Monday, November 17, 2014

Team Ratings, Rankings, and Seedings

In most of the world, table tennis is a club sport and a team sport. The two go together, as club members form teams and play against teams from other clubs. The team aspect is not that strong in the US . . . yet. But it grows slowly every year. To me, club team activity is one of the measures we can use to judge the success of the sport in our country.

We have entered the season for team competition. College teams in the NCTTA League have been competing since the first divisional meet in October, and will continue fall seasonal play through this weekend. They'll pick up again in January and February to complete the regular season, followed by regional qualifiers and then the Championships in April.

For those of us who don't have the opportunity to play in a team league, the next best thing is a team tournament, and you have three team tournaments to choose from this Thanksgiving weekend. Personally, i consider team competition to be much more fun than individual tournaments, and i strongly encourage everyone to enter one of these tournaments.

Butterfly Thanksgiving Team Championships - Hobart, IN (Chicago Area)

North American Teams - National Harbor, MD (Washington DC Area)

West Coast Thanksgiving Teams Open - Fremont, CA (San Francisco Bay Area)

But i digress. This post was supposed to be about ratings, rankings, and seedings.

Tournament players generally understand individual ratings and rankings. A player rated 1900 has demonstrated better match results than a player rated 1800. Therefore if we were to rank the players, the player ranked 1900 would be #1, and the player ranked 1800 would be ranked #2. Rankings are the numerical order of the players involved.

Seedings are similar to rankings. Rank is the order, and seed is where they will go in the draw. It is a subtle difference, and most people don't need to worry about the difference. Basically, the player ranked #1 will be seeded #1 and have the "best" draw; the player ranked #2 will be seeded #2.

How about doubles? We don't play a lot of doubles in the US, and we don't have many established doubles pairs who always play together. If we did, we would be able to use the doubles match history of those specific pairs to establish a national doubles ranking order. But since we don't, we are left with using singles ratings to determine doubles rankings. It's not ideal, but we get by - we add the individual ratings of each doubles pair together to establish a "doubles rating", and use that to rank pairs.

Team competition is quite a different story. Again, we don't have many established teams that always play together. So looking at what we do with doubles, the obvious practice would be to take the sum of ratings to find a "team rating".

This is quite easy to do, but unfortunately a numerical sum is not a very good reflection of actual team strength.

Take these two fictional teams. They'll play in the old Swaythling Cup format: 3 players per team, playing the best 5 out of 9 matches. The teams are composed of players with the following ratings:

  • Team #1: 2400, 1800, 1800
  • Team #2: 2000, 2000, 2000

If you use the concept of a "team rating", these two particular teams will be tied at 6000, which would mean that they are equal in strength. But if you play out the team contest, team #2 will win by a score of 5-3 almost every time. They will almost always lose 3 matches to the 2400 player, and almost always win 5 matches over the 1800 players. The reality is, team #2 is clearly stronger than team #1, and in many situations, using the sum of individual player ratings will not give good seedings.

A better alternative can be found by looking at how the ITTF ranks teams. You can read about their ranking methodology on the ITTF web site. In summary, they take every national team in the world, and run a simulated giant round robin between all teams. The simulation assumes that a higher ranked player will always beat a lower ranked player. Each team is then ranked according to the results of the simulation, as if the teams had actually played out a giant round robin.

Using this method of ranking gives superior results when seeding teams. A weak team sometimes brings in one particularly strong player to try to pump up the team's seeding, and if seeding is based simply on the sum of individual ratings, the effect of that strong player will be exaggerated. These issues and others are solved when seeding is performed by simulation.

Here's an example of team seeding from this year's America's Team Championship, an annual event held over Memorial Day Weekend in Rockford, IL:

These were the teams seeded into the fourth division. If the sum of player ratings had been used, the TriniAmericans would have been seeded #24 - higher than all of these teams, and actually into the next higher division. But when the matches are played out, they should fall into the middle of the pack in this division. On the other side, the Fab Four's total rating is not especially high, but against this field they are likely to win most of the contests, albeit by scores of 5-4 and 5-3.

If you play in a team tournament that i'm directing or refereeing, it will probably seed teams using this method of seeding. I've been using it for several years, and would recommend that other team tournaments adopt it.

Since we're relying on a computer simulation to rank teams, conceivably we could go one step further and consider the win probability for each individual match. For example a 2400 player has an almost 100% chance of winning against a 2000 player, and it's reasonable to assume he'll give his team one point for that match. But when someone rated 1830 plays an opponent rated 1825, the match is almost a tossup. In the simulation, the 1830 player could be awarded 0.51 points, and the 1825 player could get 0.49 points to reflect how close the match is. It's not exactly simple to do this in a spreadsheet, but it's not terribly difficult either. You just need to figure out each player's chance of winning and go from there.

Normally i don't seed teams using win probability, because it's more difficult for players to follow. However this method results in even better seeding, and in the NCTTA League we will use this method to seed teams at the College Table Tennis Championships. If you're curious, you can read more about this method on the NCTTA web site.

Not table tennis.

I don't live in New York and have yet to see the Top Spin documentary film. But i did see a lovely movie about musicians entitled "Begin Again". If you're a sentimental sap like me, maybe you'll enjoy it. 

In Memoriam.

My friend Jock Oubichon moved on last weekend, at the young age of 47. He may not have been well known in the world of table tennis, as he didn't travel much and didn't draw attention to himself, but he made a mark on everyone who knew him. He was a person who had willed himself to become a positive force - an intense yet gentle person who really loved helping people. As a table tennis player and coach, he had a great eye for human mechanics, and he loved coaching though i don't know if he ever charged anyone for his help. Several years ago i had the honor of nominating him as USATT Volunteer Coach of the Year, and if i recall correctly the USOC named him a finalist among all sports.

I didn't see much of him in recent years due to geography, but i always looked forward to meeting him at the Nationals in Las Vegas - me working at the tournament, him meeting with the sponsors of his club's junior program. I will miss him.

Ron Spencer's message about him can be read at the San Diego TTA's web site.


  1. I blogged a bit about this. See second item.

  2. Very Cool Blog Kagin! I Didn't Know you Had One! Thanks For Sharing With Us!

    Mike Mezyan

  3. Need to get more practice in order to get better ranking.
    table tennis spot