Relative Division Rankings

[Citrus Dad]

Here's some statistical measures comparing divisions. I've included both average OPRs and best OPRs, and the average of the Top 8 teams for best OPRs (I think that's a better measure of team potential). Newton is by far the weakest division. Galileo has the best overall average, but 2nd lowest Top 8 ranking. This may make it the most competitive. Archimedes is strongest overall. Curie falls in the middle in each category. (Sorry, looks better in Excel-see attached image.)

Avg. OPR Avg. Rank Max OPR Max Rank Top 8 Max Rank

Archimedes 29.5 646.1 38.7 615.0 81.7 11.0
Curie 28.9 672.9 37.4 671.2 78.9 16.4
Gallileo 31.0 577.3 39.3 576.5 78.7 17.6
Newton 26.4 718.0 35.6 650.6 68.5 34.4

[Grim Tuesday]

Thanks for going ahead and doing this, it's quite revealing. I think now is a great time to remind everyone that just because a division is weaker overall than another doesn't mean it won't produce the Champion. Who would have thought that the 4th seed from Galileo (iirc one of the weaker divisions last year, though they were all pretty balanced) would win it all last year?

[2789_B_Garcia]

Originally Posted by Grim Tuesday

Thanks for going ahead and doing this, it's quite revealing. I think now is a great time to remind everyone that just because a division is weaker overall than another doesn't mean it won't produce the Champion. Who would have thought that the 4th seed from Galileo (iirc one of the weaker divisions last year, though they were all pretty balanced) would win it all last year?

+1 this

OPR<strategy+teamwork

Particularly since OPR is calculated using numbers from qualification matches, and the dynamics in qualification matches are different from the dynamics of elimination matches, and I can only assume (this will be our first trip to the big dance) that champs quals and elims will have different dynamics than matches at district, regional and even district champs matches based solely on the fact that there will be much more variety in the robots and teams.

I always like to consider the value of good qualitative data

[xSAWxBLADEx]

Will Curie break the curse? Statistics show not. :/

[Craig Roys]

To me, this shows the divisions to be closer overall than thought. There is a range of only 4.6 on average OPR - can you tell me that a robot with an OPR of 31 is significantly better than a robot with an OPR of 26.4? Same goes for the average Max OPR...range of only 3.7. The top 8 Max OPR is quite a bit more revealing...Newton does appear to be quite a bit lower than the other 3 divisions on the top end. The fact that the overall average in not significantly lower makes me think there might be a bit more balance.

Thanks for bringing up the data - this should be a lot of fun. I agree with the previous posts about not counting anyone out...the right alliance combination who gets hot at the right time can go far.

[Citrus Dad]

Originally Posted by Craig Roys

To me, this shows the divisions to be closer overall than thought. There is a range of only 4.6 on average OPR - can you tell me that a robot with an OPR of 31 is significantly better than a robot with an OPR of 26.4? Same goes for the average Max OPR...range of only 3.7. The top 8 Max OPR is quite a bit more revealing...Newton does appear to be quite a bit lower than the other 3 divisions on the top end. The fact that the overall average in not significantly lower makes me think there might be a bit more balance.

Thanks for bringing up the data - this should be a lot of fun. I agree with the previous posts about not counting anyone out...the right alliance combination who gets hot at the right time can go far.

A 20% range in average OPRs is pretty significant. I'd expect them to be much closer, even with random draw over 100 teams in each division.

The relative rankings aren't so useful for predicting Einstein results (because those are "one-off" competitions) but rather for comparing relative difficulty for different teams to achieving top qualifying positions and easier elimination routes. For example, 1986 appears to have an easier route to the Newton Division finals than the top teams in Archimedes.

The OPRs are a strong predictor to team success. Combining offensive statistics (which are more accurate than OPRs) with quantitative defense/skill ratings we developed, we were able to accurately predict 80+% of the qualifying match results, and 6 of 7 elimination round outcomes in two tournaments. The OPRs are slightly less accurate, but cannot be dismissed. In fact they are a stronger predictor than qualifying points. We came from the #6 Alliance Captain position to win our Regional, but we picked up the best available offensive robot (an FCS) to win.

[Jaxom]

1114's scouting database has some additional info for comparing the different divisions: http://www.chiefdelphi.com/forums/sh...77#post1264177

It's on the "Divisional Summary" tab.

