Week 1 Analysis: Part 2 (Eliminations)

This thread is a continuation of the thread Week 1 Analysis: Part 1”. I was going to add this on to the original post of that thread, but it appears there is a time limit for editing posts. If mod can/would like to merge the two threads and rename it “Week 1: In Depth Game Analysis” or something similar that would be great! If not, sorry for the second thread. Next time (if there is a next time) I’ll just make 1 thread at the end of the whole week.

Week 1: Day 2 – Competition on the rise!

Today saw the conclusion of week 1 elimination matches and the first eliminations of the season! As before, I viewed as many matches from as many webcasts as possible, taking my sampling of the action mainly from the Washington DC Regional, the Peachtree Regional, the Traverse City District Competition and the Kettering University District Competition with a few matches from the Greater Kansas City Regional. I recorded data for 83 matches; 28 qualifying matches and 55 Elimination matches, and watched well over 100 in total. Before I jump into what’s changed from day one to day two I want to do a quick recap on some of the key points and data from day 1.

A quick recap.

Day one of week one saw surprisingly heavy and effective defense for qualifying matches. Many matches were decided by penalties, especially due to rule G46. Robots with effective ball control and a very agile drive system (the two tended to go together) had a significant edge, regardless of whether they kicked the ball into the goal or rolled it in. The approximate percent of matches that ended with 1 or more elevations was a mere 10.416%; elevations had a high chance of being game changers. Approximately 27.08% of matches had goals scored in autonomous and only about 14.58% of robots who attempted autonomous were successful. At the end of the day the average score per match per alliance based on my data was a mere 1.91667 points, the average based on all available match data was 2.12523 with the Kettering University District Competition having the highest average of 2.76829 and the Washington DC Regional having the lowest average, 1.40678.

So, how about day 2?

In general, the scores from day 2 saw an increase over day 1, as was expected. Some competitions saw a significant increase in the average qualifying score from day 1 to day 2, some hardly increased. All competitions saw and increase in average scores from the Qualifying matches to the Eliminations with the partial exception of the Washington DC and Bayou Regional, which saw a decrease in the average from day 2 qualifying scores to eliminations, but an increase from day 1 and overall.

Effective ball control, again, took the games by storm. A majority of points were scored by robots with effective, active ball control mechanisms that stayed in their home zone and pushed balls into their goals. Effective ball control will no doubt be a prevailing factor in every regional to come, including the Championship. The same effect could be and was in a few cases achieved via superb drivers. As expected drivers tended to be better overall, being generally more efficient and proactive in their driving. A select few teams had drivers that really shined, however they tended to drive playing defense, not offense so more often than not excellent drivers helped to preserve scores and keep scores low, not create high scores. The highest scoring matches were generated when alliances had a robot in their home zone with either and effective driver or great ball control mechanisms who would push balls in, and a robot mid field with decent kicking capabilities who could occasionally make a long distance shot but more importantly were able to give their partner in the home zone a relatively constant supply of balls. This pairing of a home zone robot with excellent ball control and a decent kicking robot in midfield to “feed” the home zone should remain an effective offensive strategy throughout all the regional competitions.

As in day 1, penalties, unfortunately, remained deciding factors in many matches. There is not much more to add from day one, G46 remains the dominant penalty. Another match killer was G17 Ball Return Timing which produced some staggeringly high penalties; the highest I saw was 26 points in penalties from this rule alone. Generally, G17 only comes into effect when an alliance forgets to return a ball completely; this penalty can be avoided by having alert human players who are well versed with the game rules.

At this point I would like to split this analysis into 2 sections, the remaining qualifying matches (which I will try to be brief with) and the eliminations. At the transition between the two I will examine how the hotly debated new seeding system comes into play, and how the defensive and offensive strategies discussed previously work into this system.

Day 2 Qualifying matches.

