A Whole Bunch of Ladder Statistics Season 5

[soccernamlak]

It’s that time again:



soccer{ball}’s Ridiculously Overdone Statistical Analysis of Ladder Season 5!

Links to Season 4 Ball Part 1 and Part 2. Season 4 TBD

[soccernamlak]

In the past two weeks, we’ve (finally) put an end to Ladder Season 5, stretching a record-breaking 10 months from April 2012 to January 2013. I, of course, blame DN.

But since it has ended, Mikesol and Aki were gracious enough to provide me the SQL file of the Ladder Database, full of tables and data. After a bit of patience and work, Aki and I were able to put all the data into csv format, and then I went to work. Here’s the fruits of my labor: stats on practically everything you could (almost) think of for Season 5 Ladder.

I’ve decided this time around to do it all in one post. We’ll start off with some basic facts. Next, we’ll look into player statistics: where you all are from, rankings, win/loss streaks, etc. We’ll follow this with some Map statistics. We’ll wrap up with the majority of my work: correlating stats with outcomes (Note: Correlations are the largest amount of analysis, comparing ratings with stats in all measures).

Before we begin, I want to point out a few things. I am still waiting on Aki to hopefully piece together a MySQL Query that I can run through the ladder database to determine:

1) Rating of player vs. typical plane used and / or percent of time that player used that specific plane.

2) Typical winning team composition overall for each game mode, and perhaps even by map.

Is this possible? Absolutely. Is it decently difficult to put together because of how ladder stores stats on games and players? Yes. In addition, since ladder doesn’t accurately count plane usage time, I’m assuming I’ll run into issues with that. But no matter.

If/When I receive this data and/or query, I’ll run some analysis on it and make a separate post. Basically, you won’t have to keep checking this thread daily to see if I’ve got new information or not.

Anyway, that being said...

Let’s get to it!

Index

Both Modes
Country Information
Geo IP Map
When Did We Play Ladder?

BALL
General Information
Player Stats
Random Facts
Map Stats
Score Stats
Correlations
So What’s The Secret To Success?

TBD
General Information
Player Stats
Random Facts
Map Stats
Correlations
So What’s The Secret To Success?

Concluding Thoughts

Player Requested Stats

These may be on different pages....

Kills vs. Goals

Part 2

Plane Types
Rating vs. Plane Type
Rating vs. Percent Use Top Plane Type
Multiple Regression

Coming Soon
Plane Setups for Winning Teams, potentially by map as well.

[soccernamlak]

Where We Live

This first section today deals with where we call home: the countries we hail from (or at least play in).

First, we’ll include the US and look at the straight numbers and proportional percent compared to the body of players:





As expected, the US dominates in shear number yet again, but is the minority majority, holding only 45% of the total ladder population across both Ball and TBD. In fact, this has decreased since last season, where 47% of Ball and 54% of TBD players came from the United States. So perhaps the EU servers have really spread out the player base a bit more and encouraged greater EU participation. So let’s look at these same numbers minus US:





No big surprises here. We see the United Kingdom at 16%, Denmark and Canada at 10%, Netherlands at 9%, ad France at 7% as our Top 5 non-US players in numbers. This really isn’t a huge change from last year in order, just a bit more players from each to push the US down a bit percent wise.

[soccernamlak]

Now, one thing I should point out before chants of “EU EU EU” come about:



Even though the US is only at 45%, the number is much higher for those who would benefit from the US server. Canada, Central America, and South America all would have better pings in a US server. Even Australia might prefer it (it’s about 100 miles closer to US server than EU server from Sydney). Just something to keep in mind. Also, 'Murca:





Anyway, moving on to the really cool part: GEO IP!

[soccernamlak]

Where We Live - GEO IP

I found a program online to batch plot IP address locations on a map. I took a screenshot of a bunch of semi-closeups instead of directing you to the actual map itself for privacy reasons. While you couldn’t directly associate the IP with a player, some player’s locations (in general) are fairly well known. Some of the IP locations plotted, I’ve discovered, are also extremely close to where the player actually lives. Others are hundreds of miles off. So for your safety, here’s a ton of pictures.