[Craig Roys]

Originally Posted by Citrus Dad

The relative rankings aren't so useful for predicting Einstein results (because those are "one-off" competitions) but rather for comparing relative difficulty for different teams to achieving top qualifying positions and easier elimination routes. For example, 1986 appears to have an easier route to the Newton Division finals than the top teams in Archimedes.

I'm not a statistics guru by any means, but on the surface I'm not sure I agree with this statement - please correct me if my thinking is off. Let me preface by saying, I believe 1986 would do well regardless of where they play as they have a tremendous robot this year. I think it may actually be tougher in Newton given that the ave max OPR for the top 8 is lower by 10-13 points from the other three divisions, but the overall average max OPR is only lower by 3.7 points. Unless I'm mistaken, isn't that saying the Newton is a little more balanced than the other fields (not stronger, just more balanced)?

[connor.worley]

Originally Posted by Craig Roys

I'm not a statistics guru by any means, but on the surface I'm not sure I agree with this statement - please correct me if my thinking is off. Let me preface by saying, I believe 1986 would do well regardless of where they play as they have a tremendous robot this year. I think it may actually be tougher in Newton given that the ave max OPR for the top 8 is lower by 10-13 points from the other three divisions, but the overall average max OPR is only lower by 3.7 points. Unless I'm mistaken, isn't that saying the Newton is a little more balanced than the other fields (not stronger, just more balanced)?

The number of really great teams in the top 8 could be smaller (lowering the average), but those great teams will still dominate.

[tdmilnes]

Originally Posted by Jaxom

1114's scouting database has some additional info for comparing the different divisions: http://www.chiefdelphi.com/forums/sh...77#post1264177

It's on the "Divisional Summary" tab.

Hey Jaxom,

Thanks to 1114 for putting this together!!!

Just a few notes from 2199. We actually played 9 matches and were 5-4 at Virginia. Although the 9th match was unofficial for us it wasn't for the teams we played against and its was our highest scoring (49 points). You also missed our Chesapeake stats. We were 6-2 and ranked 11.

[Citrus Dad]

Originally Posted by Craig Roys

I'm not a statistics guru by any means, but on the surface I'm not sure I agree with this statement - please correct me if my thinking is off. Let me preface by saying, I believe 1986 would do well regardless of where they play as they have a tremendous robot this year. I think it may actually be tougher in Newton given that the ave max OPR for the top 8 is lower by 10-13 points from the other three divisions, but the overall average max OPR is only lower by 3.7 points. Unless I'm mistaken, isn't that saying the Newton is a little more balanced than the other fields (not stronger, just more balanced)?

Some of my statement comes from more details than I revealed in my post. 1986 would be only the 2nd or 3rd best in the other 3 divisions--it has a clear lead in Newton. Galileo is probably the most competitive, with a higher overall average, but its Top 8 is lower than Curie or Archimedes. But you're correct that Newton could be very competitive in the elimination rounds, and 1986 is dependent on its alliance members.

[Craig Roys]

Originally Posted by Citrus Dad

Some of my statement comes from more details than I revealed in my post. 1986 would be only the 2nd or 3rd best in the other 3 divisions--it has a clear lead in Newton. Galileo is probably the most competitive, with a higher overall average, but its Top 8 is lower than Curie or Archimedes. But you're correct that Newton could be very competitive in the elimination rounds, and 1986 is dependent on its alliance members.

Yeah, I looked a little more closely at the OPR distributions. Newton is fairly consistently lower with the exception of a small stretch near the bottom. Hopefully, Newton can go against the numbers and do some damage - I know we plan to do our part to help make that happen!

[Gregor]

Originally Posted by Citrus Dad

...and 1986 is dependent on its alliance members.

What do you mean? 1986 is one of the few bots that is not dependant on its alliance members.

[2789_B_Garcia]

Originally Posted by Gregor

1986 is one of the few bots that are not dependant on its alliance members.

I second that after playing defense against them at Hub City.

[Jaxom]

Originally Posted by Gregor

What do you mean? 1986 is one of the few bots that is not dependant on its alliance members.

Thank you for the vote of confidence, but all bots are dependent on their partners. There were certainly regional matches where teams could win the match all by themselves (we did that some, as did a lot of other bots) but in eliminations that would be more rare. At CMP, I don't expect that to happen very much, if at all; certainly not in eliminations.

Look at the average scores for MSC. The average *losing* score in elims was well over 100. The average winning score was upwards of 150, iirc. No single bot is going to do that by itself, if being the only scorer on its team allows the opponents to play a lot of defense.