There was an overall increase in the level of competition, including a general increase in the scores, autonomous points scored and elevations. There was an increase in the average points scored per alliance per match of 1.3869 from 1.91667 on day 1 to 3.30357, based on the scores of my sampling of matches. The percent of matches in which a successful autonomous was present effectively doubled increasing by 26.4914% from 27.08% to 53.5714%. The number of matches with successful elevations had even higher gains, tripling from day 1 to day 2, increasing from 10.416% to 32.1429%, an increase of 20.7269%. It’s also worth noting that matches in which more than 1 robot sufficiently elevated was much more common than in day 1.

The following is a comparison of average score per alliance per qualifying match for every regional with available data from day 1 to day 2.

San Diego Regional:
1.9824 to 2.68 UP .6976
Washington DC Regional:
1.40678 to 2.4666 UP 1.05928
Peachtree Regional:
1.62281 to 1.64 UP .01719
Bayou Regional:
1.93023 to 2.61765 UP .68742
Greater Kansas City Regional :
2.2971 to 3.233 UP .9359
BAE Granite State Regional:
2.59259 to 3.73077 UP 1.13818
Finger Lakes Regional:
2.52941 to 3.43478 UP .90537
Autodesk Oregon Regional:
2.27419 to 3.3 UP 1.02581
Traverse City District Competition:
2 to 2.45 UP .45
Kettering University District Regional:
2.76829 to 3.5 UP .73171

Grand Total:
2.12523 to 2.92751 UP .80228

Overall, the scores increased. The average score from day 1 to day 2 based on all the available match data increased by 37.75%; .80228 points. This is a fairly reasonable increase, and I expect to see similar percentage gains from day 1 to day 2 in the coming weeks, with overall scores gradually increasing. It is probable we will see the overall increase in scores escalate around week 3-4 since we will start seeing teams who are attending multiple regional competitions back in action with a large experience advantage. Along with the increase in average score there was the aforementioned doubling of autonomous and tripling of elevating. This seems in line with the original thought that both autonomous and elevating will continue you play sizable roles in match outcome. However, I expect elevating to quickly drop off as game changers from day 1 to day 2 in future regional competitions as the number of robots capable of elevating will not likely increase while the amount of goals scored will. Autonomous will probably be a competitive edge for a select few teams who are able to get 3 or more balls consistently scored in autonomous. Currently I have not seen more than 2 balls scored in autonomous by a single robot, let alone an alliance.

The basic game play strategies used in day 2 remained about the same. The pattern of game play is not likely to change until we see repeat competitors in future weeks. Overall it seems there will be a gradual shift towards offense for the qualifying matches. For better or worse defense was just as present as in day 1 and just as effective if not more effective. The same general defense strategies were used; direct defense of the goal was most common, followed by “annoyance” type defense (getting in the way, pushing matches) of mid field robots. Although when done tunnel blocking remains an effective form of defense, especially against bots that cannot traverse the bump, there were not many instances were the tunnel was blocked. This is probably due to the fact that more bots were able to cross the bump today and those robots unable to would generally stay in their home zone or play defense in their opponents home zone. The increase in robots going over the hump appears to be due to a general increase in driving ability and a decrease of the defense strategy of waiting on the other side of a bump across from a crossing robot as was expected. Overall fewer robots were flipped in competition, significantly less toppled while traversing a bump and there was an increase in tipped robots from getting into pushing matches against robots with a lower center of gravity.

The new seeding system.

This year sees the arrival of a new seeding system. This system is a topic of debate among FIRSTers, some seeing it as unfair, others seeing it as great system. Before we relate it to week 1 we need to understand how the system works.

A robot gains seeding points at the end of every match. The points are calculated using Match Seeding Points plus a Coopertition™ Bonus. The method for calculating Match Seeding Points and the Coopertition™ Bonus are defined as follows in sections 9.3.4 and 9.3.5 of the FRC manual:

**9.3.4 Match Seeding Points **

  • All teams on the winning ALLIANCE will receive a number of seeding points equal to the penalized score (the score with any assessed penalties) of the winning ALLIANCE.
  • All teams on the losing ALLIANCE will receive a number of seeding points equal to un-penalized score (the score without any assessed penalties) of the winning ALLIANCE.
  • In the case of a tie, all participating teams will receive a number of seeding points equal to their ALLIANCE score (with any assessed penalties).

