Is there no respect for the dead? Can’t we just leave the corpse to it’s eternal rest? Why am I subjecting Vanilla TF2 to the equivalent of exhumation and being ridden around the graveyard in a macabre game of horsey horsey with this record breakingly long graph fest on the joys of i52? Just remember, if the cops come by, this is the way we found it and we’re just trying to bury it again as any civic minded citizen would do.

One year ago I started these articles reviewing the stats at i49. At that tournament an innocent and bright eyed young Demoman named War (at that time rated as one of the best Demos in Yorkshire) playing in an unfancied local team made his name and has since become an integral fixture at the top of the TF2 scene, with i52 marking his final coming of age representing Europe’s finest. And then he (and the rest of his team) fucked it up.

Hopefully I’ll rise to the challenge of summarising this awe inspiring event a little more successfully than the grand finalists, although the spectre of players deliberately selling out their team to inflate their stats purely to look good in this article has raised its barely believable head, so much of this article also contains experimental and never before seen analysis methods. Can I find a way to stop the stats being dominated by log trolls? We’ll see.

To begin, a caveat *

I’ll get this out of the way first, the Epsilon vs Immunity log from the group stages was lost, which was a fairly comfortable win for Epsi in which War assures me he did massive damage, compiled a huge total of frags (which just keeps getting bigger at every telling) and no doubt the rest of his team did well too, so any time you see an Epsilon player marginally behind another player (particularly an Immunity player) apply this asterisk and consider mentally moving them up a spot.

Sorting the statistical wheat from the chaff

My initial investigations revealed some unusual results and the reasons why weren’t immediately apparent, but without going into too much detail it appears some players took their games against weaker teams more seriously than others to a massive extent. To cut a long story short removing games involving any team outside the top 8 seemed to restore sanity to the proceedings, so sorry to Redmond, Dogs and all the others who battled into the knockout rounds – you’re fired.

There is one exception however, if we do this Epsilon lose an extra game because TC knocked Ayo Gurl into the lower bracket. Based on this TC were clearly a cut above the average mix team (admins – put the self-confessed drunks into the kiddy pool next time) and their single game against Epsilon is included as it just punishes Epsilon to have their knockout games start against Classic Mixup when everyone else had clear 5-0 wins to buff their stats. Following the same logic I also allowed TC vs Awsomniac. Of course TC don’t get included any further in this article except for leaving their fingerprints on the stats, sorry chaps.

What remained were only the hardest fought games between the best teams global TF2 had to offer… and the shattered booze addled husks of esports youth in the form of Ayo Gurl. The final decision was whether to split the groups, effectively glorified scrims, away from the knockout stage. The knockout games represent TF2 played in its most intense form with any mistakes potentially resulting in elimination where true playing skill would surely out.

Ultimately I think removing an entire day’s worth of games that thousands watched would probably divorce this article too far from most people’s experience of the tournament, and doubling the workload to analyse them separately would be prohibitive, so they’re all lumped together for this article.

The Teams

I’m doing things a little differently to begin with in an effort to avoid simply slapping on the charts and ploughing through the same shit one more time (don’t worry, I’m sure they’ll make an appearance further down). First I’m taking a look at the team overview, the damage they output, their KD and my patented officially Droso-disapproved rating.

1

I’ve got the teams arranged in order of their finish in the tournament rather than their stats, and the obvious thing to notice is that Epsilon are a little inferior to Mixup who are a little inferior to Immunity. Given that there’s a cigarette paper between these teams and they finished in reverse order what accounts for the difference from the actual results?

Epsilon were smashed completely in the grand final which supresses their stats (as well as buffing Froyotech’s), but it also comes down to games in the lower bracket. Epsilon being knocked out last left them with a knockout record of TC, Mixup, Froyo and Mixup again, whereas Immunity had Publiclir, Froyo, Awsomniac and Mixup (two of those games they won fairly convincingly although Awsomniac was more of a struggle) and Mixup had Awsomniac, Epsilon, Reason, Immunity and Epsilon again (two of those were convincing wins). Also, don’t forget the lost log.

Obviously in those convincing wins, of which Epsilon only really had one, teams were able to buff their stats, so the team stats table reflects this. The other teams in the lower part of the table didn’t get the same benefit, each one going out as soon as they faced a stronger opponent.

So what do the stats look like just in head to head games between these teams?

2

As you’d imagine from games that went to golden cap best of 3s there’s almost nothing in it but you can see that just in the context of the head to head the statistical trend does follow the results, with Epsilon having overall the best KD and DPM.

The ratings are quite suppressed in relation to team DPM, this is because usually they’re taken in the context of the top teams smashing a few weaker sides and boosting their numbers but in this case only Epsilon sneak a positive KD ratio.

