View Single Post
Old 12-14-2012, 07:25 AM   #47
Join Date: Jan 2010
Posts: 1,249

Originally Posted by goober View Post
I agree. There is nothing terrible about the algo currently used. It only is an issue with those trying to manipulate the system. So rather than trying to fix the algo forget about it- you can't fix any system if people are willing to lose matches on purpose or keep scores close. A couple years back one of the teams that went to nationals out of our region won all their matches 3-2 when they easily could have won them 5-0. When you are that much better than everybody it is easy to control the outcome. Their final match of the season when they already clenched playoffs they lost 1-4 to the last place team (one guy didn't get the "they are suppose to lose" memo) .

So what should be done is fix the incentive. These 3 things would eliminate 90% of problems over the long run while allowing those desperate for sectionals and nationals to have their shot.

1) No self rated players in playoffs.
This may penalize only legitimate self-rates that would have advanced to playoffs as someone willing to game the system is willing to throw a year and will just play that first year and sandbag to get a C rating then they are golden. I prefer to tighten the "allowance for natural improvement" as particularly at 3.5 and below it seems far too large.

That said, this rule probably wouldn't affect very many, but I'm guessing the USTA resists because they think it would deter players from joining if they know from the start they can't go to playoffs their first year.

Originally Posted by goober View Post
2) Any player in nationals bumped and benchmarked for 3 years, sectionals 2 years. It is kind of ridiculous a team that finishes top 5 in nationals only has 30% of their team bumped.
This seems a bit extreme, and I'm not sure the point of benchmarking the player for multiple years.

The only point of being a B is to indicate you played post-season against players outside your sub-flight/flight/league and thus can serve as a reference for calculations to try to normalize ratings for different areas. Making a player a B in years they didn't go to playoffs doesn't make sense.

On the bump for multiple years, this can be problematic. You have situations where a good player, A, at their level is on a team with some great players and they go to nationals. Player A usually plays court 3 doubles and wins his fair share, but isn't in the top-6 doubles players. At nationals, he gets to play a match or two after they've either clinched the semis or lose in the semis. This player A, at best a very good player at his level and certainly not low-end at the next higher level is now bumped up for 3 years? That is a huge penalty.

And this is probably pretty common with many teams that do well having a core group of players that carry the team but then have some good but not great players to fill out the roster and play in the less important matches. I was on a team that went to nationals in 2011 and had this exact situation. About half our team got bumped and the ones that didn't probably shouldn't have been. Now a few of those did continue to improve and got bumped this year, but a few still haven't. Locking those in at the higher level for 3 years is too much.

Originally Posted by goober View Post
3). No ESR bump downs only bump ups. Huge loop hole exploited so players that get bumped only have to remain at their new level for 6 months before going back down to their old level.
Early start leagues/ratings are difficult to deal with and this perhaps has some merit.

But the core issue is that the algorithm doesn't deal with sandbagging at all. There are a couple ways to deal with this.

First, there are sections where leagues that don't advance to any nationals are still counted for rating purposes. This allows players to play all out trying to get to nationals but then sandbag in the leagues that "don't count" to keep their rating down. One approach to deal with this is to consistently across sections only count leagues that have nationals. The challenge here is you end up with less data to calculate ratings from, some players may play only 1 or 2 matches in these leagues but many more in the other leagues. Also, the real sandbaggers will still find a way like the example you cited.

Second, one could adopt something like what golf does with its handicapping system where only your best results count and poor results are thrown out. This gets more complicated, especially with tennis where you are playing an opponent rather than just playing the course, but is a good approach to dealing with throwing matches to manipulate ones rating.
schmke is offline   Reply With Quote