9.3.5 Coopertition™ Bonus **

  • All teams on the winning ALLIANCE will receive a coopertition bonus: a number of seeding points equal to twice the un-penalized score (the score without any assessed penalties) of the losing ALLIANCE.
  • In the case of a tie, all participating teams will receive a coopertition bonus of a number of seeding points equal to twice their ALLIANCE score (with any assessed penalties).

Simply put, the winning alliance gets their penalized score plus 2 times the losing teams’ un-penalized score in while the losing alliance gets the winning alliances’ un-penalized score. The gripe many people have with this system is that it allows for the possibility of the losing team to get more seeding points then the winning alliance. Generally this only happens in blowout matches where the winning alliance has a very high score and the losing alliance a very low score, for example 9 to 0, but the winning alliance gets a significant number of penalties, say 4. That means in this example the losing teams get 9 seeding points (the un-penalized score of the winner) and the winning alliance teams get 5 points (their penalized score (5) plus double their opponents score 2*0). While this is a strange outcome, I would point toward what the new seeding systems promotes.

Assuming certain variables are held relatively constant, the new seeding system eventually boils down to two basic categories: the safe bet and the risky bet. With the safe bet you get less out if you win, with the risky bet you get more out if you win. One of the GDC’s trademarks is providing competitions in which teams are presented with two extremes, for example, in 08 a large hard to control object on the ground that you need to control to a very large height. Similarly, this year you have easy terrain versus difficult terrain playing off of low versus high via the bump, the low clearance tunnel and the high tower. Teams can choose to specialize in one area and do well, while those who innovate and bridge the gap between extremes do excellent. The same concept is now present in the ranking system, the extremes being the safe bet (extreme cooperation) and the risky bet (extreme competition), with teams that learn to effectively manage how they place their bets being the innovators who will likely come out on top. The “safe bet” is playing the game as a multi-alliance effort, IE working with the opposing alliance instead of against. This entails helping your opponents score points by scoring on your own goal and not defending it with the overall goal of getting as many points scored as possible. This is based off of section 9.3.4, where you will receive the un-penalized score of your opponent while your opponent gains their penalized score. So and ideal scenario would be for example a score of 17 – 0 with the winning team receiving no penalties, this would give ALL the teams competing 17 seeding points, benefiting all because of the high score. There is one pitfall to the safe bet however; the opposing team must be on-board with your alliance for the “safe bet”. If they are not you run the possibility of your “safe bet” turning into a reduced risk “risky bet” for the opposition. The “risky bet” is much more along the lines of what FIRSTers are used to, that is to say playing against the other team and playing to win with one additional step; playing at an almost equal level. Just as with the safe bet teams will always want to try and keep the scores as high as possible, but they also want to keep the scores as CLOSE as possible while still coming out on top. This is considered a “risky bet” due to the fact that keeping scores close makes you more likely to lose due to penalties or just being outplayed thus losing any advantage. For a team taking the risky bet the ideal outcome would be for the opposing alliance to have more un-penalized points at the end of the match than your alliance but have the opposing alliance lose due to penalties. For example, an un-penalized match score of 8 – 9 where your opponent scores more raw points but a final penalized score of 8 – 7 with your alliance winning. This would result in 26 (your penalized score, 8, plus two times the un-penalized score of the losing alliance, 2*9) seeding points for teams on your alliance but only 8 (the winning alliances un-penalized score) for the teams of the losing alliance. This particular outcome is a huge long shot, the more common scenario would be trying to end with slightly more points than the opposing alliance, for example 7 – 5 , so you would still get significantly more seeding points than the opposing alliance teams, assuming you win the match with penalties applied.