From this we can draw the conclusion that IMs overall tournament stats are boosted a bit against weaker teams to a greater extent than Epsilon or Mixup, that Mixup’s jetlag/exhaustion on the first day probably held their stats back a bit and that Epsilon’s destruction in the final really hurt their stats as a team and individually, plus of course the already infamous missing log.

It’s tempting to start wanting to make further caveats to the logs, particularly the final, but they represent what happened in the tournament so if you turned up drunk, late or dead it’s only fairly reflected in the stats.

Team Anatomy

I want to look at how the team’s internal organisation of the fragging classes effects the distribution of roles to give a wider context to the individual stats so we have an idea of who got the heals, what jobs they were set up to do and how they worked together.

Froyotech

3

Most of the features of this chart show that Froyotech appear to be an orthodox classic NA style team. The scouts dominate capping duties with B4nny in particular responsible for this, but even though he’s the defensive scout he doesn’t get appreciably any more healing than Clockwork or his roamer. Lansky dominates the healing received playing the strong man soldier with over 30% heals with duwatna on about 25%.

Damage output however is much more democratically spread with Lansky and Duwatna accounting for less than 50% of the team’s damage between them, with the highly aggressive Clockwork and conservative roamer Blaze doing similar work for their team.

In terms of finishing the frags Clockwork and Lansky are the main contributors with Duwatna picking up a very low share of kills, taking on the role of assister in the classic NA style. Blaze the roamer also picks up a solid share kills despite playing his team’s most disposable class, although he played it with a stronger focus on survival than most.

In terms of KD Lansky dominates survivability for his team with he and B4nny the players most likely to still be there at the end of fights. Clockwork’s highly aggressive style and role of initiating scout sees him record a relatively small share of his team’s KD, but he doesn’t do badly compared to other team’s aggressive scouts.

4

Examining the team’s assists we can see that B4nny’s role as the brain of the team has led to him being the fulcrum of a fragging trinity with Clockwork and Lansky. Only Blaze out of the combat classes has a preferred fragging partner in Clockwork by a small margin. Shade’s bodyguard Lansky is predictably his primary partner and Duwatna plays close to his Medic too. The scouts do the majority of cleanup on Duwatna’s damage.

The delineation of the flank from the combo is expressed by Blaze’s relative distance from the scouts from Lansky, and a slight preference to Clockwork over B4nny.

Epsilon

5

Epsilon exhibit some of the characteristics of the classic EU demo pocket style, but it’s safe to say that Mike is even more marginalised from a typical pocket than the supporting pocket soldier than you might think almost playing as a second roamer expected to make big plays rather than consistent ones.

The scouts dominate capping for Epsilon to an enormous degree taking on nearly 70% of the capping duties, with Mike the player giving up capping duties the most being involved in less than 10%.

Healing is dominated by the combo as it is for every team, but in this case it’s heavily in favour of War playing the strong man demo receiving over 30% of his teams heals. Kiler gets the benefit of the defensive role by being buffed a little more and presumably building uber but Bash and Tek are on the bare minimum. There were scouts who got healed less but not at the sharp end of the competition.

The team’s damage output is also dominated by War falling just short of 30%. The nature of damage ouput tends to lead to it being less clearly warped in favour of one or two particular players so this is really a big share for one player. Bash played as the aggressive scout so you’d expect a little more damage but he has the smallest share of any Epsi player. Despite being significantly less healed Mike still managed a solid 25% of his team’s damage.

In contrast to Froyotech Epsilon expect their demo not to just get damage but to finish frags, with War picking up nearly 25% of the team kills, and as a result the scouts and pocket are less important to the team in this regard. This has a similar effect on the KD balance which War also dominates. It’s noticeable that even taking 25% of team heals Mike only has around 18% of the team KD. Given his undoubted quality this is indicative of a distinctly non-pocket approach to survivability.

6

In strong contrast to Froyotech it is War closely pocketed by KnOxXx into combat who form the main fragging partnership for Epsilon, with Kiler performing the main cleanup on the Demoman’s damage. The second most important partnership is Kiler and Bash. It’s indicative of Mike’s split role, and the extent that War is pocketed in combat, that his strongest collaborator is also Kiler.

Tek and Bash worked together on the flank and neither of them saw KnOxXx much in combat at all, although Bash mixed with his other teammates far more than Tek did.

Classic Mixup

7

In some respects Mixup’s style is highly orthodox but in others it’s unusual. As you’d expect their scouts dominate capping with back cap specialist Enigma leading the way. It’s not unusual to see a Demo chipping in with a few more caps than his soldiers either.

As far as healing goes TLR is pocketed fairly strongly but most teams tend to have a more obviously pocketed player. Typically for NA, Platinum is the secondary heal getting around 25%. Enigma is left to subsist almost entirely on healthpacks.

As you’d expect the demo outputs a lot of damage but Mixup are the only top tier team who don’t have two players combining for over 50% of the whole team’s damage output. This is probably because A Seagull gets such a large amount of production from the roamer position, something he notably shares with Blaze. The scouts are obviously marginalised a bit here due to the limitations of their class.

