I set myself the task today to begin a statistical analysis for Ladles. It should be possible, with the frequency of competition nowadays and the standardization of servers and settings, to keep a running standings based on all matches teams play against each other: Ladles, leagues, wars, etcetera. It also ought to be possible to track a player’s statistics. Win-Loss percentages do not tell the full story, though. If an average team only played bad players, they would appear better than they were. Quality win strength is important. Relativity. Likewise, a player may be awful but be playing with great players and still win every match. If you can define his teammates as good or bad, based on their performances with different teammates, you may be able to define his skill.
I started out looking at the 23rd Tronic Ladle. I came up with two different scores. The first, Ladle Score, is defined as:
Rounds Played * Winning Percentage
This gives a numeric value that describes how far a team progressed in a tournament and how much they won each round. For example: a team that win’s each match 2-1 will have a worse Ladle Score than one that wins each match 2-0. Similarly a team that loses a match 1-2 will have a better score than a team that lost the same round 0-2.
The second figure, that may be a better one, is Ladle Win Quality Score. It is defined as:
Wins against Opponent A * Opponent A's Ladle Score + Wins against Opponent B * Opponent B's Ladle Score+...
This accounts for how much winning your opponents did through the entire tournament. For example: a team that loses 1-2 in the opening round to the eventual champions could very well have a higher Ladle Win Quality Score than a team that loses 0-2 to them in the Semifinal. Also, if a team loses 1-2 to the eventual champions they may very well have a better Ladle Win Quality Score than a team that lost 1-2 to a less successful team. The value is a bit recursive. Ladle Score counts winning percentage,essentially all wins. This means beating a team 2-1 improves your Ladle Win Quality Score more than beating them 2-0, but only if you finish high enough. The scores are most valuable relative to each other. Out of the context of a Ladle they are much less meaningful.
free kill took the idea further (and correcting an error I had made along the way). He exported all the scores from the Ladles going back to 18, where we started using the new template on the wiki, and sorted them. All the Ladles since then can be found at http://generalconsumption.org/armagetronad/ladle-stats. He’s released the code for how he did this. Copying over his table here is to do.
The table is sorted by Ladle Win Quality Score (LWQS), then Ladle Score, then Elimination Round. Twixted Xats 1 won the Ladle, Crazy Tronners 1 lost the final 2-3, CT2 and Wild West were the losing semifinalists, both having scores 0-2. Bearing the results of the Ladle in mind, a couple key observations:
Twixted Xats 1 and Crazy Tronners 1 were the finalists, and the have the both the highest Ladle Scores and LWQS. Plus, the only team to win a match against either finalist has the third highest LWQS. Both the semifinalists, however, have higher Ladle Scores.
When taking the time to determine Ladle Scores for all the Ladles, a LWQS for any given Ladle may be quite revealing.
zomg holers! 
Interesting, I wonder how these figures will look compared to other Ladles, especially those shorter ones, with less teams and rounds.
It would be nice to choose the “bests” after each Ladle. Best team is the one that wins the tourney, but how to choose best player or best defense? | think statistics can be helpful here.
I’m going to input all the 6v6 Ladles and create a total, draw a LWQS from that. Then see if I can get something for players.
You’ve made an error in how you calculate Ladle Score. The correct calculation is:
winning_percentage = won / matches
ladle_score = rounds_played * winning_percentage
You’ve made a mistake with winning_percentage, which you have defined as:
winning_percentage = matches / won
So using your calculation TX1’s winning_percentage = 13 / 11 = 1.18 = 118%, but of course you can’t win more than 100% of the time :).
I took your idea and made it a little more interactive: http://generalconsumption.org/armagetronad/ladle-stats
The source and tools used to create that page can be found at http://code.launchpad.net/~dlh/+junk. There you’ll also find a tool to export the Ladle results, teams, and players to JSON.
The stats only go back to Ladle 18, because that is the first Ladle that started using the bracket and team template.
I’m working on a ranking system similar to TrueSkill (http://research.microsoft.com/en-us/projects/trueskill)–there is some technical explaination at that site. However, to apply it to the Ladle, I would need a larger dataset to work with (DS Megafort / Fortress Café). Using just Ladle results wouldn’t work out well.
Correction: source can be found at https://code.launchpad.net/~dlh
ah, a knuckleheaded mistake.
great! I’ve updated the post and am going to copy over your work here.
“However, to apply it to the Ladle, I would need a larger dataset to work with (DS Megafort / Fortress Café). Using just Ladle results wouldn’t work out well.”
I think it’s time for some mainstream fortress server to have pretty stats o_o