Easy to use Offensive Power Rankings (OPR) program for mid-regional scouting

[Jared Russell]

The lack of correlation between OPR, DPR, and winning a regional is at first surprising. But if you step back and look at how several regionals have been won so far this year, teamwork and strategy seem to be the common theme.

I believe that in Lunacy, the strength of an alliance is not defined strictly by the sum of its parts.

[Bongle]

Originally Posted by Jared341

The lack of correlation between OPR, DPR, and winning a regional is at first surprising. But if you step back and look at how several regionals have been won so far this year, teamwork and strategy seem to be the common theme.

I believe that in Lunacy, the strength of an alliance is not defined strictly by the sum of its parts.

The most interesting correlation I've found is that the robots that get to go into eliminations correspond very strongly with the PM rating. At FLR, all but one of the top 20 robots ranked by PMR were either captains or were selected to go into eliminations. The correlation is less strong with OPR.

[Ed Law]

Originally Posted by Bongle

The most interesting correlation I've found is that the robots that get to go into eliminations correspond very strongly with the PM rating. At FLR, all but one of the top 20 robots ranked by PMR were either captains or were selected to go into eliminations. The correlation is less strong with OPR.

Hi Bongle,

I referenced your work in my white paper. I didn't realize you are mentoring Team 2702 now. I still associated you with Team 1281 in my presentation. I will update it. I really like your insight/explanation how to interpret these rating numbers.

I want to discuss with you about correlation between OPR and CCWM/PMR. I think we both agreed that it changes from year to year depending on the game. One way, as you suggest in your post, is to see whether the teams that were alliance captain or got picked and went to the elimination round have higher OPR rankings or higher CCWM/PMR rankings. This will tell us whether teams value pure offensive power or contribution to winning margin when they select teams. Perhaps we should exclude the alliance captains when we do this comparison.

Another way to look at correlation is how "predictive" the two different ratings are to outcome of elimination round matches. I did a study of the prediction of match results using OPR and CCWM. Through the first two weeks including those events that have complete data published, in the elimination round, the prediction using CCWM is 59.5% and using OPR is slightly better at 63.6%. One reason the correlation is not that good is because there were a lot of close matches that could have gone either way.

We should be careful not to draw conclusions from just one regional. Each of the two ratings correlate better for some regionals but not for others.

Regards,

Ed Law

[JesseK]

At first I thought that it would be unsolvable (each equation has 6 unknowns, and you'd end up with 2N unknowns and only N equations), until today I had the fairly-obvious brainwave that each match would give you two equations (one OPRred - DPRblue = scorered for red, one OPRblue - DPRred = scoreblue for blue). This approach seems like it would be more predictive of robot performance than simply doing OPR and DPR separately*. I'm going to try to implement it tonight, but if someone wants to try it themselves and report back, that'd be great

You would want to have (blueScore = blueOPR + redDPR) and vice versa. Remember, you having a low DPR is better for you, hence you want your opponents to have a high DPR. If you subtract the opponent's DPR then you're lowering your own score just because the opponents are easier to score on ... which doesn't make sense when the opponent's DPR is positive (which is pretty bad).

Some simulation data to come tonight; I have an idea for a quick and dirty Match Prediction method.

<edit> Quick match prediction using the formula above for the DC Quarterfinals:

blueScore = blueOPR - redDPR:
Teams: (RED) 234, 45, & 620 vs. (BLUE) 1111, 122, 768
Predicted Scores: Red = 43, Blue = 6
Actual Scores: Red = 126, Blue = 100


blueScore = blueOPR + redDPR:
Teams: (RED) 234, 45, & 620 vs. (BLUE) 1111, 122, 768
Predicted Scores: RED=101 BLUE=80
Actual Scores: Red = 126, Blue = 100

Note that the prediction for their second match was identical even when the new match data was added, so predicting the outcome of the second QF match when the scores are that close is nearly impossible. Predictions also don't take into account adjustments for strategy, which is probably why it's best to leave the actual statistics for Qualification Matches only.

[Ed Law]

Originally Posted by JesseK

You would want to have (blueScore = blueOPR + redDPR) and vice versa. Remember, you having a low DPR is better for you, hence you want your opponents to have a high DPR. If you subtract the opponent's DPR then you're lowering your own score just because the opponents are easier to score on ... which doesn't make sense when the opponent's DPR is positive (which is pretty bad).