The share of the kill rates is where things start to get a bit weird. Platinum has a little under 20% but apart from him it is a strangely egalitarian spread with the roamer actually picking up most of the slack the demo leaves. Then when we look at KD the trend is slightly reversed between the same individuals with Platinum picking up the slack left by A Seagull in terms of survivability.

Those of you who are familiar with the article series know I use a rating based around a proportion of KD and DPM and Mixup’s is by far the most balanced across the whole team with a standard deviation less than half of any team’s other than Reason, whereas Epsilon and Immunity’s are more than 3 times bigger. This indicates that Mixup aren’t dominated by one or two players hogging the stats and rely on high level team play across the board.

8

For Mixup it’s the scout partnership that clearly lead the way by a significant margin. Next most potent was the pocketed TLR who was Harbleu’s bodyguard and team spearhead when fighting. His main cleanup partners were the scouts. Like Duwatna, Platinum’s main partner is a scout, this time Squid although he also spent some time fighting with the medic.

Platinum didn’t really have a strong relationship with TLR indicating a high degree of autonomy for the Mixup demo, who had a wide range of fragging partners. Squid also got around spending the most time fighting with A Seagull, who saw very little of his medic during engagements.

Team Immunity

9

By contrast the antipodeans have formed themselves into similar configuration to Epsilon, moving from using a supporting Demoman to a powerhouse fragging machine. As with most other teams the scouts cap most but Termo obviously likes to ensure he has no challengers on the scoreboard by capping almost as often.

In the healing stakes Termo and Yuki absorb the most getting around 30% each. There’s very little left out of this for anybody else with just Snowblind edging up to around 15%.

At i49 Yuki was the center of damage output for his team as well as so much else, but in the new Immunity Termo hogs the damage output getting nearly 30% of his team’s total damage and Yuki is relegated to the low 20s. The roamer and scouts have a strangely equal damage output.

Termo also tops his team’s fragging output mirroring War with Yuki and Snowblind getting similar numbers. When it comes to KD then it’s the scouts and Termo who have the survivability with Yuki under 20% of his team’s KD. Aporia has the smallest proportion of his team’s KD of any roamer in the top 4 teams.

10

The cleanup crew of Sheep and Snowblind lead the way for Immunity but they’re a team without any other standout partnerships instead having a number of closely grouped pairs of players. Bonobo’s primary target in combat was Termo with Yuki a near 2nd.

Aporia is largely isolated on the flank except for Snowblind, who was everywhere, the glue between the flank and the combo classes. Sheep stuck much more closely to Termo and Yuki. Termo’s damage was like a magnet to most of the team, as was Yuki’s to a lesser extent.

Awsomniac

11

Of the two scouts Zappis is the dominant capper and as always the scouts dominate the capping duties.

The French org with no French players played with the ultimate pocket, Zebbosai swallowing his share of the heals and anyone else’s he could get his hands on. At 35% he shares the dubious honour of most heavily pocketed player with Sideshow although in the case of Ayo Gurl it’s of less strategic interest. Kaidus also still gets a respectable 28ish% and it seems the result is that Stark and Zappis have rollout heals and not much else, although this seems to be typical for Stark.

Zebbosai and Kaidus are pretty evenly matched in terms of the damage they put in to the team and they also sport the thoroughly EU style damage configuration with two big daddies taking over half the damage output with the roamer and scouts outputting similar amounts.

As you might expect Zebbo takes the largest slice of kill rates with Stark predictably the main clean up man and also as you might expect Knutsson doesn’t have a very large share.

In terms of survivability Stark as always exhibits greater survivability than you might expect from his heal share no doubt due to the kind of backstabbing cunning you have to exhibit to become that much of a cancerous gamer (this lad has a future in politics). Knutsson has only about 10% of his team’s KD.

12

There are 3 main frag pairings for AWS, Mirelin and Kaidus, Mirelin and Zebbosai and Stark and Zappis. This would chime with Kaidus’s comments on Fully Charged about how he started taking more ubers as the tournament progressed fighting with the heals while Zebbosai made moves elsewhere.

Both of the scouts worked a little more around Zebbosai’s damage than Kaidus’s but Zappis strongly favoured Knutsson. Kaidus seems oddly isolated from his soldiers although this could be the Mirelin effect getting in on as many assists as possible, although it could just as easily be the result of Kaidus’s movement around the map.

Reason

13

Amongst the Euro teams Reason seem to have taken the most inspiration from the NA scene… or is it just that their pocket and demo aren’t up to scratch? Shall I mention the scouts dominate the capping again? Why not, the scouts dominate the capping – 5 more words on the count.

Healing wise Huhy and Drackk are fairly heavily pocketed, but Rising also gets more than you might expect for a roamer with Herr_P completely marginalised.