General Worldview:



The rest are in these IMGUR albums and/or picture links:

Asia

Australia

Europe

North America

South America

It’s 33 pictures, which at Altitude Forum’s 4 image limit per post, would take me 9 posts just to show them all to you. Ew.

Anyway, takeaway here is that we come from all corners of the globe. A ton of people from the US. A decent amount from Central and South America. A good number from the Middle East, Australia, and Southeast Asia. A ton of people from Europe, from pretty much all over. And then a few people in remote locations of the world, from Iceland to Hawaii to Kansas.

[soccernamlak]

When Did We Play Ladder

Keep in mind here that the data should be taken with a grain of salt due to how EU and US inputs time into the database. Aki can explain more if you are interested. But basically games played in EU servers after a game in a US server will show up as being played earlier. Anyway...

Let’s look at BALL first. Here we see days of week:



Pretty well balanced in my opinion. Saturday being the largest contributor here, which is to be expected due to it being a weekend.

If we look at the times when it is played:



We see the biggest percentage of games played in the afternoon and evening. Note here that the times are in UTC I believe (Aki correct me if I’m wrong!), so theoretically a 21:00 game (or 9PM) would be 4PM EST.

This is incorrect

As updated from Aki:

Negitive. Times are recorded based on the parser's home time zone. This means when US games play, it saves at about -6 UTC, and the EU parser is close to 0 UTC.

If this is the case, I still think it points to how majority of games are played in the evening hours regardless of server, with American afternoons and European Evenings joining.

[soccernamlak]

Now for TBD days of week:



We something similar to BALL, except that Sunday is just a tad bit larger than Saturday. This makes sense because not only is Sunday still on the weekend, but many TBD ladder matches seemed to only start up either before APL or even SL.

Times of day:



Similar to Ball, but with ladder dying off earlier in the night than Ball ladder would.

Finally, a comparison of time of day as a percent for TBD and Ball:



Note that TBD typically peaked earlier but died quicker than BALL.

[soccernamlak]

General Information

Just some fun numbers.

For BALL:

Number of Matches: 10,243 (compared to 6,129 in prior season)

I think this is interesting, considering that Season 4 actually ended on time after 4 months, or 127 days. Season 5 (this past season) was 10 months, or 280 days long. So Season 5 was 120% longer than Season 4, but only 67% more matches were played.

In fact, if we look at it as a per-day calculation, there were 48.25 matches per day in Season 4 but only 36.58 matches per day in Season 5.

All of this suggests the possibility that either:



a) Ladder is becoming less popular and is losing to other servers

b) Altitude is dwindling in size of active members

c) SSD is banning too many people

d) EU / US ladder split is alienating more people than it is bringing them together



Personally, I think this is evidence of b: that our player size is decreasing, resulting in less active players which correlates with less ladder activity and thus less games per day.


Total Time Played: 95,053.33 Minutes = 1584.2 Hours = 66.01 Days

Percent Up-time (Total Time Played / # of Days in Ladder Season): 23.6% or about 6 hours a day

[soccernamlak]

Player Stats

First, let’s go with some straight facts.

For BALL:

Number of Players: 709 (compared to 558 in prior season)

Average Ranking: 1752 +/- 666.8 (compared to 1682 in prior season)

Average Number of Matches Per Player: 173.2 +/- 322.5 (compared to 131.3 in prior season)

Most Matches Played: Sanis (2150)

Average Win Percentage: 44.90% +/- 21.68%

Best Win Percentage >25 Games Played: 72.2% in 36 Games by A Frail Old Lady

Best Win Percentage >100 Games Played: 66.25% in 160 Games by mssv

Best Win Streak: 19 by MattMan

Worst Losing Streak: 15 by KM DW

Now, a few overall graphs for your viewing pleasure...

We first start with a view of the number of games played by a player, ordered from least to greatest:



As you might imagine, it’s a fairly nice exponentially increasing function here, which should make sense: more players try ladder for a few games while a few play ladder 24/7.

