoint. While it took them a while to get going, their auton looked pretty solid in the playoff matches. Like 2337, that auton is a nice bonus but they need to work on their stacking speed.
10. 3620 (0) - 37 - To some degree, the Averages Joes were fortunate to have the opportunity to select 2054, but they were quite capable in their own right. Without the ability to stack a new tote onto a current stack in possession, 3620 may have a lower tote-stacking ceiling than other teams, but they can make up for it with their capping skills.

• Honorable Mentions (>10 points): 245, 4381, 4967, 85, 1918

Magic 8-Ball 2015 Edition
Last year, the magic 8-ball and I had real problems trying to figure out who would win MSC. Well with all the changes that have happened, maybe this thing will do a bit better this time around.

*Gives it a shake*

“Lunacy was more of a fun game to watch than this.”

Well I guess it’s still broken.

[Loose Screw]

I find this year to be really interesting when it comes to auton and HP. This is the first year since 2009 when HP can score more points in auton. I wonder if auton will be the game-winning points at MSC finals?

I predict MSC to be crazy (as always) and in some finals matches, alliances will run out of totes. RC-snatching autons will be key for victory in finals, but teams that focus purely on capping will struggle with rankings (rely on partners to make 6 stacks).

I can't wait to see finals this year!

[hardcopi]

I know we had 2 HP, one is a better thrower and we switched him to be our full time HP after it became obvious how many points we could rack up. We averaged over 20 points a match with noodles and he even took out a robot briefly in quarters with a noodle.

[RonnieS]

Very informative post for many FiM teams. I sure hope we are able to speed things up this weekend and deliver the 3rd stack capped
-Ronnie

[stuart2054]

Very good analysis as always. We were as stated, struggling with our auton at Kentwood and the differences between practice modes and a real FMS calling the shots but we hammered it out.

We were also fortunate to have the "Joes" and "ID" as partners in the finals. The Average Joes have a great robot and team that was very consistent throughout the quals and finals. Look out for 85 "Bob" and 2959 the "Robotarians" especially and any of the top 8 from Kentwood as they all are very good and not to be underestimated. 2767's human player is unreal.

[Jack_O]

Keep an eye on the human player from The Big Red Theory team 5162. He got 33 litter points alone in one semi-final match at Gull Lake. That means he got eight of his nine throws onto the other side of the field; it really made a mess for the opposing alliance to drive around.