When we move onto damage and into kills per minute the importance of the combo diminishes to the extent that the notional roamer Rising and scout Smzi start to dominate kills in a similar way to Classic Mixup, in fact even more so. Huhy ultimately struggled for kills a bit and Drackk for survivability.

14

Reason’s stat distribution was oddly Classic Mixup-like and like Mixup the scout pair of Smzi and Herr_P lead the team by some margin. Their combinations with their soldiers are indicative of the flank/combo split, with Smzi favouring his pocket soldier while Herr_P worked more with Rising although in his case the divide isn’t quite so obvious.

Huhy and Drackk were the pocket focuses and Huhy/Bulle take the place of their 2nd most prolific partnership, Drackk actually assisting his medic fewer times than Smzi. The only other invite group team this happened in was Ayo Gurl, the team whose pocket soldier had S bound to kill.

Publiclir

15

Yes it’s not Saints, they’ve got/had a sponsor and the scene needs those so they’re called Publiclir.se, got it? Yadda yadda scouts capping.

Vani clings on to his soldier and demo for dear life giving nearly 2/3rds of heals between the two, squeezing a few extra drops from his medigun for Dennia.

In terms of damage like Reason they are one of the minority of teams whose damage share drops under 50% from the combo, but whether this is stylistic as with Classic Mixup or a possible sign of weakness from the combo isn’t clear from these stats.

Despite receiving vast heals the pocket soldier ends up with no better survivability than his roamer and with little more kill rate.

16

Vani and Smirre are head and shoulders ahead of the team as a pair and given the amount of healing Smirre received this shouldn’t be a surprise. Alba and Ond Kaja are notable for spreading out their frag partnerships across the team.

The secondary groups were Smirre and Ond Kaja and the scout pair of Alba and Dennia. Given the beatings Publiclir received perhaps it’s expected for a lot of team shape to be lost as survivors fight together on the retreat a lot of the time.

Ayo Gurl

17

Less of a team than a chronic liver condition waiting to happen, the main thing you can take from this chart (except for how much the scouts dominate capping) is how hard Permzilla was baiting his team for stats. Even taking this into account it’s slim pickings – he dominates the stats of the worst team in the competition by a long way. Hint to future would-be stat whores – i49 War did it against the good teams as well.

Sideshow matches Zebbos healing share but that’s the end of the comparison between the two.

18

Unlike the other assist charts this records who can hold their drink the best. Evidently Permzilla and to a lesser extent Hawku are the least subject to hangovers while KingOfSquirrels may have been better served by a nearby infirmary. That’s the way to reach the top of the ETF2L admin tree kids.

The Players

Sorry about the wait, you can switch off the brain now… or can you? In recognition of the status of i52 I’ll be going into more detail than ever before on player damage and kills. Yet more strange details of player behaviour remain to be revealed…

Demomen

19

This chart shows (in no particular order) the use of damage within the context of the end result, be it a kill, assist or both combined. It’s calculated by dividing the damage per minute (DPM) by the Damage Per Kill (DPK, the average amount of damage a player did before they registered a kil) in blue, Damage Per Assist (DPA, the same as DPK except for assists) in red or a combination of them both (DPKA) in yellow. The higher the number (and the longer the bar) the less damage the player needed to do to achieve the kill, assist or the combination of the two which represents efficient conversion of damage to an end result.

In terms of damage efficiency it seems there’s a strong variation between those teams that pocketed their demo and those that didn’t. The expectation is for those leading the charge is to get both good damage per kill and damage per assist because they’re fighting in close quarters and also calling their damage for others to follow up on. Those playing a more supporting role would expect their damage efficiency per kill to suffer, but to do better on damage efficiency per assist. A player roaming more might expect to get better damage per kill than assist.

The demos close to the top of the list, Termo, War, Permzilla and Smirre were all pocketed by their teams to a greater or lesser degree, and generally had good damage conversion in both assists (Perm obviously benefitting from Hawku’s ability to take his drink, as well as completely dominate his hungover team in all important kill and damage metrics) and kills.

Some readers may be disturbed to see Perm and Smirre with such large damage totals, but think of the enemy team as a dinner plate of potential damage. In kaidus’s or Platinum’s team their teammates are voracious gluttons taking huge bites leaving a relatively smaller meal on offer, whereas Ayo Gurl and Publiclir were filled with anorexic supermodels turning their nose up at the damage available in favour of the spawn queue, so a single well placed sticky at choke hits 4 targets pouring through for 200 damage just before they gun you down.

Kaidus and Duwatna show opposite sides of the coin. Kaidus took more ubers than he has during the season but Zebbosai still dominated his team’s healing during general play, so you’d expect Kaidus to be a bit isolated playing his roaming game a lot of the time apart from leading some pushes and getting kills on isolated occasions both of which would lead to limited opportunities to rack up assists. Duwatna on the other hand played a supporting role putting out the damage for others to clean up on and his relatively long red damage per assist bar shows this, whereas his damage per kill is amongst the weakest.

