Chief Delphi - 2014 Curie Division

Chief Delphi (http://www.chiefdelphi.com/forums/index.php)

- General Forum (http://www.chiefdelphi.com/forums/forumdisplay.php?f=16)

- - 2014 Curie Division (http://www.chiefdelphi.com/forums/showthread.php?t=128842)

Re: 2014 Curie Division

Quote:

Originally Posted by Gary Dillard (Post 1375477)

Maybe no penalties (especially no stupid ones) should be up there somewhere?

Look no further than the human player. If they are acting the fool and piling up penalties like a busnessman does airline miles write them off ASAP!

Re: 2014 Curie Division

Quote:

Originally Posted by Koko Ed (Post 1375478)

Look no further than the human player. If they are acting the fool and piling up penalties like a busnessman does airline miles write them off ASAP!

Yes to this. Human players should definitely not be getting penalties at this point much less drive teams.

Re: 2014 Curie Division

I've made a pretty crude attempt to factor out event "quality" from OPRs. The first stab at this simply adds the event's median (or mean) OPR to the OPR for the team. Teams that go to better events get more points.

There is probably a better way to normalize OPRs across events, but this seems okay.

Median adjusted

Code:

469, 173.87

1718, 172.53

254, 165.16

118, 156.09

4522, 155.41

4334, 153.48

68, 147.97

180, 145.42

125, 144.78

359, 141.41

4362, 139.88

230, 139.83

447, 134.78

842, 132.70

3476, 130.78

4055, 129.77

1241, 129.24

2013, 126.08

177, 126.05

2451, 125.47

987, 120.14

3015, 115.58

3812, 115.58

74, 115.37

3990, 113.58

2016, 112.03

624, 111.82

573, 109.24

1323, 108.89

1595, 107.49

1629, 107.33

11, 106.22

222, 106.17

135, 104.99

1732, 103.57

75, 101.77

1714, 100.17

2619, 98.72

2928, 95.88

5172, 95.59

4819, 94.64

2468, 93.10

236, 91.92

1676, 91.58

4915, 91.04

5191, 90.55

2648, 89.38

294, 88.08

2443, 87.46

3161, 85.55

772, 85.44

1501, 85.26

2848, 84.20

2478, 82.12

3386, 80.39

1723, 73.98

3301, 73.36

3660, 73.23

3794, 72.06

2073, 70.84

3932, 70.33

3230, 70.31

3478, 69.24

1311, 68.90

418, 68.86

5024, 68.67

4125, 67.45

1902, 65.97

3562, 65.60

191, 63.47

5125, 62.99

865, 62.59

2403, 60.94

3042, 58.18

4159, 57.95

5093, 57.02

5179, 56.96

3692, 56.30

4486, 55.74

4161, 55.72

28, 55.65

5002, 55.58

2543, 52.49

5076, 51.66

4901, 49.43

1937, 44.94

4171, 44.38

540, 44.33

2080, 43.58

4935, 38.53

4969, 36.91

2169, 35.94

2227, 32.31

5326, 30.95

3843, 30.65

1287, 29.59

5297, 29.06

5036, 28.36

4977, 25.62

1884, 14.18

Mean adjusted

Code:

1718, 187.63

469, 178.86

254, 169.93

4522, 162.90

4334, 161.09

118, 161.08

68, 152.96

359, 149.26

180, 146.12

4362, 144.87

125, 143.80

842, 141.23

230, 138.84

1241, 136.33

2451, 134.26

2013, 133.70

3476, 131.36

447, 130.91

4055, 128.79

177, 128.27

987, 127.52

74, 120.36

3990, 118.12

3812, 117.91

3015, 117.03

624, 115.52

1323, 114.47

573, 114.23

1629, 114.11

1732, 112.37

1595, 109.82

135, 108.28

2016, 108.16

11, 107.80

222, 107.75

2619, 103.71

75, 103.35

1714, 102.08

5172, 101.83

4819, 99.63

2468, 98.80

2928, 98.21

4915, 97.16

2443, 95.32

236, 94.14

1676, 93.16

294, 92.62

3161, 90.32

2648, 88.40

3386, 87.02

1501, 86.61

2848, 85.87

5191, 84.14

2478, 82.70

772, 81.38

1723, 77.94

1311, 77.83

3660, 77.18

2073, 76.18

5024, 75.77

3794, 75.33

418, 73.85

3301, 73.33

3478, 72.52

5125, 71.79

3230, 71.38

3932, 71.03

4125, 71.03

3562, 70.94

865, 67.36

1902, 66.66

191, 64.91

4486, 63.57

5002, 62.86

5093, 62.01

3042, 61.77

2403, 61.52

5179, 61.50

4159, 60.46

2543, 59.50

3692, 58.21

28, 57.99

4161, 57.91

4901, 53.40

4171, 49.96

540, 49.69

2080, 47.54

1937, 46.59

5076, 45.24

4969, 42.27

4935, 42.16

2169, 41.91

3843, 37.93

5297, 36.91

5036, 34.99

1287, 34.64

2227, 34.15

5326, 32.30

4977, 28.90

1884, 19.84

I've attached a file that contains all event OPR stats.

Re: 2014 Curie Division

Quote:

Originally Posted by Andrew Schreiber (Post 1375422)

Tactical flexibility. Never underestimate the value of forcing your opponent to think on their feet.

So much this! Changing up your strategy between or even during a match can throw in a mind game to the opponents that could swing a match.

Re: 2014 Curie Division

"I've made a pretty crude attempt to factor out event "quality" from OPRs. The first stab at this simply adds the event's median (or mean) OPR to the OPR for the team. Teams that go to better events get more points."

Or... teams that competed at events with less defense played get more points?

Re: 2014 Curie Division

Quote:

Originally Posted by JB987 (Post 1375627)

Good catch. I only thought about the positive case.

This is correct, assuming less defense equals more points, which it doesn't always. It's hard to statistically isolate events that are defense heavy from events that don't have a lot of offensive power.

As I said this is a super crude way of doing this, and it may be rewarding points for the wrong things. I think the teams who went to DCMP events are unfairly rewarded as those events were made up of a selected pool of teams.

Anyway, all the raw data is there.

Re: 2014 Curie Division

Thanks for crunching the numbers, Tom. Any data in correct context can serve a useful purpose...

Re: 2014 Curie Division

Anyone on Curie want to join in on a scouting alliance between teams on different fields? (Details can be found here: http://www.chiefdelphi.com/forums/sh...+championships)

We've got teams on the other 3 fields interested, but could use some representation from Curie. Thanks!

Re: 2014 Curie Division

Quote:

Originally Posted by Tom Bottiglieri (Post 1375563)

The first stab at this simply adds the event's median (or mean) OPR to the OPR for the team. Teams that go to better events get more points.

If I'm reading this right, wouldn't this exacerbate the issue? You're adding the high average of a strong event to an already inflated OPR.

Rather, you'd want to measure how strong the event was and correct OPR with that. Simplest way would be to take the difference of the event's average OPR and the world average OPR, and subtract that from the team's OPR at that event. This would lower OPRs at stronger-than-average events and raise them at weaker ones.

Something like:

OPR - (Event Avg - World Avg)

Re: 2014 Curie Division

There are really two opposite factors that affect event to event OPR comparison. One is if an event does not have many teams that are able to possess the ball, everyone's OPR will be lower. This seems to be what George is interested in. The other is that at an event with a lot of good teams, there's only so many points you can score due to cycle time. A good scoring robot and two decent passers will likely score similar to 3 all around good robots. This reduces the OPR of the all around good robots at an event. This seems to be what Tom was addressing.

Re: 2014 Curie Division

Tigertrons YouTube page has videos from Mt. Olive, and I am making it my mission to get the MAR Champs matches posted by Monday. (edit at home, upload at work ;) ) Unfortunately, we don't have video from Chestnut Hill, sorry.:(

Re: 2014 Curie Division

Quote:

Originally Posted by Joe Ross (Post 1375873)

Among the other problems with OPR this year is the assumption that teams are trying to maximize their overall score. Many teams at times sacrificed potential truss and high goal points to maximize assist points rather than total score.

Re: 2014 Curie Division

Quote:

Originally Posted by cbudrecki (Post 1375898)

FYI, Team 1676 has all of the full-field match videos from Mt. Olive, Clifton, and Bridgewater-Raritan posted on their YouTube Channel. They are also working on getting all of their MAR Champs footage up as well, which will hopefully all be posted by tomorrow.

Re: 2014 Curie Division

Quote:

Originally Posted by George1902 (Post 1375696)

I am trying to award teams who played at harder events and had their OPR lowered because of the defense they played against. This method does exactly that and brings up teams who played at high caliber events closer to the top. It doesn't award teams who had high OPRs at their event where no one else could score as much. I'm not sure how good of a metric it is other than kind of being a 2nd order sort on who played well at "good" events.