With ratings plotted the same way:



We see how the average rating hovers at 1752 overall and closer to 2000 for players with greater than 20 games played. In addition, it’s a tertiary equation, ensuring how the majority of players are grouped close to the average, with disproportionally less players ranked really high or really low.

The win percentages:



Works the same way: an average around 50% with very few players losing badly or winning big.

Finally, frequency of streaks >2:



Similar to last season, we see a good deal of streaks up to 8 or so; from there it drops off pretty quickly. Not surprising, as ladder is actively trying to get you to your supposed skill level quickly and then stop your streak (or in other words, try to get you to win or lose more quickly depending on which way your streak is going).

[soccernamlak]

Random Fun Stats

For BALL:

Most Lopsided Game (aka Longest Game with 6-0 score): 

Ball Funnelpark, taking 17m 43s for a 6-0 result, or 177 seconds per goal.

Closet Game (aka Shortest Game with 6-5 score): Ball Maze, taking only 4m 50s for a 6-5 result, or 26.4 seconds per goal.

Average Number of Ping Kicks Per Game: 0.166

Average Number of Leaves Per Game: 0.127

In other words, worst case about 1 out of every 3 games has someone getting disconnected, but realistically probably closer to 1 out of every 5 or 6.

Total Ping Picks: 1,704

Average Ping Kicks Per Player: 2.4

Average Ping Kicks Per Player with >0 Ping Kicks: 5.1 in 334 players, or 47.1% of player database has >0 ping kicks.

Total Leaves: 1,301

Average Leaves Per Player: 1.83

Average Leaves Per Player with >0 Leaves: 5.68 in 229 players, or 32.3% of player database has >0 leaves.

Player with most ping kicks: Cpt. Slow (88)

Player with most leaves: Nip (75)

Player Most Likely To Get Disconnected (Ping+Leaves / Games) with >10 games played: Chubsy 66.7% of the time

Player Most Likely To Get Disconnected (Ping+Leaves / Games) with >50 games played: 2achary 32.7% of the time

[soccernamlak]

Map Stats

For BALL:

First some basic information.

Shortest Game: 2:05 on Skylands (compared to 2:24 on Cave in prior season)
Longest Game: 27:11 on Lost City (compared to 30:42 on Mayhem2 in prior season)
Average Time: 9:16 (compared to 9:07 in prior season)

Now let’s look at some images. First we have our data of how many times each map was played, in both numerical and percent form:





As you can tell, Cave, Planepark, and Asteroids are the clear favorites. Every other map was played roughly the same amount of time, except Cutesy, which was played a grant total of 24 times in 280 days, or about once every two weeks.

I also looked at minimum, average, and maximum playing time for each map:



For minimum playing time, Cutesy stands out as the highest (i.e. don’t pick this map if you want a short game) followed by Snow, perhaps suggesting the sheer size of these maps and/or ease of the defense to prevent quick transition goals.

However, interestingly enough, Cutesy almost reigned in the lowest average time played, narrowly being beaten out by Darkwar, which is expected considering the layout of Darkwar.

I also decided to look at how lopsided maps were based on left-team win percentage:



Normally I’d say Cutesy is lopsided and favors the right team (which wins over 57% of the time), but with only 24 games played, it’s a bit too early to tell.

However, with the rest of the maps played >500 times, Lost City is the clear winner of being biased on team placement, with the left-side winning 55% of the time, no doubt helped by the center-rock formation. At 54% to the right team, Funnelpark comes in second, which is a bit odd seeing that it’s a pretty well mirrored map.

However, what’s probably the most interesting stat of all of this is that every map except 2, the right-hand team wins more. So want to win more? Hope ladder puts you on the right hand side of the map. This becomes even more biased to the right side of the map when you consider that all the top played maps favor the right side.

[soccernamlak]

I also decided to look at how the time of games over the season stacked up. Since at the beginning of the season everyone is effectively not ranked, does ladder tend to overstack one team creating shorter on average games?

To find out, I looked at all of the game times over the course of the season in order:



That’s a lot of data points!

