paper: Championship Division Strengths

[Ed Law]

Thread created automatically to discuss a document in CD-Media.

Championship Division Strengths by Ed Law

[Ed Law]

From the weighed average OPR numbers, I put all 4 divisions on a chart and sorted in descending order. I also added the Michigan State Championship numbers as a comparison.

Using weighed average OPR data, the mean of each division is
Archimedes 17.27
Curie 21.04
Galileo 21.51
Newton 20.80
MSC 24.45

The MSC numbers are not the OPR the 64 teams got at MSC. They are the same weighed average OPR numbers projected to Week 7 and included the MSC OPR numbers.

Looking at the graph, I have the following observations
1) Archimedes is the weakest division in the whole range.
2) The kink/knee is at around 9 or 10 just like last year, meaning the top 10 teams drop off at a much faster rate in OPR than the remaining teams.
3) The original unstretched MSC curve is on the low end of all the divisions in terms of strength unlike last year. The reason it was still so exciting is it truncated off the bottom 23 teams.

You can change the numbers easily to use other ways to rank teams like best OPR or most recent OPR. Just copy and paste into the appropriate cells.

[IKE]

Soooo, maybe MSC will not be more competitive than the divisions this year. I for one am extremely excited about this. One of the more interesting things we learned at MSC is that if you haev an alliance capable of putting up over 100 points, you stand a really good chance of taking your match.

Going by the data:

Galileo has the highest probability of the #1 alliance sweeping their way through (but there are 3 players wanting to be in that tango).

Newton will likely have the most "seasoned" alliance on Einstein. There will likely be at least 4 incredibly strong alliances in that division battling it out for the spot of Einstein. They are also the most likely to ahve a very strong 3rd (look in the 20-26 range).

Curie will also have extremely strong 3rd picks, but it softens a bit quicker than Newton, thus the winner of that division won't have the practice that Newton will.

Archimedes... the bad news is that according to the data... we are quite a bit softer than the rest. The good news? We are strong enough. Using my team and 67 as an example. I don't think you need OPRs in the 70s to make a winning alliance. 70+70+30 does not equal 170 pts in this game. If your alliance has the ability to get between 100-120 points, you have a decent chance of going all the way.

For the event as a whole:
The elims will be even better than MSC. The top 30 or so are so strong in all of the divisions (these are the teams likely to make up your 24 elim teams). this is extremely exciting. Qualifying will have a lot more variety. There are quite a few softer teams coming to the world Championship. Many of them may have a 2-3 lb "secret weapon" that they are bringing along, so expect quite a few teams to bubble up into that 30-40ish OPR range.

I do predict that 3/4 #1 Alliance captains will have double autonomous capability.

[Jared Russell]

Originally Posted by IKE

I do predict that 3/4 #1 Alliance captains will have double autonomous capability.

Ironically, they may never need to use it in elims. There are easily 24 consistent single ubertube-capable teams in each division.

[Gdeaver]

This is all based on offensive analysis. Defensive strategy has been unusually absent and many times employed ineffectively. At some of the later regionals Defense became much more well implemented. If smart alliances employ effective defensive strategy and the human players get smart, I contend that you will not see 100+ scores.In fact 80 would be a good score. If teams employ smart defense they can shut down the offensive power bots. Then the game changes and uber tube and minibots are critical. If defense comes on at the championships, then the winners maybe the last bots functioning. If the play continues to be a gentleman's game of putting up tubes and no defense then pick the 25 top opr's 3 of them will be your champions and every one can sit in the stands and watch them. I would like to see some rumble in this game.

[Kevin Sevcik]

I'm going to side with history here and predict that we're going to see more defense in the Division Elims and especially on Einstein. With the strong alliances likely capable of scoring 3 ubers and 4+ logos on a field littered with tubes, the game would quickly come down to the minibot race. Once you've hung 3 high ubers and 4 logos, logo pieces are almost inconsequential, compared to a 1st place minibot. To beat a 1st-4th minibot finish, you need to be an entire low logo ahead of the opposition. That's a pretty tall task, even with this caliber of teams. Better to bail on logos after you fill the top two rows, and position for the minibot race and putting D on that faster minibot. Once an alliance determines they're not capable of winning a heads up minibot race, expect to see some heavy defense played on hostbots heading to the towers. Of course, the opposition may be thinking the same thing... Which is where strategy comes in.