Some simulation data to come tonight; I have an idea for a quick and dirty Match Prediction method.

<edit> Quick match prediction using the formula above for the DC Quarterfinals:

blueScore = blueOPR - redDPR:
Teams: (RED) 234, 45, & 620 vs. (BLUE) 1111, 122, 768
Predicted Scores: Red = 43, Blue = 6
Actual Scores: Red = 126, Blue = 100


blueScore = blueOPR + redDPR:
Teams: (RED) 234, 45, & 620 vs. (BLUE) 1111, 122, 768
Predicted Scores: RED=101 BLUE=80
Actual Scores: Red = 126, Blue = 100

Note that the prediction for their second match was identical even when the new match data was added, so predicting the outcome of the second QF match when the scores are that close is nearly impossible. Predictions also don't take into account adjustments for strategy, which is probably why it's best to leave the actual statistics for Qualification Matches only.

Hi Jesse,

I think the general consensus in the CD community agrees that OPR should be calculated with qualifying matches only. The extra matches by some teams who made it to elimination round will bias the results if they are used.

Ed

[DanDon]

Hey y'all...

Can anyone run the program for the NYC regional and post the results?
I reformatted and haven't gotten a chance to install VMWare Fusion on my mac again.

Thanks a bunch!
Dandon

[Vikesrock]