Now that the seeding system has been fully explained and the strategies analyzed, we can see if it applied week 1. Below is the win/loss ratio of the top seed for each week 1 regional:

Team #2543: 8/1 = 8 (Won 80%, Lost 10%, Tied 10%)
Team #1727: 4/3 = 1.3 (Won 44.4%, Lost 33.3%, Tied 22.2%)
Team #1912: 8/0 = 8 (Won 80%, Lost 0%, Tied 20%)
Team #2345: 10/1 = 10 (Won 90.9%, Lost 9.1%, Tied 0%)
Team #1922: 8/2 = 4 (Won 80%, Lost 20%, Tied 0%)
Team #1551: 10/1 = 10 (Won 90.9%, Lost 9.1%, Tied 0%)
Team #997: 7/2 = 3.5 (Won 77.8%, Lost 22.2%, Tied 0%)
Team #1918: 9/2 = 4.5 (Won 75%, Lost 16.7%, Tied 8.3%)
Team #67: 11/1 = 11 (Won 91.7%, Lost 8.3%, Tied 0%)

Below is some of the average scores of the alliance of the #1 seeded team versus the average score of their opposing alliance:

Team #2543: 3.9 | 2.2
Team #1727: 3.88 | 1.88
Team #1912: 4 | 1.7
Team #67: 6.92 | 3.42

Below are the same stats for the lowest ranked seed from the same competitions as from the previous set of data:

Team #1372: 1/6 = 0.17 (Won 10%, Lost 60%, Tied 30%)
Team #3150: 1/7 = 0.14 (Won 11.1%, Lost 77.8%, Tied 11.1%)
Team #2556: 3/4 = 0.75 (Won 30%, Lost 40%, Tied 30%)
Team #1025: 2/8 = 0.25 (Won 16.7%, Lost 66.7%, Tied 16.6%)

Team #1372: 1.1 | 2.1
Team #3150: 0.33 | 2.22
Team #2556: 2.2 | 1.8
Team #1025: 1 | 3.17

The general match strategy teams chose to use for their qualifying matches is one that matched the traditional competition, IE trying to win the match by as much as possible; thus there was much defense played by bots who did not have great kicking or ball control abilities. Many people, me included, were surprised by the level of defense for the qualifying matches. While it did seem high I don’t expect to see it go away until either the last regional competitions or the championship, mainly because in some cases that’s what some robots are best at and they would want to showcase their defense capabilities for others teams’ scouts. This is understandable especially because defense becomes hugely important in the eliminations where all that matters is getting the higher score. Looking at the seed stats above it does not seem either of the strategies discussed above came into play. This is due to the condition I mentioned at the start of the seeding sytems analysis that those strategies apply only with certain variables held relatively constant. These variables are mainly the level of competition and the average scoring capabilities of the robots competing. Without all the robots being generally as competitive as the next and having the same average scoring capabilities as the next, the two “bets” don’t always come into play. Because the skill levels and scoring capabilities came in such a wide ranges it allowed teams to just go for the highest scores possible. Just as both “bets” are contingent on scoring as many points as possible, just in different ways, those robots that could consistently score high ended up seeded highest since they just physically gained more seeding points that other teams. Their high win/loss ratio reflects their efforts to win by scoring high. All they had to do was score higher on average than their opponents, which can be seen in the match scores. Conversely the lowest seeded team had the lowest average score. If you notice the average opponents score for both the lowest and highest seeds were about the same, it was mostly their average score that affected their rank. Regardless of strategy it is obvious that goals = seeding points. Until the level of competition evens out, such as in later weeks or nationals, teams who can score goals the best or play their alliance partners and opponents to get maximum seeding points will continue to seed well; and so long as the level competition keeps a wide range there will continue to be a general correlation between match wins and rank.

The Eliminations.

The elimination matches saw and even greater increase in the level of competition as was expected. As from day 1 to day 2 there was an increase in the percent of matches that ended with elevations and in matches with a successful autonomous as well as an increase in the average score. The percent of matches with successful autonomous modes increased from the day 2 qualification matches rates of 53.5714% to 67.2727%, a gain of 13.7013%, with an even larger difference between the overall qualifying match average to the eliminations average. The same was again true with successful elevations, day 2 having 32.1429% and the eliminations seeing 40% of matches with successful elevations, an increase of 7.8571%.

