On TBA what does DPR and CCWM stand for? I knew OPR but don’t know what these one mean or whats good or bad
2 Likes
DPR stands for Defensive Power Rating. Its the same as OPR except they calculate the metric using the OPPOSING alliance’s score. So, if you have a 10 point DPR, that means that on average, you’ll account for 10 points of the opponent’s score. Likewise, if all three members of an alliance have a 10 point DPR, then, on average, the opposing alliance is likely to score 30 points against this alliance. Obviously, DPR has no revellance if you don’t play defense. CCWM stands for “Calculated Contribution to the Winning Margin”. This too, is calculated the same as OPR except this time it is the difference between the winning and losing scores. A CCWM of 10 means that, on average, you generate 10 points more than opponents. An alliance of 3 teams all with CCWMs of 2 would, on average, win by a 6 point margin. OPR and DPR are both acronyms that came from the NBA. Don’t blame FIRST for these confusing labels. So, if you are looking for a way to calculate DPR or CCWM, just apply the method to getting OPR but use the opposing alliance score or the difference in scores in your “scoring matrix”.
Not necessarily. Very strong offense bots can also have low dprs especially in games with shared gamepieces due to their offensive actions hindering their opponents. However, DPR is not a very reliable metric in FRC for indicating actual defensive ability.
1 Like
Fixed that for you 
5 Likes
Sorry … the word is “relevance”. I’ve been ticketed by the grammar police. Anyways, OPR and DPR become irrelevant if scoring is not linear … and FIRST games are never linear, just less or more linear depending on the availability of game pieces and their scoring. OPR also becomes misleading whenever a team is not playing offense all the time (like to acquire a ranking point, help a teammate or play defense). DPR is then even less so because you CAN’T play defense all the time (during autonomous, opponent is in a ‘no touching’ zone, endgame, etc.). Yes, if one were looking for a defensive robot, DPR could be a metric to go by but still, a good team could get a low DPR just because the opponents decide to defend against their alliance every time. That DPR would have no significance. CCWM is a more relevant metric here because it covers both ends of the game.
I still don’t get how teams can have a ridiculously high OPR like 60+
4 Likes
Ok OPR is calculated way different than what I thought
1 Like
There are two ways to calculate OPR. The first is shown by “Thor5090”. Create a matrix of independent variables (a “1” for each team in the alliance) and a pencil matrix of the scores (the dependent variables). Then process a linear regression on the two matrices (in Excel, its the pencil matrix of dependent variables, and the other one of independent variables when using the LINEST function). The other way is to create a “Match Matrix” (a square symetric matrix of teams that played together) and the total of all the scores each team accumulated (Scoring Matrix). Then you want this match matrix times the OPR matrix to equal the scoring matrix. Take the Inverse of the Match Matrix to both sides and you end up with the OPR equals the inverse of the Match Matrix times the Scoring Matrix. I don’t know which method is more accurate but the 2nd method can only be applied once the inverse of the Match Matrix is possible (after enough matches has been played).
1 Like
I like using DPR to see who has the “worst” schedule. The higher the DPR, the worse your schedule was.
The reason you sometimes get outlandish OPRs is when not enough matches have been played and recorded to properly introduce a trend. Remember, the coefficients produced by the OPR calculation is to minimize the error when reproducing all the scores after the OPRs are calculated. Also, OPR is the CALCULATED contribution to the offense. It should be a guide when you generate your stack charts from scouting (which are the ACTUAL contributions to the offense). Sometimes, this is misleading. I’ll use the NBA for an example. Say you have a great point guard who dishes and breaks down defenses with his offensive skills. He only occasionally shoots the ball but he mostly dribbles past the defense and dishes a great pass to the power forward or center who makes the bucket. The power forward or center gets the points. But when you calculate OPR, the point guard gets a higher OPR than his actuals (points per game) because his contributions increases the score. This is why coaches in the NBA like OPR because they find the best combination of players. A team may have a higher OPR than say, actuals, because their play produces situations that the alliance can score more.
1 Like
OPR is fairly representative in most games where scoring is largely linear, such as this year. They generally conform with actual scouting data, with the large deviations happening either when a team has either an easy or difficult schedule or if its alliance partners consistently under or over perform in their matches (which happened to 1678 in 2013 Curie). In certain years the OPR hasn’t been particularly useful, as in 2017 with differential scores for gears as the alliance progressed through the rotors, and in 2018 when the time based scoring was different between the scale and switch, plus the exchange.
As to high OPRs, those are consistent with the actual points those teams are scoring in matches. If a team scores 5 cargo in auto plus taxiing (22 points), 20 in teleop (40) and a traverse climb (15), that’s 77 points.
1 Like
Didn’t even notice the misspelling, my brain autocorrected for you 
In FRC definitely use “actual” contribution in place of OPR if you have it.
The contribution, like in the NBA, is useful when you have a good knowledge of player positions. Like you may want the best in contribution for a point guard, but if you looked at it and the entire list was power forwards and you filled the roster with just those, you wouldn’t have a team that works well together. Roles in FRC are not often defined in any useful way like basketball.
1 Like
Richard_McCann, I totally agree. I think the game design people WENT OUT OF THEIR WAY to make OPR for the 2018 game (Power Up) to be as irrelevant as possible. The scoring for that was the least linear in the history of FIRST. It used to be, that Ed Law would provide a metric he felt best to represent the success of a team but I think he gave up that year. I should point out here, that there are lots of other metrics one can use to scale the competitors. There’s ELO and AWAR and Correlation just to name a few that one can use w/o having a squad of scouts. I’ve had to use these recently when my team, 1983 Skunk Works, had a severe reduction in resources (ie. scouts) and had to rely on general statistics to make important decisions.
1 Like
Ok now what does ELO and AWAR stand for? I’ve seen ELO before but never seen AWAR
Aggregate Wins Above Replacement – thread.
So AWAR is a more long term look at a team. I know before this year our team would’ve been down around 0 or 1. We made a tremendous leap this year and hope we can keep it up but it’s hard to say considering we just graduated 50% of our team and with covid we took a big hit on recruiting. We are starting a couple FLL and a FTC teams up so hopefully we can make another leap in a few years. But for now we’re hoping our state championship win this year will bring in more high school students for now so we can continue to grow.
1 Like
And for ELO, I first call Wikipedia for an overview…
…And then I @Caleb_Sykes for the FRC applications.
1 Like