So here is how the FRCTop25 will work: All voting will be done now through Monday. We will split and recompile votes for each champs and present where teams fell on Tuesday the 18th for Houston Champs and Monday the 24th.
After both championships are complete we will have a 2017 wrap up show where we will present the FRC Top 25 composite poll. Your votes this week will be then combined with the expert poll and we will release the results of both of these on May 2nd.
**The experts poll will be from the amazing Looking Forward group. ** After the May 2nd show Looking Forward will release their entire analysis of their Top 25. We are excited for this partnership effort.
Check out the results live on at
Post below if you want to make your pitch on a team that should be in the FRC Top 25.
So, I love this list, mainly because we’re on it. I’m not really sure what it means, though. I searched through Chief Delphi to try to better understand what ELO is. All the posts assumed people already knew what it was. I found this to explain ELO online:
"The better your opponents are, the more points you get if you beat them.
The worse they are, you get less if you beat them."
This is pretty simple but it makes sense. Caleb, could you explain in more detail what goes into ELO for Robotics? Thanks.
Since I’ve been getting this question quite a bit recently, I think I’ll make this a summary post that I can link to whenever I get this question.
Elo ratings are a common metric used in sports and online gaming. The rough idea is that the Elo model creates a prediction for each match, and after each match, teams which exceeded their expectations get an Elo boost and teams which underperformed on their expectations see their Elo drop. I have adapted Elo for FRC by doing the following:
I use the winning margin for each match as the measurement to build on. The winning margin is standardized between years by scaling according to the standard deviation of each year’s scores.
I account for both qual and playoff matches, but weight playoff matches less heavily than qual matches.
After each season, I revert every team’s rating a bit towards 1500.
Rookies and new veterans begin with a rating of 1350.
A full description of my model can be found in the “Instructions and FAQ” sheet of the FRC Elo 2008-2016 book.
I would do it the other way around. How come you weight quals heavier than playoffs? Playoffs you have absolute control over your alliances and your path to victory. Quals sometimes luck plays a big factor.
Poll is now closed. 374 votes for this poll. In comparison we had 209 last year. Thank you to everyone who voted. We are looking forward to revealing who took the FRCTop25 for the Houston Champs.
Every modification I made to Elo was done to maximize the model’s predictive power. Giving playoff matches as much weight as quals matches reduces the predictive performance both for playoff matches in that series as well as qual matches at future events.
My best guess as to why this is the case is that, since we don’t get the random team assignment like we do in quals, there is not nearly so much useful information in playoffs as there is in quals. Once you build your alliance for playoffs, there is no way to separate out individual teams’ contributions, so it makes some sense that the ratings might be slower to respond to strong performances.
Fair point if you’re targeting predictive power. My own Elo system I use purely for measuring performance rather than looking towards the future, so I don’t weigh quals heavier than playoffs. Thanks for the explanation.