Here's the NYC results sorted by +/-

New York City
New York City data retrieved from local HTML file
01.    Team 1155	W:6	L:1	OPR:40.13	DPR:7.71	PMR:32.43
02.    Team 0694	W:5	L:1	OPR:48.45	DPR:16.43	PMR:32.02
03.    Team 2753	W:5	L:2	OPR:40.41	DPR:10.78	PMR:29.63
04.    Team 2344	W:6	L:1	OPR:37.71	DPR:11.32	PMR:26.39
05.    Team 0354	W:5	L:2	OPR:30.65	DPR:5.93	PMR:24.72
06.    Team 0371	W:5	L:2	OPR:24.19	DPR:0.05	PMR:24.14
07.    Team 0271	W:6	L:1	OPR:33.62	DPR:10.62	PMR:23.00
08.    Team 0056	W:6	L:1	OPR:34.82	DPR:13.50	PMR:21.32
09.    Team 1396	W:5	L:2	OPR:16.90	DPR:-1.69	PMR:18.59
10.    Team 3059	W:4	L:3	OPR:16.46	DPR:-0.88	PMR:17.34
11.    Team 0395	W:4	L:2	OPR:34.57	DPR:17.66	PMR:16.90
12.    Team 1302	W:5	L:2	OPR:19.31	DPR:3.41	PMR:15.90
13.    Team 0335	W:5	L:2	OPR:19.48	DPR:3.61	PMR:15.87
14.    Team 0555	W:4	L:3	OPR:31.73	DPR:16.13	PMR:15.60
15.    Team 0743	W:3	L:4	OPR:28.15	DPR:13.62	PMR:14.53
16.    Team 2265	W:3	L:4	OPR:13.59	DPR:0.03	PMR:13.56
17.    Team 2933	W:4	L:3	OPR:27.06	DPR:14.36	PMR:12.70
18.    Team 1807	W:6	L:1	OPR:23.10	DPR:10.70	PMR:12.40
19.    Team 0358	W:2	L:5	OPR:9.30	DPR:-3.01	PMR:12.31
20.    Team 0237	W:5	L:2	OPR:27.48	DPR:16.54	PMR:10.93
21.    Team 0270	W:4	L:3	OPR:16.16	DPR:6.91	PMR:9.25
22.    Team 0806	W:5	L:2	OPR:16.18	DPR:7.07	PMR:9.12
23.    Team 0041	W:3	L:4	OPR:29.08	DPR:20.49	PMR:8.58
24.    Team 1660	W:6	L:1	OPR:25.19	DPR:16.74	PMR:8.45
25.    Team 3017	W:4	L:3	OPR:9.20	DPR:1.35	PMR:7.85
26.    Team 1796	W:4	L:3	OPR:24.28	DPR:17.44	PMR:6.84
27.    Team 1257	W:3	L:4	OPR:18.69	DPR:12.12	PMR:6.57
28.    Team 1211	W:5	L:2	OPR:14.03	DPR:7.46	PMR:6.57
29.    Team 2681	W:5	L:2	OPR:14.22	DPR:8.45	PMR:5.77
30.    Team 0375	W:7	L:0	OPR:28.39	DPR:22.82	PMR:5.57
31.    Team 1881	W:3	L:4	OPR:14.22	DPR:10.59	PMR:3.62
32.    Team 0527	W:4	L:3	OPR:16.86	DPR:13.26	PMR:3.60
33.    Team 1237	W:2	L:4	OPR:12.71	DPR:9.29	PMR:3.42
34.    Team 1156	W:3	L:4	OPR:12.07	DPR:9.84	PMR:2.23
35.    Team 2285	W:3	L:3	OPR:23.47	DPR:21.81	PMR:1.66
36.    Team 2601	W:3	L:4	OPR:18.98	DPR:17.41	PMR:1.57
37.    Team 2070	W:3	L:4	OPR:7.30	DPR:6.57	PMR:0.73
38.    Team 0640	W:2	L:5	OPR:8.66	DPR:11.47	PMR:-2.81
39.    Team 1230	W:5	L:2	OPR:11.28	DPR:15.17	PMR:-3.89
40.    Team 0369	W:5	L:2	OPR:9.11	DPR:13.57	PMR:-4.45
41.    Team 2579	W:3	L:4	OPR:5.34	DPR:12.70	PMR:-7.36
42.    Team 1862	W:2	L:4	OPR:-0.63	DPR:6.93	PMR:-7.56
43.    Team 0333	W:2	L:5	OPR:7.73	DPR:16.31	PMR:-8.58
44.    Team 1563	W:1	L:6	OPR:22.48	DPR:31.97	PMR:-9.50
45.    Team 2577	W:3	L:4	OPR:20.45	DPR:30.12	PMR:-9.67
46.    Team 3111	W:2	L:4	OPR:12.75	DPR:22.44	PMR:-9.69
47.    Team 1989	W:3	L:4	OPR:2.76	DPR:13.64	PMR:-10.88
48.    Team 1600	W:3	L:4	OPR:11.61	DPR:23.95	PMR:-12.34
49.    Team 1520	W:1	L:6	OPR:2.07	DPR:14.43	PMR:-12.37
50.    Team 2205	W:2	L:5	OPR:11.63	DPR:24.03	PMR:-12.40
51.    Team 0329	W:4	L:3	OPR:-3.86	DPR:9.24	PMR:-13.10
52.    Team 2573	W:2	L:5	OPR:6.82	DPR:19.95	PMR:-13.13
53.    Team 1880	W:0	L:7	OPR:3.65	DPR:19.53	PMR:-15.88
54.    Team 0263	W:2	L:5	OPR:20.96	DPR:37.03	PMR:-16.07
55.    Team 3053	W:2	L:5	OPR:9.44	DPR:26.20	PMR:-16.76
56.    Team 1340	W:1	L:6	OPR:10.14	DPR:28.07	PMR:-17.93
57.    Team 2554	W:3	L:4	OPR:4.92	DPR:24.93	PMR:-20.00
58.    Team 0759	W:4	L:3	OPR:2.52	DPR:22.78	PMR:-20.26
59.    Team 0380	W:2	L:5	OPR:0.99	DPR:21.58	PMR:-20.59
60.    Team 0421	W:5	L:2	OPR:1.94	DPR:25.20	PMR:-23.26
61.    Team 3004	W:1	L:6	OPR:6.56	DPR:30.20	PMR:-23.65
62.    Team 1635	W:2	L:5	OPR:8.29	DPR:35.37	PMR:-27.08
63.    Team 2895	W:3	L:4	OPR:2.33	DPR:31.59	PMR:-29.26
64.    Team 3112	W:1	L:6	OPR:0.70	DPR:33.29	PMR:-32.59
65.    Team 0334	W:1	L:6	OPR:-6.96	DPR:32.80	PMR:-39.76
66.    Team 1698	W:0	L:7	OPR:-1.64	DPR:39.21	PMR:-40.86