Your other option is tube defense/denial. If your opponent is weak on throwing, assign someone to steal tubes or herd them into LANES. Just make sure you've got a smart team on this task. You can cripple an opponent a lot faster focusing on just their triangles or circles or squares. It's pretty hard to finish a logo without any circles, after all. If your throwers can consistently land tubes in your zone, then you've got a good chance of catching an opponent off guard and denying them a logo or more. If your throwers can't hit the zone, consider holding a piece back for hand delivery, and stealing as many as you can in the meanwhile. Because of this, I'm actually expecting to see a lot less tubes littering the field on Einstein.

(I hope half this theorizing pans out, so I can cancel out my early "Most everyone will slot load" theorizing.)

[MysterE]

I think the main reason that we won the Bayou was because of defensive strategy. To be honest - our robot cannot score 100 pts in a game individually, but we could hold a team that consistently scored at 70 - 100 in a game down to 30 - 45 points. I'll be interested to see how many other teams opt out of scoring in lieu of defensive strategy.

[IKE]

Originally Posted by Jared341

Ironically, they may never need to use it in elims. There are easily 24 consistent single ubertube-capable teams in each division.

Yep, this is absolutely correct, though the consistency of a couple teams double autos and crowding issue (2 is company 3 is a crowd in this years game) may want them to use the double.

When we did our game analysis for competing in Michigan, we knew we would need it for Kettering. Thought it would still be useful and possibly necessary for Troy. We figured it would come in handy for MSC, but completely unnecessary for Elims. Ironically some chassis improvements reduced its reliability after Kettering, but we should have it back to reliable for the Championship as I think it will be key for at least 2-3 qualifying matches.

[Joe Ross]

I wonder how much of the low OPRs in Michigan are due to amount of minibots in Michigan. A 2 second minibot would probably get 1st or 2nd in 90% (or more) of matches world wide. However, in Michigan, it might be down to 3rd or 4th.

For example, more then 50% of teams in Michigan had a lower OPR at the state championship then at one of their districts. I haven't looked at teams who had their 3rd or 4th event at regionals to see what percentage improved, but I'd bet its much higher.

If Michigan teams were given a 5 or 10 point minibot bonus to bring them inline with other events, I bet the line would be similar to last year.

[Vikesrock]

Joe, I think that's probably a significant factor in the OPR of both Michigan teams and others and can skew the numbers either way. At events with few reliable working minibots even a 2 to 3 second minibot can substantially inflate a team's OPR by taking 1st in most or all of their matches.

[Chris Hibner]

Originally Posted by Joe Ross

I wonder how much of the low OPRs in Michigan are due to amount of minibots in Michigan. A 2 second minibot would probably get 1st or 2nd in 90% (or more) of matches world wide. However, in Michigan, it might be down to 3rd or 4th.

For example, more then 50% of teams in Michigan had a lower OPR at the state championship then at one of their districts. I haven't looked at teams who had their 3rd or 4th event at regionals to see what percentage improved, but I'd bet its much higher.

If Michigan teams were given a 5 or 10 point minibot bonus to bring them inline with other events, I bet the line would be similar to last year.

I would also think that this can be attributed to the amount of good teams and the diminishing returns of scoring in this game.

For example, let's say your event has all robots that can each put up 6 tubes and all have < 1.5 sec minibots. Let's say that every match ends up 120 - 100. Then every robot would end up with an OPR of about 35, which is good, but not breathtaking.

Throw one of these robots into a weak regional where the 2nd best team can only put up 2 tubes and only a few teams have minibots. A 6-tube machine with a good minibot and consistent autonomous should put up 78 points per match by themselves. That would give them a very high OPR (65+) in the weak regional, but thrown into the tough regional described above it would probably be ~35.

[Ed Law]

Originally Posted by IKE

Soooo, maybe MSC will not be more competitive than the divisions this year. I for one am extremely excited about this. One of the more interesting things we learned at MSC is that if you haev an alliance capable of putting up over 100 points, you stand a really good chance of taking your match.

Going by the data:

Galileo has the highest probability of the #1 alliance sweeping their way through (but there are 3 players wanting to be in that tango).

