Offensive Power Rankings for 2008

[Beldo]

Bongle-

Are you going to do recreate the rankings after all the regionals are done and before the finals in Atlanta?

[Kyler]

Originally Posted by neshera

Hey Kyler:
Do you want me to send a slide rule with Nathan tomorrow to help you out?
Nesher

Lol I think I can have the computer do most of the math for me. I just have to understand it... thats my hope at least.

[Bongle]

Originally Posted by Beldo

Bongle-

Are you going to do recreate the rankings after all the regionals are done and before the finals in Atlanta?

I could re-do them once this week's regionals are complete, that's not a problem.

I'll do an Atlanta-only one after ATL too if there is demand.

[IKE]

I am not a programmer, but here is an interesting challenge for someone with some time on their hands. Once Atlanta divisions are posted, use the OPR to pick the winning alliance assuming that the rankings and draft follow the OPR (alliances would be 1,2,24-3,4,23-5,6,22......). Obviously the OPR will pick 1114 and some other powerful hurdler as the winners, but I am curious how closely it may actually predict things.

[Bongle]

Originally Posted by IKE

I am not a programmer, but here is an interesting challenge for someone with some time on their hands. Once Atlanta divisions are posted, use the OPR to pick the winning alliance assuming that the rankings and draft follow the OPR (alliances would be 1,2,24-3,4,23-5,6,22......). Obviously the OPR will pick 1114 and some other powerful hurdler as the winners, but I am curious how closely it may actually predict things.

To go even further:
Use the match scheduling algorithm to generate a schedule, then simulate each division's games, assuming that the alliance with the higher OPR always wins. Now you've got more accurate simulated rankings, because you may even have a weaker team get into the top 8 due to luck with the scheduling. However, you'd have issues because in the real world there'll be draft rejections which you can't plan for.

Then do the division playoffs and Einstein simulation. I may do that after my first exam when I have a long time off.

[IKE]

Hate to ask for more, but here goes:
I noticed our OPR drop dramatically when week 1 was added in (we had a lot of technical difficulties). Is anyone planning on doing an OPR ranking only using a teams last regional? This may more acurately show how the teams will perform at Atlanta....

[XaulZan11]

Originally Posted by IKE

Hate to ask for more, but here goes:
I noticed our OPR drop dramatically when week 1 was added in (we had a lot of technical difficulties). Is anyone planning on doing an OPR ranking only using a teams last regional? This may more acurately show how the teams will perform at Atlanta....

That's a really good point. I know when I'll prescout for my team, I'll focus on the team's last regional because it usually the best predictor to see how a team will preform. Maybe you can weigh the different regionals (such as if a team attends 2 regionals, the first regional may be 30% while the second will be 70% of thier total score).

[OZ_341]

Originally Posted by IKE

Hate to ask for more, but here goes:
I noticed our OPR drop dramatically when week 1 was added in (we had a lot of technical difficulties). Is anyone planning on doing an OPR ranking only using a teams last regional? This may more acurately show how the teams will perform at Atlanta....

Good point! We are definitely a much better machine now then we were at Chesapeake. Our drivers also have more confidence in the machine.

[Kyler]

Well, just something interesting I have found. Its to be expected and nothing major but it dies go to show being very percise (adding 4 decimal points) Can change things quite a bit in ranking the teams. To test my own Power Ranking program I ran the GTR results and found my results were different that bongles by up to a point (might not seem like much but once you get to the lower power rankings it is 10 % +) and it changes orders a lot. Here's what i found.



Nevermind, I entered one score wrong. My results are the exact same but with 4 excta decimal points.



I only included a few differences because the teams start to change then and I was too lazy to sort that all out. Also, a big THANKYOU! to everyone that helped me understand this stuff.

[Jacob Plicque]

Kyler,
Will you post your program code for review? I think that your are on to something by adding precision to your calculation.

[Guy Davidson]

Wee, I managed to break java

My inversion algorithm for the 1465x1465 or so matrix runs out of heap space. I'm currently working on an alternative way to get p. Right now, I'm going from Mp = s to p = Minverses. I might try to create an augmented matrix and rref it to get the solution. Any other ideas?

EDIT: Nevermind, I just solved the equation Mp = s.

Greg, I now have the OPR's generated from the entire csv dump you generated. Is there any way you could generate other csv dumps to analyze? By regional, by week, over the last three weeks, etc., all sounds like interesting sets of data to analyze.

[Guy Davidson]

I have the rankings generated from Greg Marra's csv dump of the TBA database.

Top 50:

1114	66.10642467
2056	54.11894734
233	50.73081127
330	48.61296306
987	48.40552663
39	48.06177403
525	45.48888639
67	44.23000818
1731	42.99226989
1124	42.40054838
217	42.10281885
103	41.29156933
40	40.86592771
175	39.4420245
191	39.06592804
469	37.90116922
20	37.75470786
79	37.65042852
1717	37.34613148
141	37.32830996
1024	37.2668216
968	35.93916079
1086	35.8662985
1126	35.81116812
368	35.65581359
494	35.0189302
2171	33.98399753
25	33.89508807
1625	33.79257371
365	33.63738509
254	32.91821392
1718	32.54432471
47	32.49881588
93	32.23816608
1477	32.14182287
1251	31.57996364
1806	31.33327422
16	31.18613493
69	31.04681657
1629	31.01151072
33	30.94122998
1418	30.31278837
195	30.2915486
27	29.90208154
383	29.89513167
71	29.87541039
287	29.76950247
171	29.71637571
148	29.4612061
877	29.34477514

The entire spreadsheet is attached. I'm working with Greg Marra on getting some more specific data sets and analyzing them, and after I do that, I'll post the results. Any specific requests?

[cziggy343]

for some reason, i dont think that ours is right... b/c we have had three hurdles or more in 75% of our matches and it shows that we score about 26 points per match, which w/ three hurdles (including lines) we should be just a little under 30... but of course i dont know what went into the compiling

[IndySam]

Is this data supposed to be your average score or your potential average? What exactly is your number supposed to mean?

[Bongle]

Originally Posted by IndySam

Is this data supposed to be your average score or your potential average? What exactly is your number supposed to mean?

It is an estimate of how many points your teams contributes per match. Note that it isn't simply your total score / 3, but by using the magic of linear algebra, you can solve for something a bit more accurate. It can help weed out robots that simply had good partners, and should be able to discern robots who many not have placed well but still contributed each round.

for some reason, i dont think that ours is right... b/c we have had three hurdles or more in 75% of our matches and it shows that we score about 26 points per match, which w/ three hurdles (including lines) we should be just a little under 30... but of course i dont know what went into the compiling

If you took a substantial number of penalties (1 every couple matches), that'd result in your offensive rating going down. This is because in addition to adding to an alliance score by hurdling, you'd be subtracting from it due to penalties. It's not a perfect number because we don't have an infinite number of matches to base it off of, but it seems to correlate nicely with observed robot performance.