I tried lowering the K-factor for the ELO calculation, but this reflects too poorly on players who only have 1 game. It was previously 64, lowering it to 32 (like in most chess rankings) didn't work. I have now set the K-factor to the value of the destroyed ship (empties are counted as being worth 35 here). This means a player is rewarded / penalised more for losing or killing an expensive ship, and less for just a scout.
This just in - 02/02/10: I changed the ELO rating on the dogfights, to only consider kills and deaths made in the first 15 minutes of the game. This is because later in the match, technology advantages such as blast cannons, blink, etc, and base attacks, make dying less about lack of skill, and more about necessity. This will effect the 'dogfight rating' of every player in the leaderboard. Nothing has been forgotten, score is still an adaptive equation, and I may change this back if player's don't like this I also changed the way total score is worked out - it is now almost entirely dogfighting score, Any scrap score over 200 is counted as 200, and only a quarter of that is used for the rating, so a player's rating can never be boosted by more than 50 by scrap score. Both of these changes were made so that the player stats, a measure primarily of skill, are not too adversely effected simply because the enemy team gained the upper hand in the strategy game. NOTE: The 'dogfight score' will not be accurate until players have many games played. Good players will be underrated, bad players will be overrated. As time goes on and games played goes above 10 or so, it will become accurate. In the mean time, since I (DarkFox) personally play in every game, and thus have a dogfight rating progressing more quickly to it's accurate value quicker than other players, I've modified my base dogfight points for killing or losing a ship from 64 to 6.4. This will have no effect on the final rating I might acheive, but it will make me progress there at 1/10th of the speed of other players.
This set of statistics pages were created with the intent of making team balance in online strat games easier to acheive.
These pages list statistics for players that have taken part in games hosted by {uscm} members. It also includes results from games whose servers have sent their log files to DarkFox.
I (DarkFox) am happy to include any log files from any players, just message me in-game or on the battlezone community forums and send the files along!
It's incredibly hard to judge a player's relative value to a team in battlezone 2. This is in part due to the fact that games are played in teams, but more significantly it's down to the fact that even a team of the best players will lose if their commander is new to the game. As such, these statistics don't actually take winning or losing games into account at all. Winning is certainly an advantage on the leaderboard, because it usually affords the winning team stronger technology with which to win fights and destroy targets, but these stats try to display individual player skills, whether they were playing with stacked teams or not.
The players on the leaderboard are rated according to their 'score'. This score is a simple sum of 2 other entries: Dogfight score, and Strat score.
Dogfight score is an ELO rating based on a player's dogfighting victories. This means that how many kills or deaths a player acheives is not used - a player's dogfighting score is based on the strength of the foes he kills or loses ships against. Killing a poor player will not result in much of a change of score. Killing a vet will result in a large score increase. Likewise, dying at a vet's hands will not penalise a player nearly as much as losing a ship to a poor player. ELO ratings start out at 1800. A chess world champion grandmaster, dedicating a lifetime to their game, can't reach much higher than an ELO score of 2800, so in battlezone 2 expect only the very, very best players to get a score greater than 2500 or so. 90% of players will be in the range of 1500-2100.
Statistically, a difference of 200 in score represents a 75% chance of the better player winning a dogfight.
'Scrap score', in simple terms, is the difference a player makes to his team's scrap for each half hour of playing (a typical game length). For example, if in 30 minutes a player loses 100 scrap worth of ships, but manages to take out 200 scrap killing enemy units, then his total scrap difference is 100, and thus this translates to a scrap score of 100. Scrap score can be negative as easily as it can be positive.
There are 4 complications with the scrap score, though...
1. Scavengers offer no penalty for loss - building and losing scavengers is part of the game. Note that players are still rewarded 20 scrap for killing a scav.
2. Similarly, turrets and service trucks offer only half their scrap to the player that destroyed them - this is because destroying turrets or trucks is usually down to team effort in winning field battles, as there's no skill to be measured in taking out an undefended truck or turret. Commanders will still be penalised for the full value of the lost unit though (which goes some way to balance the extra scrap and combat ratings a commander acheives with his guntowers & turrets).
3. Any buildings, including extractors, are not featured in logfiles, and thus are unfortunately never included in any of these statistics.
4. Finally, scrap score is capped at 200 for calculating a player's total score, to prevent turtle commanders from being rewarded too highly.
Taking a combination of dogfighting score and scrap score is important - a team without the ability to fight the enemy in ships is gauranteed to lose a game, however eager to kill scavengers they are! Likewise, a team of incredible dogfighters, that play every game using lasers to get kills, are not going to have much luck stopping the enemy from beating them in economy. Both skills are important, and hence why both are used here.
Since the standard deviation for the dogfighting ELO rating is roughly similar to the standard deviation of scrap score, no further modifications are necessary - a player's score is simply the sum of the two.
Stats program history:
This stats program has been through many iterations, and finding a good way of rating players has proven very difficult. It started out rating player's kills and deaths, but this led to massive scores for players who had joined a game, gained a kill, then left.
Newer versions then added a weight for games played, but however minor the weight became as games played grew, there was always an unfair advantage for players that played more regularly on uscm servers (me, DarkFox, being the obvious example - I did not deserve the number one spot, but I held it in previous versions simply because I (obviously) played on my servers more than anyone else).
Another problem that arose was players that had a massive kills/deaths ratio, and had played plenty of games, but who were over-cautious and had very few actual kills. These players were being rated the very best, despite the fact that their value in a team was highly questionable - sitting in an enemy base sniping will score a player perhaps 1 kill per game, but with no ships lost, this leads to an inflated rating. To circumvent this problem, the old rating system was again weighted, this time to favour players that were more aggressive. Ultimately finding a good balance was too subjective a matter, and the rating score became nothing more than my personal educated opinion.
The current version, written 5 years later in 2010, works differently. The scrap benefit per game is an obvious and relevant statistic to judge a player's teamwork. Additionally, the dogfighting uses an ELO rating, which will converge on a player's true skill after enough games have been played, but which is also reduced by competitors scoring kills, keeping it in check (and dynamic) despite how many games have been played. Note that the starting rating of 1800 is, by definition of it being the starting value in a logarithmic probability scoring system, the 'average'. Players do not start on a low score, and half of the players are expected to, ultimately, settle on a score below this.
All together, this means the current version of these stats pages are *not* specifically weighted to favour players with a lot of games, but it might take 5-10 matches before the dogfight rating will become an accurate measure of skill. (for those who understand the ELO system, we're using a K-factor of 64 here, which is quite high, and allows rather rapid changes of score).
Note that, due to the fact that scrap score can be negative or positive, with zero being 'neutral', a final score of 1800 can be considered roughly average.
A player's racial icon is given according to the player's most used race. An upgraded icon is given to those players that play as the race for over 3 hours. The racial icons are drone, warrior, thunderbolt, sabre, xypos and xares.
Kills, including sniper shots, on enemy _pilots_ are *not* added to your kill statistics.
: Award given for high kills/hour
: Award given for high snipes/hour
: Award given for good scrap score
: Award given for highest positions on leaderboard
: Award given for above average ISDF gametime
: Award given for above average scion gametime
: Award given for above average hadean gametime