View Full Version : Week2 cumulative Twitter stats & OPRs
Stats and OPRs based on Twitter Qual match data as of Sat 3-8-14 20:05:49 ET.
The usual Twitter data caveats apply.
NOTE: Look for updated data in more recent posts in this thread
Stats and OPRs based on Twitter Qual match data as of Sun 3-9-14 14:54:20 ET.
The usual Twitter data caveats apply: Twitter data is incomplete and may contain errors and redundant data. Twitter data is not official. However, it contains information not available in the official Match Results and Team Standings data, such as foul points for each match.
Explanation of "Quartile" for those not familiar with the term:Example:
win final score:
102 average (mean)
420 max
131 Quartile3
95 Quartile2
67 Quartile1
100% of the winning scores were less than or equal to 420
75% of the winning scores were less than or equal to 131
50% of the winning scores were less than or equal to 95
25% of the winning scores were less than or equal to 67
25% of the winning scores were between 131 and 420
50% of the winning scores were between 95 and 420
75% of the winning scores were between 67 and 420
Elim Stats based on Twitter Elim match data as of Sun 3-9-14 14 20:03:13 ET.
See previous post for Qual stats and OPRs.
The usual Twitter data caveats* apply.
* Twitter data is incomplete and may contain errors and redundant data. Twitter data is not official. However, it contains information not available in the official Match Results and Team Standings data, such as foul points for each match.
For comparison, here's the combined final score OPR using the official FRC MatchResults data (not Twitter) for Week1 (http://www.chiefdelphi.com/forums/showpost.php?p=1353206&postcount=1)+Week2 (http://www.chiefdelphi.com/forums/showpost.php?p=1354665&postcount=3).
For comparison, here's the combined final score OPR using the official FRC MatchResults data (not Twitter) for Week1 (http://www.chiefdelphi.com/forums/showpost.php?p=1353206&postcount=1)+Week2 (http://www.chiefdelphi.com/forums/showpost.php?p=1354665&postcount=3).
It appears team 2386 is missing from your list?
Edit: It looks like you're missing the entire Alamo regional
It appears team 2386 is missing from your list?
Edit: It looks like you're missing the entire Alamo regional
Good catch. A file was missing from the folder when I ran the script. I'll post a correction shortly.
Edit:
Update posted (http://www.chiefdelphi.com/forums/showpost.php?p=1356229&postcount=4).
According to the log file created by my script which processed the MatchResults data there were a total of 2124 qual matches and 1112 teams at 26 events in weeks 1 & 2.
Compare that to the Twitter data for weeks 1 & 2: only 1822 qual matches and 994 teams.
Here's a spreadsheet of the "OPR" of all the data columns for all the Qual matches in the Twitter data as of Sun Mar9 20:03:13.
"OPR" (least squares fit) was computed for each team not only for their alliance scores, but also for their opponent's alliance scores.
So for example:
column B minus column F gives the CCWM
column B minus column C gives the "unpenalized score"
(B-C) - (F-G) gives the "unpenalized CCWM"
You can add whatever other computations might interest you.
The usual Twitter data caveats apply.
See attached plot of A∙x-b alliance unpenalized score residuals vs percent of alliance scores, taken from Twitter data dated Sun Mar9 20:03:13 ET containing 1822 Qual Matches with 944 Teams.
A∙x~b, where:
A is the binary design matrix of the alliances;
x is the Team unpenalized OPR* score vector;
A∙x is the alliance unpenalized OPR* score vector;
b is the alliance unpenalized actual score vector;
A∙x-b is the residual vector (difference between A∙x and b)
From the graph you can see that 65% of alliance scores are within ±20 unpenalized points of the OPR. 37% are within ±10 points.
*L2 norm (least squares) solution of A∙x~b
MrRiedemanJACC
12-03-2014, 19:14
Per our previous conversation, I am assuming most teams from Gull Lake aren't represented very well in this because of the lack of twitter data? (I'm thinking 2054 should be pretty far up there on OPR)
Per our previous conversation, I am assuming most teams from Gull Lake aren't represented very well in this because of the lack of twitter data? (I'm thinking 2054 should be pretty far up there on OPR)
What does the pronoun "this" refer to? Your post is linked to Post8 (http://www.chiefdelphi.com/forums/showpost.php?p=1358276&postcount=8), which does not show individual team OPRs at all.
??
Robotmmm
12-03-2014, 20:18
Is there a way to denote that teams that attended week 1 Mount Olive District had matches played that were 20 seconds short?
Is there a way to denote that teams that attended week 1 Mount Olive District had matches played that were 20 seconds short?
I think you just did :)
On the other hand, if what you meant was "somehow factor that into the statistical calculations to account for the anomaly", I'm not sure it would be possible to obtain a consensus agreement among CDers concerning a statistically valid way to do that.
Robotmmm
12-03-2014, 20:45
That is why I said denote. There is no easy way to change the stats. Just talking about an asterisk type scenario.
I'm sure the answer to this has been posted and discussed a million times, and I apologize... but is Hatboro-Horsham missing?
I'm sure the answer to this has been posted and discussed a million times, and I apologize... but is Hatboro-Horsham missing?
Sorry. There is no Twitter for Hatboro-Horsham (PAHAT). This usually means there were internet problems at the venue.
Many folks would like to see the Twitter data saved on the local FMS computer* until it can be archived and made accessible to teams doing scouting and for students interested in statistics who would like to analyze the data.
Ask around and see if there's anyone on your team who has any influence within FIRST. Maybe he or she can get the attention of someone at FIRST who has the interest and authority to make a constructive change.
* instead of being deleted at the end of the event
MrRiedemanJACC
12-03-2014, 21:30
Ahh yes sorry. I was referring to the excel file in your post #7 in this thread.
Ahh yes sorry. I was referring to the excel file in your post #7 in this thread.
There are only 16 Qual matches in the Twitter data for Gull Lake.
But that results in OPR scores which are lower than they would otherwise be only if those 16 matches were lower-scoring matches compared to the ones which are not in Twitter data.
Of course, if GullLake turns out to be a relatively high-scoring event compared to future events that teams from Gull Lake participate in, then the higher-scoring matches missing from the Twitter data will affect the OPR going forward for those teams.
In the case of 2054, their OPR from the incomplete Twitter data appears to be higher than from the (complete) MatchResults data.
http://www.chiefdelphi.com/forum/showthread.php?p=1356165
vBulletin® v3.6.4, Copyright ©2000-2017, Jelsoft Enterprises Ltd.