Platinum and Huhy struggled with efficient conversion of their damage in the tournament as a whole, although Platinum arriving late and getting straight on the beers on day one might have had something to do with that. Judging from the spread of his assist stats Platinum seems to play a sort of roaming supporting demo role so he sees the worst of both worlds statistically.

The yellow bar in the graph above shows the combined efficiency of demo damage being converted into results. The assist part is partially team reliant as you need someone to clean up and so its inclusion isn’t all about the individual, but that’s also true of the damage per kill – good support from the team still help players get kills for less damage output. So if we adjust DPM by the damage-per-kill conversion efficiency we get an adjusted damage figure: useful damage

In blue we have the original damage per minute and in red the adjusted damage per minute. Duwatna is the big winner here as his close–to-equal-bottom DPM is turned into a respectable mid table adjusted DPM. Demomen who weren’t the center of their team’s activities and had team mates capable of cleaning up the opposition without them suffer. Still there is a team component, if your team does better general damage there will be less damage needed to claim a kill, but adjusting for this is a much more complex exercise beyond the scope of this article.

20

Asking readers to mentally adjust for things like the role explained in the team charts above and explored in the text is probably too much to expect, so I started to experiment with some ways of showing this more clearly. What does it look like if we adjust this new DPM value by the share of that player’s team’s DPM they account for? For example if a Demoman accounts for 20% of his team’s damage then he’s not playing the same role as one who gets 30% of his team’s total DPM.

To do this I’ve divided the Adjusted DPM calculated in the chart above by the % of the team’s DPM output that player accounted for making the wildly optimistic assumption that the % share is accounted for entirely by the tactical role they fulfilled. Failures in a player’s innate quality may be to blame, but hopefully those shortcomings will still shine through.

21

Duwatna wins again, and the strongly performing players in a weak team start to see their credentials eroded and high quality Demos who played a less well supported roaming style start to converge with them in the middle of the pack. Huhy gets the short end of the stick, but this really only chimes with common opinion of his Lan, not a very good one from a personal perspective.

The question is whether I want to use this as a basis for later charts such as a rating. In some respects it helps rule out players compiling stats by baiting their own team, but it may also punish players who have done well in bad teams and possibly unfairly boosts players whose teammates make things easier for them. I think we’ll just have to add this as another flavour for comparison.

22

Moving onto the kill statistics we have the rate at which kills were gained in red, and the KD achieved by each player in blue.

There are some broad similarities between the patterns established in damage, although some players such as Kaidus, Platinum and Duwatna, the players who sacrificed some damage output in their role in the team, get back some compensation here.

Termo also has a huge lead in KD, whilst getting an equal kill rate to War, Perm, Smirre and Kaidus he boasts a vast superiority in survivability. He came to i52 to take the Demo crown and by all metrics it’s hard to argue with the results so far… although there is that missing log.

However having accepted the idea of adjusting stats by team role for damage above, I did the same analysis on survivability.

23

In this chart the original KD is in blue, but the KD adjusted by the team % of KD the player had (which we’re saying reflects sacrifices they make for their role) is in red.

Duwatna rockets into the lead, adjusted massively upward to account for his relatively small share of the team KD pie and personally benefitting from the extent to which his team pockets the seemingly indestructible Lanksy.

Kaidus and Platinum also flex their muscles joining War who flirts with mediocrity but they all still play second fiddle to Termo. Permzilla is brutally punished by the awfulness of his team as he was practically the only player with a positive KD anyway, but it’s good news for Huhy who manages to regain a bit of respectability.

All of which leaves us with enough nonsense to produce multiple ratings that should suit every taste. The rating number was established largely as a way to order players with diverse DPM and KD stats and to give some kind of idea as to an effective proportion of the two. It follows the basic formula of (DPM * 0.6) * (KD * 0.8)

To recap clearly in the new experimental system DPM is divided by Damage per kill and then DPM is divided again by the percentage of the whole team’s DPM that player had, and the same percentage scaling is done to the player’s KD. However the results aren’t entirely to my liking.

One of the holy grails of statistical analysis is to link healing rates to other statistics but it never works without leading to a hugely disproportionate penalty to the players who were healed a lot as they don’t get any compensation for their half of the job – keeping the med alive. However the current situation lends us an opportunity to bring healing percentages into the calculations.

DPM and KD are being adjusted by the share of it that player contributed to his team in an effort to account for the limitations of the role they’re being asked to play. However part of that role, and one that might help shine a light on under/over performance, is healing, so I’ve averaged the DPM share and KD share with the healing share. So if a player got 25% of his team’s damage and got 15% heals then the DPM modifier we’re dividing the player DPM by is 20%.

