Request to change TBP system

[FTC12676]

current TBP system is not perfect and is being manipulated by some teams. I am creating this thread for discussion and suggestions for what can replace current ranking system. I personally prefer TBP should be your own alliance score and not dependent on opposing alliance. This way team will be ranked based on their own performance.

[Mentor 777]

I agree! I found it quite interesting watching the Detroit Championship. Gluten Free maxed out their score each time and broke the world record several times over the course of their 9 rounds. However, with TBP, they ranked 5th.

In contrast, Lanbros and Kraken Pinion both scored for their opponents to increase their TBP. That strategy paid off in the rankings since they came out as number one and two. However, I believe it would have been more interesting if the TBP system was based on your own alliance score. That would encourage teams to go for the best round they could do. While Gluten Free's record would have been very difficult to beat, these other top teams could have given them a run for the money if they hadn't been encouraged to spend time increasing their opponents' score to increase the ranking.

The theory that a win against an alliance scoring 400 points should be worth more than a win against a team scoring 40 points would still hold true if the tie breaker points were based on your own alliance score.

[FLARE]

I think the current system was especially skewed at Detroit (and probably Houston) because of the lottery teams as well.

We knew we would have a difficult time as soon as the match schedule came out. But it became very clear afterwards, when looking at the final numbers on ftcstats.org. Of the top 20 ranked teams, we were paired with 0 of them and opposed 5 of them. Our qualifier ranking was #27, but we are #9 for our division based on OPR. Random luck of who you get paired with and against should not play such a large role in the final outcome/rankings. Though I wish each team's own scoring was tracked for rankings, OPR or alliance total would seem to be much better than the current TBP system.

[BSV]

The bonus ranking point system used in FRC is quite good. It rewards teams who win the match, but it has an extra level or reward for teams that complete a specific task either individually or as a team. For instance, if you could fill a rocket in Deep Space by yourself (or with a teammate, but not required) then you got a bonus RP. The other bonus RP was for climbing at the end and required teamwork, but if you built a high-climber you could probably count on one of your teammate to get the low climb even if they couldn't otherwise score.

The secondary tiebreakers go to points scored doing different tasks rather than an arbitrary thing that the other alliance did.

It's definitely a model worth considering.

[Alec]

Four alliances in elimination rounds is grossly inadequate for an 80 team division. There should be eight alliances competing in elimination rounds per 80 team division. Also, the finals should be best three out of five; not two out of three. Fixing TBP (if it can be fixed) won't fix the underlying problem of having only four alliances compete in elimination rounds.

[norinftc4uLori]

To keep it simple - Would using the High Score as a tie-breaker work?

[Cheer4FTC]

The issue with using only High Score is that 2 undefeated teams that paired on an alliance could have the same high score. It also means that anyone paired with a super-star team gets a huge break in that ranking (e.g., all of Gluten Free's partners at Detroit Worlds had a good High Score largely because they were partnered with Gluten Free).

I suggest replacing TBP with the average of a team's top 1/2 qualification round scores (so with 5 or 6 qual matches at an event, take the average of their best 3 scores, and with 9 or 10 qual matches at an event, take the average of their best 5 scores).

There are a bunch of more complicated or hard-to-explain options (OPR-related, or a team's 2nd best score, or the average of 2nd and 3rd best scores, etc.), but I think the FTC HQ folks have asked for proposals that are easy-to-explain.

[Mentor 777]

I suggest replacing TBP with the average of a team's top 1/2 qualification round scores (so with 5 or 6 qual matches at an event, take the average of their best 3 scores, and with 9 or 10 qual matches at an event, take the average of their best 5 scores).

I like that suggestion! Simply using an alliance's own score is better than using the opposing alliance's score. However, your suggestion also takes into account the many things that can go wrong in a match, even with an excellent robot! Connection issues occur very frequently at competitions, even at the World Championship level. By using the top half of the scores, a team would not be penalized so harshly for an issue that is mostly out of their control. While there are things that can be done to mitigate the connection issues, even the top teams ran into this problem.

[NEOFTC]

I strongly agree that there should be 8 alliances on each field at Worlds. Too many teams did not compete at all on Saturday. As far as tiebreakers, I suggest cumulative autonomous points. I would also like to see the rankings displayed just like they do in FRC. Divide by the number of matches. That way you don't have to wait until the end of a round to see how things really shake out.

[Cheer4FTC]

I strongly agree that there should be 8 alliances on each field at Worlds.