The game play strategies used in the eliminations were not too dissimilar from the strategies used in the qualification matches. Unlike qualifications, winning the match is a requirement, making defense a must. Teams that added dedicated defense bots to their alliances had a sizable advantage. The most common defensive strategy was to send the defense robot to the opposing alliances home zone to block goals and push around the scoring robots. One slight variation of this strategy was sending a robot with a powerful kicker into the opposing alliances home zone to kick balls to either the middle zone or their home zone, leaving no balls for the opposition to score while increasing the number of balls available for their alliance to score with. The basic scoring strategy was similar to the one that evolved from qualifications, where an agile robot with a very good ball control mechanism stayed in the home zone pushing balls in and a decent kicking robot stayed in mid field “feeding” the robot in the home zone. Robots with great ball control again had the edge, those who were agile and had great balls control really shined. I personally did not see any robots with swerve drive in the finals, though a sizable portion did have mecanum; those with a good mecanum system and solid ball control did well.

Penalties were still present and consistently a deciding factor in elimination matches, although the average penalty amount appeared to decrease. G46’s, the main source of penalties, went down quite a bit, occurrences of other penalties stayed relatively constant. G17 still killed a few matches with huge penalties. G17 is possibly one of the most easily avoided penalties. Human players MUST stay alert and return balls quickly.

There were significantly less tipped robots in the eliminations. A majority of robots in the eliminations had no trouble going over the bump; in fact crossing the bump became a very common occurrence as alliances shifted teams between the home zone and the middle zone, as well as between offense and defense. The original strategy of waiting on the other side of a bump to upset an incoming robot more or less completely disappeared in the eliminations. Generally the only robots that tipped on the bump were those who had trouble straddling the bump in order to deploy their lift mechanisms. A majority of tipped robots came from defense. With the bumpers this year being up much higher and the center of gravity needing to be higher as well, any robots that get into a pushing match generally end up with one end in the air, causing many to tip. When the defense tips the robot doing a majority of the scoring, especially early on, it can drastically affect the outcome of the match. The defense should be careful when getting into pushing matches as tipping an opposing robot may result in a G36 ROBOT to ROBOT Interaction penalty.

The increase in average elimination match scores increased sizably with the exception of the Washington DC Regional and the Bayou Regional which actually decreased from day 2 to eliminations. The average score per alliance per match from the matches I recorded data for was 4.09091; an increase of .78734 from the day 2 qualification match average and over double the day 1 qualification matches average. Below if a comparison of the average day 2 qualification scores to the average elimination scores for each regional:

San Diego Regional:
2.68 to 3.375 UP .695
Washington DC Regional:
2.4666 to 2.30952 DOWN .15708
Peachtree Regional:
1.64 to 3 UP 1.36
Bayou Regional:
2.61765 to 2.25 DOWN .36765
Greater Kansas City Regional :
3.233 to 4.11905 UP .88605
BAE Granite State Regional:
3.73077 to 4.4545 UP .72373
Finger Lakes Regional:
3.43478 to 3.725 UP .29022
Autodesk Oregon Regional:
3.3 to 3.30952 UP .00952
Traverse City District Competition:
2.45 to 3.15 UP .7
Kettering University District Regional:
3.5 to 5.28571 UP 1.78571

Grand Total:
2.92751 to 3.52184 UP .59433

I expect this general rate of increase from day 1 to day 2 to eliminations to remain about the same, with a slight overall increase in scores as the week’s progress. I also expect autonomous to get better as the regional competitions progress. It does not seem that autonomous currently has much of an impact on the match outcome and I do not expect it to in future competitions; although it will progressively get better it will likely remain not significant enough to swing a match from the start. Successful elevations were also much more common in the eliminations but they are not generally huge game changers, they seem to be more of an insurance system against penalties. For now with the average scores as low as they are they retain the potential to swing a match but I expect soon that the average score will overtake the value of elevating, especially in these eliminations matches where all the top scorers will end up.