Yeah, I know it’s not easy to follow but it makes some sense, and once it’s done they’re then combined in the same format as the original rating to produce a single rating number.

Whether any logic has survived the scaling, adjusting and general dicking around we’re about to find out….

24

So take your pick, in the blue corner measured against the bottom axis we have the original rating that works straight from the raw stats. It’s the only one I’ve done some mathematical work from to prove its broad (as in long term over hundreds of rounds) efficacy and to its detriment Permzilla and Smirre are rated very highly (above players who are unquestionably their betters), but in the yellow corner we have a new contender adjusting raw stats to take into account production, the team role and healing support.

All the adjustments are effective at producing a ranking that’s debatable but not obviously warped in the same ways as the raw stats and does a much better job at passing the eyeball test, with the Demomen anyway. With no disrespect intended Termo, Duwatna, War, Kaidus and Platinum (OK, maybe Perm is just in there but it’s close) is a more credible top 5 than Termo, War, Permzilla, Smirre and kaidus (with Platinum clearly behind in 6th barely ahead of Huhy and the Lan winner nowhere).

Whichever one you like Termo is right at the top, he wasn’t overly healed, he wasn’t baiting his team, he didn’t pad his damage stats, he was just good. Duwatna benefitted the most by adjusting his stats to his role, in the supporting demo role he was enormously efficient in assisting his team but some of the credit for his rating must go to his cleanup crew – Duwatna’s rating is, more than most, an expression of his team’s cohesion. Given that he played a supporting role in his team is it really fair to rate him so highly?

It does expose something of a weakness in this system that’s a little difficult to explain clearly. Because Duwatna’s stats are scaled to his team role, and because his team have the strongest collective stats by quite a bit, then all their players end up being scaled to that collective strength which may hide the individual skill level too much. Either way Duwatna certainly did everything his team asked of him, and given the chance he racked up massive stats against Epsi in the final.

It seems War must lose his global i49 crown and settle merely for European domination, but he remains firmly ensconced in the international demoman firmament whichever way you approach his stats, perhaps the rediscovery of a lost log and a better final might have kept him closer to the top.

Finally we have a way of looking at the stats that moves Kaidus into a fairer position for his skill level. Relatively modest healing support, a lack of a traditional support role trailing in the wake of his pocket and highly efficient death match production add up to a much higher position than his raw stats suggest he should have, and is only behind Demos who are debatably stronger.

For the Demoman for whom there is a suspicion of baiting, Permzilla, the new way of looking at things reduces him to something of an also-ran on the international stage. In all seriousness he performs his team role adequately, but his team role was the only fully conscious player left.

No one hurts more than Huhy. Make sure you give him a big hug at the next Lithu Lan. This is bad news for his team mates who will have picked up the slack in their share of the team’s output, but they can always hang their hats on the original ratings system – only they know if their tactical system is to blame or their team mates ineptitude.

Pocket Soldiers

Usually I bung all the soldier in together but for i52 we have to put on our statistical Sunday best and make more of a formal recognition of roles, so we’ll start with the pocket soldiers and try to make sense of the various styles of pocket on show – the carnival strong man, the pseudo roamer and Sideshow.

25

Regarding kill efficiency Lansky is ahead with Zebbosai in close pursuit but apart from Sideshow there’s a strangely even grouping in the middle of the soldiers with nothing in it really between them. One unusual feature of the damage per assist numbers are Mike’s figures being as low as Sideshow’s. With all due respect there’s no way Mike’s deathmatch will have been equal to Sideshow, so it must illustrate a lot of periods of relative isolation.

26

The effects of adjusting to production on the raw DPM figures is largely meaningless at the top end, Lansky manages to stretch his lead a little but it’s in the mid table the main action happens. Drackk’s 4th placed DPM is pushed back into the pack due to a lack of end product, and Mike’s raw figure still leaves him third on total output but by a smaller margin. Sideshow is also ejected from the general main group and is pushed down the stairs and locked in the cellar.

27

Looking at the kill metrics there’s not much between the two strong man pockets Zebbo and Lansky on kill rates but it’s on survivability that Lansky makes a huge stride forward recording a tournament KD in excess of 2 – an extraordinary achievement.

Yuki and TLR are joined at the hip in 3rd but Mike, sent consistently bombing and left to fend for himself results in him recording a totally lacklustre KD. Of course Sideshow falls well below the standard. Remember kids, don’t bind S to kill.

28

Looking at raw KD values compared to the values adjusted by team role, it’s Yuki who gains the most leapfrogging Zebbosai who is pushed into 3rd. TLR loses some ground to Mike who got his stats boosted a little in recognition of his role as a sacrificial lamb.

Incredibly Sideshow makes a comeback – Drackk I can only apologise, it’s a side effect of Perm dominating his team’s KD.

So it’s looking good for Lansky to have top rating but we’ll see how adjusting for healing and combining everything into a rich creamy rating dough affects things.