Then I looked at just the first 250 games:



For the overall times, you can see the cluster between 3 minutes and 12 minutes, but no real rise or decrease in match length over the season. If we look at the first 250 games, we see roughly the same outcome: no correlation between first few games and elapsed time. Meaning that ladder is doing a pretty good job at ensuring we aren’t having too many overstacked and quick games, despite players not really being ranked yet at the beginning of the season.

Finally, for fun, I decided to look at the length of games in Cave, ordered from shortest to longest:



I was very pleased with this graph, as it shows an almost perfect tertiary fit for the distribution of game lengths. In other words, most ball games fall within some sort of standard distribution with a small percentage of the games going by very quickly or taking a while.

[soccernamlak]

Score Stats



Now to consider score distributions on these maps.

First, we look at the percent of games that ended with a 6-0, 6-1, etc.



Now, if scores of the losing team were completely random, each piece should be 16.67%. However, as we can see from this pie chart, only 4%(!) of games ended with 6-0. Out of over 10K games, that’s ridiculously small. In fact, only 15% of the games were lopsided (6-0 or 6-1 outcomes) with a whopping 45% of games close (6-5 or 6-4 outcomes). I think this is a pretty good testament to ladder’s sorting system, assuming a closer game equals more balanced or evenly matched teams.

Probably a bit surprising is how well this mirrored last year’s data:



Note how the amount of 6-0 games actually decreased while the number of 6-5 games increased. Sure, 1 percent might not seem like much, but considering the number of games played, you’re looking at about a 25% increase in 6-0 games despite 67% more games being played.

How does this compare for each map though? I looked at percent of bad, okay (6-2 or 6-3 outcome), and good games for each ball map:



The map with the highest percent of bad games came from Lost City, Grotto had the largest number of okay games, and Cross had the highest number of good games. I’m excluding Cutesy here due to the fact that only 24 games were played last season on this map. Being that it is in the random rotation for this season, it’ll be interesting to see how it fares.

The one thing I mentioned last year was the introduction of Reef and Skylands. They had not been played enough in rotation to determine if they would be comparable quality wise to the other maps. Well, as you can tell, the maps stacked up well with the heavy hitters this year, again providing some evidence that the maps do provide for close games, despite any complaints on the layout.

In fact, overall, these aren’t that widely varied from one another, hinting that team composition and ladder balancing might have more of an effect on the score than the map layout itself.

[soccernamlak]

Correlations

For BALL:

Now for the absurd amount of data. What influences rating in ladder? Hold on tightly, we’re about to go through a ton of data...

First, we’ll look at rating versus number of games played:



Not a strong correlation, but a bit better than last year by 3%. Again, you need to play a decent amount of games before ladder can start sorting you comparatively to other players, but once you do, things tend to work out, assuming you play semi-consistently throughout the season.

This correlation, as expected, even weakens as we look at only players who played on average less than one game every 5 days (aka greater than or equal to 50 games):



An R-squared of 0.2473 isn’t terrible considering the cluster near the beginning, but again not a strong correlation regardless.

Now let’s consider the offensive and defensive strategies. First, do average kills per game matter?



Not really for the most part. If we assume “great” players had a rating of >3000 last season, then we can see a wide range, from 15 kills to 55 kills per game.

Assists paints a similar story:



Again, a pretty wide range of assists per game for top ranked players with very little correlation with rating overall.

[soccernamlak]

Deaths paint a similar, but perhaps more constricted picture:



Here, top ranked players only varied in a range of about 25 deaths per game, with everyone forming a sort of bell curve here. All of this (kills, assists, and deaths) I talked more in depth about last year, concluding that these statistics don’t really matter too much so long as you are actively participating in offense or defense, indicated of course by actually killing, helping kill, and getting shot at.

Take this display with a grain of salt because of how ladder counts deaths:



Still thought it was cool regardless to see how kills, assists, and deaths per game stack up with rating.

Now let’s look at offensive measures. First, rating vs. goals per game:



A bit weaker than last year actually, with the majority of top players either not really scoring or scoring almost closer to 2 goals per game. The “elite” players (>4500) had at least one goal per game.

