Predictions Week 3: Creating Flow

[Looking Forward]

If variety is the spice of life, Ultimate Ascent is one spicy dish. The 2013 FRC challenge has produced far more strategic and design variety than any game since 2004. While slight differences in features led to different strategic and tactical approaches to many previous games, there's been a great deal of homogenization in FRC in recent years. Generally speaking, there haven't been more than two or three "classes" of common robots since the start of the 3v3 era. 2013, however, already has at least four distinguishable classes, with a multitude of sub-sets of each; pyramid shooters (with and without ground loading or the ability to move underneath the pyramid), full court shooters (tall and short), defenders (shot blockers, lane cloggers, and under pyramid), and climbers (20, 30, and dumpers).

In large part, it's because "do everything" bots simply aren't plausible for the vast majority of the FRC population this season, even counting the FRC royalty. From 2008-2012, building a robot capable of all the game functions was well within the reach of a very large portion of veteran teams. And where sacrifice was required, it was relatively minor. For instance, a team might forgo travelling under the tunnel in 2010 in favor of a hanging mechanism. 2013 is a beast of a different nature, with wildly differing tasks that require significant effort to accomplish in a satisfactory manner. Even the top tier teams have aspects they completely neglected to pursue, and mechanisms that aren't as consistent as we've come to expect. How many top tier teams had 30-pt climbing mechanisms that never saw action at their first event?

This increase in the variety of robot is leading to a plethora of different alliance structures and strategies. While teams typically stick within the style of play that develops at their event, that style varies dramatically depending on the quantities and effectiveness of particular types of robots. Northern Lights was all about full court shooters, as a number of top flight teams were draining shots from the opposite corner. If you were to bring that up with a competitor at most other events this weekend (save perhaps Waterford), it would differ tremendously from their experience. Some alliances are all about run-and-gun, while others try to slow the pace to emphasize autonomous and the end game. It will be interesting to see how much these strategies converge or diverge as the season progresses, and how each field at Championship takes its own personality depending on its demographics. Ultimate Ascent is shaping up to be one of the most dynamic and interesting games in recent memory.


Bullet Points:

• A popular opinion around Chief Delphi is that this season is great for upsets. The statistical accuracy of that sentiment is contentious, though. It depends very much on how you define your criteria. In 2012, the #1 seed won 43 regionals/districts (64.2%). So far in 2013, the #1 seed has won 11 regionals/districts (57.9%). The mean winning alliance seed in 2012 was 2.014, this year it's 2.737. There's movement towards the underdogs, but it's not huge. If you evaluate it based on which seed wins each elimination series, the difference is even less. In 2012, the higher seeded alliance won 70.0% of the time. This season, the higher seed is winning 67.4% of elimination series. Only time will tell if the variations here are just noise, or if Ultimate Ascent does (slightly) favor the underdog compared to Rebound Rumble.
  note: All this data was compiled before the results from Israel were in.
• An interesting fact emerged as that data was being compiled. Life's not easy as a member of the #2 alliance. Not only has the #2 only once this season (Israel), but when #2 meets #3 in the semi-finals, #3 actually holds in the advantage over the past two seasons. #2 and #3 have split eight meetings this year, but #3 won 23 of 39 meetings last year.
• Using the #1 seed winning percentage alone is a flawed way to evaluate the parity of elimination matches. 2005 was the sole year of 3v3 competition that didn't feature the serpentine draft, and thus was obviously very difficult for underdog alliances to leave with a victory. Yet, only 50% of regionals ended with the #1 alliance taking home banners. Why? The #2 alliance took home 12 of the 30 regional wins, and the 5th-8th alliances went completely empty handed.
• The Detroit Red Wings popularized a forechecking scheme known as the "left wing lock" in the NHL. While nothing like what's being employed in FRC, the term fits a popular strategy. Most notably used by the high powered winning alliances in Toronto and San Diego, the strategy employs a defensive bot clogging up the left lane next to the downfield pyramid, thus making it more difficult for the other alliance to reach their protected loading zone. Beyond that, the defender is in great position to block full court shots, and allows them to be 84" tall. While certainly an important function of the alliance as a whole, it's not terribly demanding of the third alliance member and works well at events where the field thins heavily by the last few alliance selections.
• Speaking of variety, I guess I should be thankful that 1503 has one of the most original and fiendishly simple shooters out of any team this season. Without any vision processing or video feedback, the pilots are really going to have to trust their instincts with their all-or-nothing approach to scoring. They should be among the best scoring machines at Pittsburgh, but don't get too cocky, Sparfox. If 620, 888, or 2614 can hit their stride, they could seriously challenge the visitors from Canada. One can hope 1503 aligns with their perfect wing mate, 3193, FalcoTech.
• After a disappointing quarter-final exit in Palmetto, 2363 is looking to rebound in Virginia. Triple Helix will be among the top competitors in a field that is thinned out by being on the same weekend as North Carolina. 1610 and 1403 will also be in the mix, but the visitors from Connecticut (195) have the potential to be the biggest threat or best ally for Triple Helix. Especially if the Cyber Knights are allowed to find their groove from full court.
• Even some of the best teams are having issues reaching the top of the pyramid quickly and consistently. The Simbotics had multiple matches where they were not well enough aligned on the corner post to advance all the way to the top. It'd be interesting to see the average points scored per attempt among teams who have successfully completed a 30-pt ascent.
• Given the challenges of climbing, it's especially impressive when a rookie can not only reach 30, but dump in the pyramid goal as well. 4451 might be the best rookie sensation so far this year, after being the #2 selection at two events (Palmetto and Orlando), and reaching the finals at the latter. Another South Carolina team, 343, will attempt to replicate their success at Peachtree this week, but will need to show improvement from Palmetto (where they frequently struggled to make it beyond the first rung). Halfway across the country in Detroit, 1023's multi-purpose hopper will provide them the ability to score in the pyramid goal while climbing, but they also bring a shooter to the table.
• Bad scouting and wonky alliance selections are not something new in 2013. Just ask any team that does scout well. But one of the worst examples of letting a great team fall too far in the draft was 910 at Kettering. Foley Freeze will likely be a much earlier selection in Detroit this week, but will have plenty of competition. 469 and 1023 will both be competing for the first time this year at the district, and will be among the best teams there.
• Kansas City may well come down to who can raise above 10. Practically every team has some sort of method to hang from the 30" bar, but a handful of the top competitors at GKC are aiming for more without giving up shooting. 1986 is the definite favorite (backed by their ground pickup and autonomous shooting), but weren't climbing beyond the first rung when they won Hub City. 1625 and 1730 both have mechanisms to escalate the pyramid, but it remains to be seen how quick and effective they are on the real field. 935 actually uses two different devices on their ascent to 20 points, a "passive" hook and then a pair of motorized tape measures with hooks attached to the end.
• Autonomous continued to play a massive role in deciding matches during week two. Four of the top contenders at the more competitive of the two MAR districts this week (TCNJ) will have the ability to load additional discs from the ground during autonomous. 2590 made the most of this on way to victory in Hatboro-Horsham. 25 and 103 were inconsistent, both in autonomous and tele-op, but were incredibly dangerous when they were on their game, during week one. 2016 is taking their sleek looking machine out of the bag for the first time.
• It's not often that teams "change numbers." Yet, Team Pwnage did just that, as they changed from their previous number (2949) to 2451 this season after splitting from their previous sponsor organization. Hopefully they can continue to carry 2949's reputation as a budding star when they compete at GKC this weekend, despite the new number.
• There's little doubt that 2539's bot would be able to compete at the MAR championship, but after a second round selection and quarter-final exit at a very deep Hatboro-Horsham event, they're behind the 8-ball. The Krypton Cougars are going to have to step up their game and put up some serious points in the MAR standings at Chestnut Hill this weekend to reach Bethlehem.
• There are other teams at Chestnut Hill with work left to do after Hatboro-Horsham. 341 took home gold during week one, but after significant issues with their camera and shooter jams, there are still plenty of questions about Daisy. 1218 was the quickest 30-pt climber at the event, but their shooter left a lot to be desired and their suction cup pickup went un-utilized. If Daisy and Vulcan don't step up, the greater the chances for 3974, 486, and 1640 to win go up.
• Perhaps no event this weekend lacks a clear favorite heading in as much as St. Louis. There are a number of teams that should be able to score well, but it would take one heck of a driving performance for any of them to dominate. Watch for 1208, 1094, and 1985 to be in the thick of the action.
• Only seven teams in Montreal have a number under 3000. The event is tremendously young, but only sports four rookies. Should be interesting to see how much these teams were able to pick up in their first couple years in FRC. 3387 and 3985 have relatively simple, but effective, shooters that will likely be among the better scoring machines at the event.
• North Carolina is 1519's home away from home, as Mechanical Mayhem is venturing down to the regional for the third time in the past four seasons. Reaching the finals in each of their previous two appearances may have something to do with them enjoying the event, and if their full court shooter can hit the mark, it's within reach again.
• Some of the historically strong teams at the Virginia regional are taking their talents to North Carolina this year. 122, 435, and 1086 will all be players in deciding who takes home gold from Raleigh. Blue Cheese will have the best odds of the bunch, but it will be far from a sure thing.
• Despite the combo of two of the more successful teams in Texas over the past couple years on the Lonestar #2 alliance, it's hard to call the stacked 1-2-3 punch #7 alliance (231, 192, 1429) beating them an "upset." Never the less, a quarter-final exit always leaves a sour taste in the mouth of a team the caliber of 624. CRyptonite has higher aims this weekend at Peachtree, and their robot has the potential to match. But they'll need to execute consistently in order to best the likes of 2415, 1648, and 1311.
• When your standards of success are as high as those of RUSH and the Thunder Chickens, just reaching the finals at a single event is a disappointing season. Both 27 and 217 were hoping that 2013 would be better than 2012, but neither managed to get the results they wanted at their first event of the season. We haven't seen the best from either team yet, and there's a chance that either (or both) could be in the winner's circle at St. Joseph's if they can work out the kinks in their machines. They're going to have to catch-up to 2474, as Excel was terrific at Traverse City (#3 seed, 1st pick, and champion).
• If the competitors are ready right out of the bag, there should be no shortage of firepower at Boilermaker this weekend. Speedy and streamlined human loading shooters like 359 and 868 look to be among the best in the field, but don't count out 234, 829, 1646, and 1732. But it's THRUST that may be the favorite. 1501's 2013 machine has the potential to be the best we've seen from them, especially if they can corral additional discs during autonomous.