29

It doesn’t make much of a difference from the raw figures, the main feature is that after eroding all of Mike’s raw stats he finally ends up almost inseparable from Yuki when his healing disadvantage kicks in. However the soldier twins who dominated i49 are both marginalised along with everyone else by monster stats by Lansky.

At i49 he was somewhat outclassed but all of Froyotech’s problems were solved this time around letting the NA beast flex his wiry muscles as the almost indestructible battering ram that led his team to victory.

Zebbosai took a massive amount of heals but just about justifies them by securing 2nd although it becomes a lot more marginal, and we finally bury Sideshow while Drackk and Ond kaja lose touch with the mid table.

Is it reasonable for Lansky to have such a huge lead? At i49 there were more soldiers playing the heavy healing battering ram role but the strategic fashion at i52 revolved more around pocket demos with pocket soldiers expected to display more mobility and independence whilst protecting the medic.

Zebbosai was the only other top level example but it’s telling how Froyotech’s tight play reinforced their individual quality whereas Zebbos deep mono-uber style often left him isolated.

Roaming Soldiers

Moving onto the roamers we’ll see if anyone can stand up to the might of A Seagull or the perfectly coiffured Blaze.

30

There’s a separation at the top with the NA pair and Reason’s Rising forming a little group of their own with the kill efficiency making up the small differences are between them. Surprisingly cAPS who had a horrific ETF2L season along with many of his team is a little ahead of the pack on damage.

Possibly most surprising is Tek’s relegation to the status of a roamer also-ran, and considering how close he was to death (both through alcohol poisoning and Sideshow’s driving) Kos is relatively close to the others.

31

Looking at the DPM levels adjusted for role Seagull leads damage output marginally ahead of Blaze and Rising, and although the pattern largely follows the raw damage, Tek manages to squeeze in front of Knutsson while cAPS stays in that middle spot of his own.

32

The kill rates very much shadow the damage output although incredibly kos manages to lift himself off the bottom and kill things faster than both Tek and Knutsson. I’m not saying a lot of thought went into it, but they went to the spawn queue.

Blaze is a little behind Seagull and Rising on kill rates as he was with the damage but with the KD figures we begin to see some differences, Blaze showing a little more survivability than Seagull and both clearly ahead of Rising. Tek also stands out from the mediocre on this stat.

33

So adjusting KD for the role the players fulfil for their team there are two big movements, one of them to be partially expected in that Aporia is generally recognised as a decent roamer but one change terrifying for the credibility of this exercise is Knutsson’s KD suddenly leaping into the upper echelons of roamerdom, ahead of even the mighty Seagull and Rising.

At the top of the list Blaze suddenly puts a clear margin between himself and his challengers, and Rising begins to fall back into the pack.

34

Looking at the final rating figures they generally follow the raw stats, although adjusting for his team role and healing support Tek did significantly better than his basic stats suggest and it’s yet another occasion where his close proximity to an IM player reminds us of that lost log.

It also appears Rising had better resources to work with than Seagull or Blaze, but he may be slightly suffering at the hands of Huhy’s collapse over emphasising his role in his team. Either way it’s an impressive showing for a player who has been away from the top level for a little while.

cAPS doing well may surprise many but it’s not just a result my mathematical mangling, his basic stats are solid too.

Of course roaming is not all about base statistics, the roamer is the play maker, the game changer who makes things happens that radically change the course of a round by flipping the flow of the game on its head.

35

Even with the extraordinary quality of roaming soldiers at i52 medic kills and uber drops at the hands of a single player are rare, so the rate is measured per hour, rather than per minute. Medic kills in blue are measured on the bottom axis, uber drops in red are measured on the top axis.

We can see immediately that the highly successful stats, although relatively conservative in nature, posted by blaze don’t include prolific actions against the opposition’s most high value target. There’s nothing wrong with this, but it does mean they were accrued without storming the most strongly defended point in contrast to A Seagull.

cAps who posted pretty good general stats also appears to have done so from a position of relative safety.

The big winners here are Rising and Tek, with Rising obviously attacking the Medic often and with some success, but the real standout is Tek’s Uber dropping capabilities nearly 3 times as often as his closest competitor, not just bombing the medic but making it happen faster than they can right click. And these are just Tek’s stats on soldier – don’t forget the backstab against Mixup on Granary mid.

Despite Tek’s heroics A Seagull’s extraordinary general stats matched with his big play ability make him stand out.

Scouts

Although there is variation in scout role it’s much more fluid than with the soldiers so the traditional big scout list will carry on for i52.

36

The two Froyotech scouts stand atop the damage conversion metrics once again benefitting from their team’s tight play and focus fire. It’s worth noting that B4nny’s total figure in particular is boosted by his Damage per assist efficiency which is the highest amongst all scouts while his damage per kill is matched by Smzi and Snowblind. Is this a result of genuinely being everywhere at once or is it a result of spamming his scattergun at anything that moves and picking up the odd extra assist to boost the overall number from that chip damage?