If we look at just players above 50 games played,



We see a stronger correlation now, but still relatively weak. Our “great” players above 3000 still varied widely between 0.25 and 2 goals per game, but there’s a suggestion here visually that “carrying” your team by scoring more might equate into a higher rating.

[soccernamlak]

If we look at goal assists



We see an interesting pattern that while statistically is insignificant, visually hints that great players at least average 0.5 to 1 assist per game. However, averaging more than that is either not possible or not recommended, giving rise to this idea of a team sport versus individual carrying, at least when it comes to movement and passing.

What about ball possession per game against rating?



Again, not really a strong correlation, but top ranked players still at least have to touch the ball.

I even looked at rating versus the average number of goals per possession per game:



Not much here. Top players and average players still clustered around the same values. Which also hints at just how ball is played as a team sport. There aren’t many players who can just quickly tap the ball and score many goals. But on the opposite hand, if you have the ball for a while, chances are you’ll probably end up scoring.

Actually, come to think about it, while rating isn’t dependent on goals per possession time per game, is there still a correlation between average goals and average possession time?



Indeed there is! The longer you hold a ball per game, the more goals you’ll score on average. Therefore I conclude that ball hogs will carry us all to victory....

soccer{ball} - Misrepresenting statistics since 2011. 


[soccernamlak]

Next we look at the passing game. First, receptions per game versus rating:



Nothing much here. People who receive passes still do really well or really bad. And some high rated players don’t even get that many receptions!

What about actual passing?



Again, nothing strong here, but we at least see a visual trend that high rated players to tend to pass slightly more than average ranked players.

Next, we look at defensive strategies: number of ball carrier kills per game:



Again, while the number of ball carrier kills per game center around 10 or so for most players, it does show that our top ranked players aren’t doing any worse than about 6 kills on the ball carrier / game.

The more interesting visuals from ball carrier kills is this in my opinion:



I looked at the percent of ball carrier kills compared to total kills per game on average versus rating. Despite where you ranked, it seemed that the majority of players hovered around 28-30% of their kills on the ball carrier. Which is interesting considering that BALL needs a dual strategy: kill the ball carrier but also kill potential receivers of a pass.

[soccernamlak]

This graph holds true even if we look at players with greater than 50 games played (keep in mind the X-values....I’ve expanded it a bit just to show you how clustered it is)



Again, many people cluster around this 27% mark. I think this tells us more about how BALL is played versus how players are rated or skill level.

Finally, we look at recoveries per game:



Again, not a whole lot of correlation here, but to be top ranked, you do need to actually attempt a decent amount of recoveries per game.

[soccernamlak]

So What’s The Secret To Success?

FOR BALL:

You want to increase your ranking, but how? Well, based on my analysis, you can't concentrate solely on scoring or killing, for example. The best altitude players are the most well-rounded

So, what do you need to do to help improve your ranking?

1. You actually have to play games throughout the season. Don’t expect to play 10 games in 10 months and be in the Top 25. This should be obvious.

2. You do need to kill, but again don’t focus on killing. Top rated players varied widely on kills / game. This also goes for assists / game as well.

3. You are going to suffer plane damage and die while playing ball. Top ranked players averaged anywhere from 30-50 deaths per game based on the direct data, probably a bit less than that considering how ladder records death. Still though, it goes without saying: don’t play TBD in ball. Attack the opponent’s goal, but don’t die immediately with every push.

4. Speaking of killing, you are going to kill the ball carrier around 28% of the time, no matter how good or bad you are. That’s just the way it ends up being.

5. Most top rated players are offensively minded, not defensively. So you’ll need to average around 0.6 or more goals per game, and around 0.3 or more assists per game, to start being considered an elite ball player.

6. This also means that you’ll need to actually grab the ball at times; don’t be afraid of it and pass as soon as you touch it.


7. You’ll need to receive passes, make passes, and make recoveries from kills throughout the game.


Thus, in conclusion, balancing scoring, assisting, killing, passing, recovery, and dying will help you rise up in the ladder ranks faster than just playing a sheer number of games. So being more well-rounded helps than a specific skillset. One other thing to notice is that based on how ball is played and these results, offensive-minded players will be rated higher overall despite the fact that defensive-minded players (high kills, low deaths, low goals, etc) are just as crucial to victory. I wish there was a way to represent their contributions in ladder, but I’m not sure of the best way to display this.

