Tuesday, August 21, 2012

Statistically Speaking

Lame post title alert! Oh. It's, um, too late isn't it? You've already seen it. D'oh! Look, it's late and the truth is I just don't care enough tonight to come up with a better title. I am, of course, about to comment on the latest batch of stats from the PSP and available for your entertainment, enlightenment and amusement over at PBAccess. And I am (surprise, surprise) gonna point out what seems like a weakness or three in the way the cumulative rating is calculated. (Please note I am offering up this mild criticism despite the fact 5 of the top 10 rated players are on Damage's roster--or maybe because 5 of the top 10 are on Damage's roster. Anyway the current standings do play a part in my thinking.)
Let's review first how the final player rating is achieved. Win Rating is 15% of the total. G Rating (confirmed kills) is 30%. Winfluence is 15%. Survival Rating is 22.5%. Points Played is 7.5%. And the Penalty Rating is a negative modifier that somehow counts 10% to the total Player Rating.
Before I go any further it's disclaimer time. Yes, the stats are a great idea. In general they are good for competitive paintball and given that the whole ball of wax is brand new and still being developed I'm not hating--just suggesting a couple of changes might be order at some point. 'Kay?
I have no problem with Win Rating--except that by this point in the season there needs to be a minimum number of points played to count towards a Top Gun or Best Player award. In baseball hitters don't qualify for the batting title unless they have 400 or more at bats. Similar concept. If a player has a great Win Rating but is only in 35% of his team's points there's a disconnect there.
G Rating by the league's admission is the most difficult stat to get right which is why it's confirmed kills. They aren't pretending to claim each and every elimination is being accounted for and yet it's worth almost a third of the final rating. That seems like kind of a lot of room for error. I'm personally also not thrilled with the message it sends to young players regardless of position that the greats must be the killers. Out of context it's an excuse to be selfish and not a team player.
Winfluence is, I think, almost deceptive because it is based on a player's winning percentages in comparison to his team's winning percentage. That means a few outlier or better players on mediocre teams will potentially have higher winfluence numbers than the players on the best teams because those teams tend to be consistent--among other things. And at 15% I think it is valued too highly.
Survival Rating isn't bad but it also isn't one size fits all. Wire leads are at high risk from the outset and tend not to survive points even when their team is successful and if the back player gets caught early it's a relatively much bigger deal. Which isn't to say there's anything fundamentally wrong with Survival because I don't think there is. It just needs to be tweaked a bit to make it a more telling stat by role. (This was something Matty mentioned in his original explanation of the stats and how they were determined.)
Points Played is really a modifier that adjusts scores based on percentage of a team's total points a player plays but at 7.5% it's negligible. For example Player A plays 40% of his team's points his points played rating is 3. Player B plays 60% of the same team's points and his points played rating is 4.5. Does 1.5 points sufficiently distinguish between those two players? Maybe but given the other number in play I think it undervalues the more valuable players.
The penalty one is fine, I guess. Of course the impact of penalties is at least partially accounted for already in points won. And if the stat guys count as a point played any point the player begins in the box then the impact of that player's penalties on winning or losing the points he plays is part of his Win Rating. Of course penalties are bad so taking a modest ratings hit isn't the worse thing either.

Here we are--closing in on year one of stats for competitive paintball. I think it's working out okay so far. You've just read what I'd consider changing--anything you'd like to add to the list?


Mark said...

If we could see how many kills are not confirmed versus those that are, it would say more about that stat.

Anonymous said...

Scroll to the bottom of the table. Players with fewer than 12 points played are excluded from the season stats.

Grant said...

I suggested on the pbaccess forum that they try and add assists - because for every kill there most likely is another player pinching out that eliminated player to create the elimination opportunity. Also some sort of stats on successful Vs unsuccessful bunkering moves would give some insight into this aspect of the game.

Cade said...