Or maybe 4 divisions of 40 teams each instead of 2 divisions of 80 teams each, with each division still having 4 alliances competing for the division championship?

Then the 4 division winners play each other round robin with the top 2 after the round robin round playing best-of-three for the championship.

FRC does this right now, except they have 6 divisions instead of 4.

[Or maybe even a simple double-elimination tournament for the 4 division winners?]

[FTC12090]

I dislike the current system as well. Incentivizing scoring for your opponent... I mean how much more insulting can you get? Not to mention the incentive to intentionally lose to hurt your opponent.

[FLARE]

I dislike the current system as well. Incentivizing scoring for your opponent... .

With the current scoring system, this is a strategy that makes sense, unfortunate though it may be. The 2 years with the beacons, scoring for opponents was actually a highly-used strategy and did not seem frowned upon at the time.

This season we did this at our first qualifier and were penalized for egregious behavior. After that is was clarified as legal on the forums, but our team was afraid to do it after that because of how it might appear to other teams and refs. Then of course we saw it in Detroit, where due to scoring for the opponents, the top 5 teams ended up in an order that seems quite skewed compared to actual robot performance.

Though I strongly agree that the TBP system should change, if it does not then I think it should be made very clear from the season beginning that scoring for the opponent is an acceptable strategy.

[JoAnn]

Lots of ideas in this thread! We've been looking at TBP, and have some criteria for what we'd consider.

1. Must be easy to explain to everyone.
2. Must reward teams for winning more difficult matches.
3. Must not require a significant change to the way referees score matches.
4. Must be fair.

These are a lot, I know, and everyone might not agree that these criteria are important. Remember that strategy plays into this. Some seasons we allow teams to score for their opponents, and some years we don't (depends on the game). A team with an amazing robot, and world record scores still needs to pay attention to the best strategy for their team at each event, and even for each match.

JoAnn

[RollerCoaster45]

Switching over to something like they have in FRC would be a good way to meet all 4 of these criteria. 2 points per win, 1 point for a tie, zero per loss and you can earn up to two additional ranking points for completing some on-field task. Then using TBP would be fine, because it would be possible to make it very difficult to earn all 4 ranking points so you would be rewarded for winning hard matches, but also doing well yourself.

[Cheer4FTC]

2. Must reward teams for winning more difficult matches.

The main problem with this criterion is that the simplest way to measure match difficulty is by looking at opponent score. But if you're going to look at opponent score, then that will motivate teams to score for their opponents and will motivate teams to take non-GP actions like asking other teams to throw a match so one of their rivals' matches doesn't look as difficult.

You could look at more complicated measures of opponent strength (like the average final rank of opponents, etc.), but this could again possibly motivate teams that are not likely to be alliance captains near the end of quals to throw matches. [Unless FTC adopts scoring more like the FRC district model where qual rank overall gives teams more district points so a team that can't be an alliance captain is still motivated to rank as highly as possible at the end of quals.]

If we had a way to measure this #2 criterion in a fair and reliable way that could not be gamed, it would seem like a reasonable criterion. But the "easy to explain" ways of measuring winning difficult matches seem to be easy to unfairly game.

Can anybody think up a way to accomplish both criterion #1 and criterion #2 in a way that doesn't provide a method and motivation to game the system??

The other problem with criterion #2 is that whether a team has difficult or easy matches is not up to the team! A simple example is GlutenFree at Detroit Worlds. They likely could have defeated nearly every opposing alliance with a normal or even weak partner, but their match schedule didn't have a lot of difficult matches for them to play. Should they be punished for their relatively easy match schedule? Ranking them poorly because of things out of their control would seem to violate criterion #4.

RollerCoaster45 I like the FRC-type ranking too, but I don't see how that measures how hard the match was to win. It's just another way to measure how good an alliance is at scoring points or achieving a particular scoring achievement, which could just as easily be measured by looking at a team's average score, or auto score, or endgame score, etc. Thoughts?

Bottom line: TBP is primarily needed to rank a set of undefeated teams after quals to determine alliance captains. To me, measuring their capability of scoring in some way seems more fair and sensible than measuring who they happened to play against (especially given the motivation for non-GP gaming of this latter measurement).

JoAnn I suggest that at least you add an additional criterion #5, something like "must motivate teams to win and score as many points as possible for their alliance in all matches, and not provide motivation for teams to take actions counter to these goals and to overall Gracious Professionalism and fair play."