With the first week of competition at a close I would say we’re in for some exciting competition in the coming regional competitions and the championship! Until the playing field levels so to speak, the new seeding system will not come into play much as some fear it will; teams will continue to go for high scores and there will continue to be defense in the qualification matches. If and when the playing field levels out I expect a whole new level of depth to come into play that will separate the teams with good robots and good strategies from those with simply a good robot. Ball control will continue to reign supreme and when paired with particularly agile robots may become deadly, especially if they can shoot from mid field where there will be significantly less defense. Week 3-4 should see a jump in the average scores of these matches as we will start to have experienced teams play for the second time, and as seen but the large increase in scores from day 1 to day 2, experienced drivers and teams will bring a lot of points to the table from the get-go. Defense will remain an integral part of this year’s game throughout the entire season, with a particular emphasis in the eliminations.


  • *Austin Steeno *(Team 39 and 2837)

GO FIRST! :smiley:

As before comments, questions, addition data are always appreciated! Do you think my analysis was flawed? Accurate? Also, would you be interested in me continuing to post my analysis in future weeks? For all I know I’m just a FIRST enthusiast who’s rambling on needlessly! :ahh:


Incredible writeup, Austin! Most of the points you made were along the lines of what I noticed from watching webcasts, as well.

Regarding seeding points: your allusion to the typical complicated/simple set up that FIRST competitions have is very interesting… I never made that connection. I would think a lot of teams would need to think that way before the 6v0 approach becomes really popular.

The near-zone close range scorer, when combined with a very competent feeding bot from mid-zone looks like the combination that takes most matches. I expect it to become more prominent as teams get better at driving/ball control. I did see a few matches that were focused on heavy defense and relied on end game bonuses for the win. These will probably be phased out in a few weeks, but might still be a common occurrence in eliminations. I suppose we shall see!

EDIT: Thought I would put up the data I was playing with. It’s just the qualifying results/rankings from all the week one regionals. I put some notes in for seeding points and match scores, but they were mostly personal notes so that I could see how the game was changing over the course of a regional.


ok…all I can say is WOW!

Great work Austin.

Perhaps the autonomous will shake out like 2008. It did not seem decisive at the early regionals in 2008. By nationals it was critical.

In 2010 with only one bot doing scoring a point or 2, it is not a major game changer. but with 2 or 3 bots scoring 1 to 2 points in later week eliminations or at Nationals , it becomes decisive.

Elevations are a different deal. With effective scorers, perhaps it will be more of a wash between scoring 3 more goals and doing the lift.

Mentor Team 39

Hanging is huge during eliminations MORE than the seeding matches.
The playoffs tend to have more even matches, and the alliance that has a hanger on their alliance vs. one that doesnt is significant.
Our alliance in SD had a broken winch in the playoffs and no one on our alliance could hang at that point. 1717 and 599 could do it consistently this past weekend. Having to outscore your opponent by 3 points or more in order to win was extremely difficult, and was the reason why we lost matches in eliminations.
Remember, two effective opposing alliances especially in the finals both have great scoring bots and effective feeders.

I cant wait to see the robot that can do 5 pointers (and yes, there will be one soon) in time for CMP.:slight_smile:
At CMP, I believe that every alliance will make sure they get at least 2 hangers (either 4 or 5 points) at the end. Otherwise, they dont have a chance.

Great write up, but I think you forgot P’tree regional’s top seed.

Team #1466 went 9-0-1 during qualifications.

The above with 1625 in the back is probably my dream alliance. Some fancy little swerve or mecanum striker combined with a hybrid offense / defense middleman should be ideal for a few weeks to come.

Another great post Austin! Just a tip, you might want to save yourself (and us!) the headache and include less significant figures in your percentages, having percentages out to the 4th decimal place is a tad daunting. Other than that, I am a huge fan of the data you pull together and your very thorough observations. They reenforce some of the things I’ve seen and put other things into a whole new light. Thanks!

Indeed, G46 is THE common penalty. Teams, no matter what you did, chances are good you’ll ride up on a ball, so start thinking NOW about what you can do to prevent it a little.

G17 isn’t just unconscious human players; those big penalties happen when a ball is scored by the system but not recorded as going back onto the field for some reason - it them becomes impossible to satisfy the DOGMA and wow, do those penalties add up fast.

But, we did see one trident operator actually leaning on the trident, watching the match, while his alliance partners literally handed him the balls scored. Needless to say, penalties lost the match for them.