[Yipyapper]

A popular opinion around Chief Delphi is that this season is great for upsets. The statistical accuracy of that sentiment is contentious, though. It depends very much on how you define your criteria. In 2012, the #1 seed won 43 regionals/districts (64.2%). So far in 2013, the #1 seed has won 11 regionals/districts (57.9%). The mean winning alliance seed in 2012 was 2.014, this year it's 2.737. There's movement towards the underdogs, but it's not huge. If you evaluate it based on which seed wins each elimination series, the difference is even less. In 2012, the higher seeded alliance won 70.0% of the time. This season, the higher seed is winning 67.4% of elimination series. Only time will tell if the variations here are just noise, or if Ultimate Ascent does (slightly) favor the underdog compared to Rebound Rumble.

First of all, I love reading these week after week.

The only thing I wanted to point out was the above blurb, since the results only go down by about 0.6 points seedwise for regional winners, but the big thing that was noted on CD concerning upsets was that the second week had a good number of sizeable upsets. The results you have include the 1st week, which fared much better for the higher seeds and skewed the overall data to favour the higher seeds.

Without the 1st week, the mean winners had a seed value of 3.364, which is more than a full seed below the 2.014 in 2012. This also doesn't show the details with the low seeds reaching the finals several times, which would show that upsetting teams in general (not just winning upsets) seems to be at an all-time or near-all-time high.

[Anupam Goli]

Despite the combo of two of the more successful teams in Texas over the past couple years on the Lonestar #2 alliance, it's hard to call the stacked 1-2-3 punch #7 alliance (231, 192, 1429) beating them an "upset." Never the less, a quarter-final exit always leaves a sour taste in the mouth of a team the caliber of 624. CRyptonite has higher aims this weekend at Peachtree, and their robot has the potential to match. But they'll need to execute consistently in order to best the likes of 2415, 1648, and 1311.

Awesome, got a shout out in this week's predictions. Hopefully we can contend with the others mentioned. I'd watch for 4509, 2974, and 1746 at Peachtree as well. 2974 was solid when they were practicing. I don't know much about 4509, but they seem to have some connections with 1771.

[Akash Rastogi]

I'd watch for 3928's debut at GKC this weekend. That group of mentors + what sound like some motivated students should be one hell of a combo.

http://www.teamneutrino.org/2013/pre...ease/imag2567/

[Chris is me]

The Detroit Red Wings popularized a forechecking scheme known as the "left wing lock" in the NHL. While nothing like what's being employed in FRC, the term fits a popular strategy. Most notably used by the high powered winning alliances in Toronto and San Diego, the strategy employs a defensive bot clogging up the left lane next to the downfield pyramid, thus making it more difficult for the other alliance to reach their protected loading zone. Beyond that, the defender is in great position to block full court shots, and allows them to be 84" tall. While certainly an important function of the alliance as a whole, it's not terribly demanding of the third alliance member and works well at events where the field thins heavily by the last few alliance selections.