Why is this an issue you ask? Great defensive players paired with okay offensive players will ultimately lose more games, as you can’t defend every push successfully forever. You need good offensive players to carry, push, and score. These offensive players end up being the highest rated in ladder.

[soccernamlak]

General Information

Just some fun numbers.

For TBD:

Number of Matches: 1,592

Total Time Played: 15,528 Minutes = 258.8 Hours = 10.78 Day

Percent Up-time: 3.85%

Player Stats

First, let’s go with some straight facts.

For TBD:

Number of Players: 385 (compared to 339 in prior season)

Average Ranking: 1549 +/- 252.2 (compared to 1578 in prior season)

Average Number of Matches Per Player: 41.32 +/- 70.51 (compared to 54.95 in prior season)

Most Matches Played: Blarg (406)

Average Win Percentage: 43.24% +/- 24.60%

Best Win Percentage >25 Games Played: 72.41% in 29 Games by Mikesol

Best Win Percentage >100 Games Played: 62.5% in 128 Games by Aya

Best Win Streak: 13 by DN

Worst Losing Streak: 14 by Tjerki


Now, a few overall graphs for your viewing pleasure...

We first start with a view of the number of games played by a player, ordered from least to greatest:



As you might imagine, it’s a fairly nice exponentially increasing function here, which should make sense: more players try ladder for a few games while a few play ladder 24/7. Perhaps more interesting is the how much greater a difference there is between “active” ladder players and non-active players in TBD versus BALL.

With ratings plotted the same way:



We see how the average rating hovers at 1549 overall and close to 1700 for players with greater than 20 games played. In addition, it’s a tertiary equation again, ensuring how the majority of players are grouped close to the average, with disproportionally less players ranked really high or really low.

The win percentages:



Works the same way: an average around 50% with very few players losing badly or winning big, but perhaps not as neatly packed together as we saw in BALL.

Finally, frequency of streaks >2:



Similar to last season, we see a good deal of streak up to 8 or so; from there is drops off pretty quickly. Not surprising, as ladder is actively trying to get you to your supposed skill level quickly and then stop your streak (or in other words, try to get you to win or lose more quickly depending on which way your streak is going).

[soccernamlak]

Random Fun Stats

For TBD:

Average Number of Ping Kicks Per Game: 0.11

Average Number of Leaves Per Game: 0.13

In other words, worst case about 1 out of every 4 games has someone getting disconnected, but realistically probably closer to 1 out of every 8 or 9.

Total Ping Picks: 178

Average Ping Kicks Per Player: 0.462

Average Ping Kicks Per Player with >0 Ping Kicks: 1.84 in 97 players, or 25.19% of player database has >0 ping kicks.

Total Leaves: 200

Average Leaves Per Player: 0.519

Average Leaves Per Player with >0 Leaves: 2.7 in 74 players, or 19.22% of player database has >0 leaves.

Player with most ping kicks: Fluffy (11)

Player with most leaves: Fluffy (25)

Player Most Likely To Get Disconnected (Ping+Leaves / Games) with >10 games played: Tosconi 27.2% of the time

Player Most Likely To Get Disconnected (Ping+Leaves / Games) with >50 games played: Fluffy 19.3% of the time

[soccernamlak]

Map Stats

For TBD:

First some basic information.

Shortest Game: 2:16
Longest Game: 35:33
Average Time: 9:45

Now let’s look at some images. First we have our data of how many times each map was played, in both numerical and percent form:





As you can tell, Cave and Asteroids are the clear favorites here, easily being played at least 3x more often than any other map. 42% of the games were either one of these maps, and that’s a lot.

Interestingly, Lost City was the clear loser here, which I thought was odd considering the groans every time Grotto or Core is selected by random.

I also looked at minimum, average, and maximum playing time for each map:



For minimum playing time, Cave was the shortest, which is probably expected based on the map design, but Grotto wasn’t far off, a bit surprising considering it is a neutral map that is fairly wide.