Finally, some interesting reading... We love the feedback, and at the end of the year we will be taking a step back and looking at the full data set to make adjustments to our pre 2012 assumption in preparation for the 2013 season. Considering this is the first full season of stat collection we have made some sizable steps in the right direction, but we understand there will be continued review to refine a balanced ranking system. I like the idea of having different weightings based on player position to get a better picture of the league’s best Front, Mid, Back players, but these changes will come in time.

Side note on Confirmed G’s: The Stats team have improved their data capture of G’s from 86% in Texas (first attempt with 4 spotters) to 94% in Phoenix (6 spotters) to 97/98% in Chicago and MAO. Some things that we are currently talking about is an update to allow for better tracking of shared G’s and a negative impact for ‘friendly fire’ of your own teammate.

Just like any team out there the Stats crew will walk the field to understand and anticipate the primary / bump positions players will attempt. They review field shooting lanes for angle, drop and bounce shots to make sure proper credit is assigned. They also pre-plan their observation points and make match by match adjustments depending on how the field/teams are playing. All of this in an effort to improve our accuracy during data capture and provide both teams and fans with the highest quality player data in the sport of paintball. (Drops microphone and walks off stage… hahahaha)

Anonymous said...

@ cade yes!

Baca Loco said...

first anon
I'm not advocating leaving any players out of the stats I'm advocating a minimum qualification standard to be considered the "Best Player" acording to the stats at season's end.

Thanks. When you say 97/98% is the claim that your guys are only missing 2% of kills or unable to confirm 2% of kills? (Or is that the same thing?)

Anonymous said...

The requirement for the batting title in baseball is 3.1 PA per team game, or 502 in a full season.

400 ABs was used by the AL from 1938 to 1944.

Baca Loco said...

Thanks for checking Wikipedia. My point remains the same.

Nick Brockdorff said...

Changes I'd like to see:

The survivability stat should be impacted less negatively, for players running wide on the breakout.... and conversly impacted to a greater extent for players not running wide.... maybe even a survivability bonus for those that make it far on the breakout?

What is "wide"? - well, up to PBA, but I'd say any further than the second grid from the centre, and you are going far.

On the kill count, spotters notwithstanding, I agree with Paul it should weigh a little less in the overall ranking, simply because there has to be a big margin of error there. Even for experienced players, it is hard to confirm kills when watching the game, especially breakout kills, where lanes are often doubled or tripled. I am thinking quite a bit of assumption goes into getting as high as 98 %?

I also think players need to be classed differently than "front, back & mid" as they are now. There are a lot of fronts that get stuck in a corner playing back, while the back plays front.... it's the nature of the game these days.... so fixing players in those 2 ancient definitions, is not reflective of the modern game.

I'd suggest instead breaking each player down in how much time he spends in each area (or zone) of the field.

First, I'd define players as "snake side, stand up side and mid" (and yes, the versatile ones will figure in all 3 columns). Secondly, I'd then give each player a "front rating" and a "back rating", which shows how effective they are in each aspect of the game.

Front would be defined as "when you have a teammate behind you" and Back as "when you have no teammates behind you".... So a back moving up to lose out the game, will stay a back, while a back overlapping his front, becomes a front for that part of the game.

Migh be too complicated to track - but I believe it would give stats that are more indicative of each players abilities and success.

raehl said...

I would point out that, no matter how the rankings are calculated, there will always be discussion about whether they are "right" or not. It's obvious that #5 is better than #50, but is #4 necessarily better than #5? You could argue that forever, just like you can argue who is the best forward, or batter, or tight end, or receiver... stats are a start to the discussion, not the end of it.

Anonymous said...

Like hockey. If you are on the ice when you're team scores,plus 1. Just the same if you're team gives up a goal minus 1. Add in G's and thats it.

Anonymous said...

I heard that the announcer is actually the guy that knows all this stuff, and that he drip-feeds the stats-guys information between plays. They write down what he says and provide it to the webcast.

Just what I heard...