We used this strategy with at least one alliance partner for almost every match this season. It is by far the easiest and safest way to pay penalty-free defense in this game. A very aggressive line + man-to-man approach is a bit better, but it takes a lot more skill to execute.

Robots that are under 30" tall have an advantage for this basic strategy because they can "chase" teams trying to go around the blocker through the pyramid.

[nuggetsyl]

Originally Posted by Looking Forward

25 and 103 were inconsistent, both in autonomous and tele-op, but were incredibly dangerous when they were on their game, during week one.[/list]

I agree 100%. I can also say this, look for us (103,25) to be much more consistent for week 3. We have several issues we needed to work out like the tower being lower then any other regional and bag motors failing the carpet coming up and breaking our intake. We made ajustments in our 6 hour window and look foward to this week.

[artdutra04]

Originally Posted by Yipyapper

First of all, I love reading these week after week.

The only thing I wanted to point out was the above blurb, since the results only go down by about 0.6 points seedwise for regional winners, but the big thing that was noted on CD concerning upsets was that the second week had a good number of sizeable upsets. The results you have include the 1st week, which fared much better for the higher seeds and skewed the overall data to favour the higher seeds.

Without the 1st week, the mean winners had a seed value of 3.364, which is more than a full seed below the 2.014 in 2012. This also doesn't show the details with the low seeds reaching the finals several times, which would show that upsetting teams in general (not just winning upsets) seems to be at an all-time or near-all-time high.

It's still too early to tell. With small data sizes (only 8 events in week 1 and 12 events in week 2), the data is bound to be noisy.

A better comparison would be to track previous games week by week to compare how noisy their data was.

[Yipyapper]

Originally Posted by artdutra04

It's still too early to tell. With small data sizes (only 8 events in week 1 and 12 events in week 2), the data is bound to be noisy.

A better comparison would be to track previous games week by week to compare how noisy their data was.

Oh, for sure; I was just noting that it was an interesting second week.

[Lil' Lavery]

Originally Posted by Yipyapper

Oh, for sure; I was just noting that it was an interesting second week.

Not outside the norm, though. #8 upset #1 3 times during week 3 last season.

[Alpha Beta]

Originally Posted by Looking Forward

Kansas City may well come down to who can raise above 10. Practically every team has some sort of method to hang from the 30" bar, but a handful of the top competitors at GKC are aiming for more without giving up shooting. 1986 is the definite favorite (backed by their ground pickup and autonomous shooting), but weren't climbing beyond the first rung when they won Hub City.

We anticipate giving the field reset crew more of a work-out with the belay hooks this weekend.

Enough good teams here to create some very unfavorable match-ups in qualifications. Can't wait to see how it all plays out.

Web-cast = http://www.more.net/first-robotics

[nuggetsyl]

In another thread (lessons learned from week one) Sean posted about defence and how it was easy to do. What I am shocked about is how bad teams are at playing defence this year. Maybe it is the increased use of Mech or omni wheels, I am not sure. I though you would have seen more defence this year becasue of the win loss system. I am tempted about making a video of how to play defence with this years game.

[BigJ]

Originally Posted by Looking Forward

don't get too cocky, Sparfox.

I see what you did there

[Koko Ed]

They didn't merit a mention in the preview but take my word for it: don't sleep on 3015 in Virginia. They are a prolific scoring machine and they will be heard from in the eliminations there.

[nikeairmancurry]

Originally Posted by Looking Forward

[*]When your standards of success are as high as those of RUSH and the Thunder Chickens, just reaching the finals at a single event is a disappointing season. Both 27 and 217 were hoping that 2013 would be better than 2012, but neither managed to get the results they wanted at their first event of the season. We haven't seen the best from either team yet, and there's a chance that either (or both) could be in the winner's circle at St. Joseph's if they can work out the kinks in their machines. They're going to have to catch-up to 2474, as Excel was terrific at Traverse City (#3 seed, 1st pick, and champion).

217 is not competing this weekend. TBA lists them, but FIRST does not. Plus that would be a fourth district they would be attending.

[mrmummert]

Its nice after so long a time to even be mentioned here. And yes at least
1610 is wary and knows what 3015 could do. I see some of the New York
teams and a few locals to do well at VCU this year.