Average playing time varied a bit more than Ball, with the higher average and longest playing maps dominated by the neutral bombs. This, of course, is expected.

I also decided to look at how lopsided maps were based on left-team win percentage:



Just look at Woods! Almost 60% of the wins by the right hand side. Very intriguing, again considering the mirror image of the map. Mayhem and Fallout come close to 57% win-rate for right and left respectively as well, again fairly interesting considering the uniformity of the maps. Perhaps most surprising is how, on average, the win rate for each side is closer to 50% overall, compared to Ball where the right side was favored to win more often than not.

[soccernamlak]

I also decided to look at how the time of games over the season stacked up. Since at the beginning of the season everyone is effectively not ranked, does ladder tend to overstack one team creating shorter average games?

To find out, I looked at all of the game times over the course of the season in order:



Then I looked at just the first 250 games:



For the overall times, you can see the cluster between 3 minutes and 14 minutes, but no real rise or decrease in match length over the season. If we look at the first 250 games, we see roughly the same outcome: no correlation between first few games and elapsed time. Meaning that ladder is doing a pretty good job at ensuring we aren’t having too many overstacked and quick games, despite players not really being ranked yet.

Finally, for fun, I decided to look at the length of games in Cave, ordered from shortest to longest:



I was very pleased with this graph, as it shows an almost perfect tertiary fit for the distribution of game lengths. Again, we saw similar things with BALL: how most games fit within a certain distribution with very few outliers on either end.

[soccernamlak]

Correlations

FOR TBD

Right, TBD time. Let’s start with games played against rating:



Some correlation here as we saw in Ball, but this is to be expected to start off with. We see a decrease in the correlation:



when we look at greater than 50 games played.

Now let’s look at killing strategy in TBD. First kills:



We see a greater focus on killing in TBD, with top ranked players (>2000 rating) clustering around 20-25 kills per game, highlighting the increased focus on killing in TBD compared to BALL.

Assists are a bit more varied, however:



Really varies from 10-30 assists per game. So, a greater importance, we see, is on actual kills versus helping kill players.

[soccernamlak]

Dying is predictable:



Good players don’t die as much as average players, but just because you don’t die doesn’t necessarily mean you’re a top ranked player.

For fun, XP per game:



This is probably the most interesting out of all the stats, showing that top ranked players get the most XP per game, due to a combination of kills, assists, and bomb hits no doubt.

Now we look at bomb running. First let’s look at rating versus base damage per game:



A slight correlation here, but top ranked players are all over the place, indicating the greater importance of defensive players and offensive players on a team being crucial to winning.

Bomb carrier kills per game:



Nothing really stands out here except that all players tend to have around 4 kills on the bomb carrier per game.

[soccernamlak]

Bomb diffusing:



Again, nothing really telling here except that top ranked players do tend to group higher up than normal, average players.

I decided to also look at bomb carrying time per game:



Again, no strong correlation, bringing us back to two more distinct player types in TBD: Defensive and Offensive.

Bomb Pickups:



and Bomb Releases:

[soccernamlak]

and Bomb Hits:



Are even less correlated than bomb carrying for the same reasons.

However, if we look at average base damage per hit:



We see perhaps not so much a correlation with rating, but how most people tend to have around 20 damage per hit per game, with some distinct vertical, linear marks at common bomb hits (made by those who played only a few times).

If we look at this further:



We see how the number of base damage in the game directly correlates with average hits per game. Expected of course, but nice when the statistics confirms it.

Finally, for fun:



Just a nice image showing rating against pickups, releases, and hits per game. Notice how they all get exponentially smaller as we go from pickups to releases to hits.

[soccernamlak]

So What’s The Secret To Success?

FOR TBD:


You want to increase your ranking, but how? Well, based on my analysis, you need to focus on assisting and killing while maintaining low deaths. You also need to figure out if you are a better defensive or offensive player.

So, what do you need to do to help improve your ranking?