Newton will likely have the most "seasoned" alliance on Einstein. There will likely be at least 4 incredibly strong alliances in that division battling it out for the spot of Einstein. They are also the most likely to ahve a very strong 3rd (look in the 20-26 range).

Curie will also have extremely strong 3rd picks, but it softens a bit quicker than Newton, thus the winner of that division won't have the practice that Newton will.

Archimedes... the bad news is that according to the data... we are quite a bit softer than the rest. The good news? We are strong enough. Using my team and 67 as an example. I don't think you need OPRs in the 70s to make a winning alliance. 70+70+30 does not equal 170 pts in this game. If your alliance has the ability to get between 100-120 points, you have a decent chance of going all the way.

For the event as a whole:
The elims will be even better than MSC. The top 30 or so are so strong in all of the divisions (these are the teams likely to make up your 24 elim teams). this is extremely exciting. Qualifying will have a lot more variety. There are quite a few softer teams coming to the world Championship. Many of them may have a 2-3 lb "secret weapon" that they are bringing along, so expect quite a few teams to bubble up into that 30-40ish OPR range.

I do predict that 3/4 #1 Alliance captains will have double autonomous capability.

I agree with others who posted on why the MSC teams have lower numbers after MSC. This is especially true because I was using weighed average when Week 6 event has more weight than earlier weeks. It has to do with the game itself.

Last year, there was no limit on total points because the soccer balls get recycled onto the field. That's why the a team's OPR did not "suffer" because everybody gets better. However this year, there is a maximum how many points an alliance can score. When other teams start to contribute more instead of only 1 team like in earlier weeks, the OPR numbers get spread out.

I think Michigan teams will do well at the Championship. Other high OPR teams outside of Michigan will find out at the Championship that they are not the only one who can score 6 tubes and have a <2 seconds minibot.

The other thing about the minibot, at the regional/district in earlier weeks, it is seldom to have 3 alliance partners who all have a minibot to deploy. Sometimes there is only one team. So if you have a minibot, you got to deploy it every time and so every time you are on the field, you contribute a lot of points. However at MSC, almost everybody has a minibot and only two got to score. Although OPR does not care which two teams score it, it gives credit to the whole alliance. However the score is not higher when your team is there anymore because of your minibot. Hence every team get less OPR points from the minibot. For those who did the study by separating out the minibot OPR, it will be very clear.

[JesseK]

In what ways can we actually try to normalize each regional to the others with respect to OPR? At "harder" regionals the Rank Scores would be higher... is there anything discernible in averaging the two or by performing other operations using the two?

E.g. [high OPR,med RS] could be less than [med OPR, high RS] between regionals?

[Ed Law]

Originally Posted by JesseK

In what ways can we actually try to normalize each regional to the others with respect to OPR? At "harder" regionals the Rank Scores would be higher... is there anything discernible in averaging the two or by performing other operations using the two?

E.g. [high OPR,med RS] could be less than [med OPR, high RS] between regionals?

RS is the average of unpenalized score of your opponents if you win or your score if you lose. In a way it does show how strong the event is. Look at the RS numbers of MSC. Even the bottom ranked teams have 50s and 60s. I think the average RS number is a good indicator but the relatively small standard deviation of the RS number means that all the teams are quite even.
I think comparing the average RS number and average OPR number would be a good indication already. If we have to perform some operations using the two, I will not take the average of the two because they are unrelated. You can't average apples and oranges. If I treat the RS and OPR as vectors, I would take the dot product then divided by the number of teams. Basically multiplying RS and OPR of each team and add them up, then divided by the number of teams to get the average of the product. Actually OPR is from each team and RS is from the whole alliance so RS should be divided by 3. However if all events are calculated the same way and compared to each other, the factor of 3 does not matter.
Anybody wants to try this out?

[Joe Ross]

I looked at the OPR for the 1st, 2nd, and 24th teams from each division. This would be the perfect alliance if opr perfectly reflected teams scoring ability, there was no advantage to playing defense, teams seeded perfectly, and teams did alliance selections perfectly. I also did the #8 alliance (under the same conditions), to show the depth of the divisions.

Archimedes 136 105
Curie 154 118
Galileo 169 114
Newton 162 122
MSC 139 103