1. Killing is important. Top rated players had at least 15 kills/game. You can't just solely sit back and run bombs: you need to be pro-active. And if you can't kill, you need to work on assists (13/game). Even partially killing another player is better than not killing at all. Note, though, that killings is far more important than assists. Top rated players and bottom rated players could have the same number of assists per game, but the killing bottomed-out at a much higher rating.

2. Expect to die. Here, dying doesn't matter to an extent: most players hovered around 20 deaths/game. Not dying that much means you are hurting yourself and your team (trying to hide/run from fights could be detrimental to your victory). Dying too much means you aren't helping on offense or defense (especially since TBD isn't a 3 second respawn).

3. Figure out if you are offensive or defensive as a player.

3a. If you are defensive, 4-5 kills/game should be on the bomb carrier with 1-2 diffuses per game. In addition, your kill/death ratio should be higher than your offensive counterparts. However, you still should attempt a few bomb runs in the game: top rated players in a defensive position still attempted offensive pushes when necessary.

3b. If you are offensive, you really don't need to worry about killing the bomb carrier (perhaps only 12% of your kills) or diffusing. Instead, worry about high number of pickup, drops, and hitting the base; all things a defensive player shouldn't really worry about. However, don't focus on high damager/base hit: top rated players varied heavily between light to heavy damage. The key here is to at least hit the opponent's base.

[soccernamlak]

Concluding Thoughts

Before finishing here, I want to point out again the offensive/defensive standpoint in TBD. This is completely opposite of Ball. We see in TBD that high rated players are either focused on killing bomb carriers and diffusing the bomb, or picking up, dropping, and hitting the opponent's base. This is evident by the wide range of values for each of these graphs of top-rated players. In Ball, however, we saw that top rated players were ones who were offensive in nature, maintaining a high (compared to other players) goals/game and assist/game ratio.

I think this is evident by the game play mechanism at work here. In TBD, there is a high value placed on both offensive and defense. As TBD games are typically much longer in length than Ball games, greater patience is required. In addition, a strong defense is necessary due to the longer times of offensive runs. Whereas in Ball, the quick pace means you need defense, but best players are constantly pushing in offensive runs.

Again, this presents a fundamental issue with rating systems between the two modes. I think a good, strong defense is needed for both modes, but only TBD ratings rewards defensive-minded players. If they’re playing ball, on the other hand, they won’t come anywhere near the Top 25 on average.

I hope to confirm these “How To Do Better Posts” once I get data on rating vs. preferred plane and common compositions of winning teams.

I hope this shines some light into Season 5 of ladder, perhaps one of the longest so far. Let me know if you have any questions or want to see additional data!

[soccernamlak]

reserved reserved last one i hope. i'm in trouble if not.

Edit: Got it. Yay me.

[Clapon]

yes you are

[soccernamlak]

Originally Posted by Clapon

yes you are

Pssssshhhh I got dis.

[Clapon]

I've been holding back every urge in my body sense f to not post, I'm surprised I made it to the end tbqh

I look forward to seeing all the stats!! ^^

[soccernamlak]

Originally Posted by Clapon

I've been holding back every urge in my body sense f to not post, I'm surprised I made it to the end tbqh

I look forward to seeing all the stats!! ^^

I appreciate it Clapon, thank you! I have all of it typed out; I just need to edit it in. So within the hour

[soccernamlak]

Ok, just finished all the final edits. Took an hour just to reserve posts and edit/format it all in correctly. Yikes. Anyway, hope you all enjoy! I'll try to have a second post on plane composition for winning teams and high rated players hopefully soon, or whenever Aki gets a free moment to breathe.

I appreciate your comments and feedback in advance! If I'm missing something you're interested in, let me know!

[classicallad]

This analysis is very helpful, thank sox <3

[DMCM]

That IP map is pretty cool.


I can tell exactly which one I am, but I have no idea who the other 6 Portuguese players are.

[DICK van DYKE]

You're a very beautiful nerd Soc. I'm looking forward to more

[mssv]

Great job!

Is there a way to see the relation between kills and goals in ball pls?

[AzzedariuSuiradezza]

Lovely read. This is why Altitude should have a bigger playerbase.