Clockwork on the other hand has clearly the best damage conversion to kill. Once a season he mercs in ETF2L and posts league leading stats but he never plays enough to make the end of season review. The question was whether the European teams would adjust and play to control him – they didn’t. Other standout scouts with good damage to kill conversion are Smzi, Snowblind and of course Stark, hater of wasted damage.

37

Although damage output isn’t particularly a goal for scouts when adjusted for kill efficiency (not including assists) we can see that Clockwork establishes a significant lead. Smzi moves into 2nd slightly ahead of B4nny and rounding out the top 3 in damage.

There’s no redemption for Bash here, his DPM adjusted for efficiency is no better than the raw stats as his ETF2L season struggles continued into Lan.

The second tier elite group is largely unaffected although Stark’s outstanding efficiency helps him bridge the gap from the mediocre to the elite group, while Zappis is detached. Sheep notably is in the lower half but even adjusted for production damage is only a small part of the game particularly for scouts – they need to survive, clean up and cap.

38

It’s no surprise to see Clockwork at the top of the kill rates but we can also see the price to pay in lesser KD compared to the very best survivors. Smzi just edges out B4nny on kill rate but B4nny has a far higher KD leading all scouts playing a relatively cautious game but making sure he picks up significant cleanup frags.

For European TF2 watchers it’s no surprise to see Stark right up there and Sheep also posts elite stats (as does his IM counterpart Snowblind) in this area having lost out on aggression. Dennia makes a surprise appearance in a bid to get out of the dungeon, but we have yet to adjust for his role in the team or the level of support he received.

39

Adjusting KD for the role played in the team makes Sheep the big winner far exceeding the performance his team role dictates. Other winners are Enigma, clockwork and Bash finally achieves some respectability. The biggest loser is Smzi who takes a significant hit which is partly down to Huhy having a poor tournament over emphasising Smzi’s importance. Kiler also takes a hit.

So we come to the final reckoning for scouts and there’s a problem. This one.

1

The fact is that certain scouts that build uber get a lot of heals that have nothing to do with being buffed for combat and don’t really affect the game in any other way we’re measuring for the players, yet we’re adjusting scouts for heals received. What can we do?

In my inimitable style we can make a massive bodge. I’ve assumed that the scout on the team with the highest healing got it by uber building with the boston basher and the scout with the lower healing for the normal scout heals, so the least healed scout on either team is the healing % of the team’s total heals I’m applying to both scouts.

What does this mean? That any scout that built a lot of uber gets a fair crack of the whip, but any team where the scouts built equally will probably be penalised a little. This makes the ratings here less perfect than the others but they’re still interesting…. look, if you’ve read this far it’s too late to start arguing the toss.

40

The top two is the same under any system and there’s no denying that Clockwork was the top scout of i52 and therefore top scout globally. His play style embodies the perfect mix of enormous DM ability and timed aggression, more than simply a thorn in the opposition’s side during clean-up he was consistently initiating and winning engagements. B4nny played the perfect foil to this, relatively passive but devastatingly effective.

The biggest contrast between rating systems is Stark leapfrogging out of the middle of the pack into 3rd, and this is down to his ability to get the job done in a heal-less desert. He had one of the lowest healing per minute totals and when I checked to see if he was scavenging health packs to make it up he was one of the lowest users of those too. Add to this that he got a relatively large proportion of his kills against demo and medic targets and it’s hard to argue against him receiving recognition. If you like the idea of taking healing support into account then you’re a fan of Stark’s i52. My honest opinion is that even as a % of the team heals the effect is too strong, but I have yet to discover a mechanism to solve this.

Another more modest improver over his raw stats is Sheep but considering how many of the other top scouts suffer he actually gets bumped up into the top 6 joining his partner Snowblind.

Amongst the scouts suffering a little under a modified rating Smzi slips to 4th and may justifiably prefer the traditional raw stats considering Huhy’s relative lack of success has probably over emphasised his role in the team. Even so, top 3/4 scout in the world caps two years rising through the ranks to the European elite and with the decline of some of the older guard he’s now firmly established as one of the top dogs.

Mixup’s scout pair get a hard time if you take the adjusted ratings over the raw stats with Squid ending up getting slightly the better of TFTV’s overlord Enigma. Most of this rating loss is down to the atypically egalitarian stats their team approach produces over emphasising how much they should be doing for the team then punishing them for not doing it, and maybe they both built uber equally. However amongst all the teams Mixup played most explicitly to win without thought to aggression and KD and this meaningless charade won’t mean anything to them individually.

Kiler also gets a tough break compared to his raw stats, but Bash had a disappointing tournament overall regardless of the method of measurement. I know, I know he makes the space, but should he really be trying to push them into their spawn? In fact there’s no saving grace for most of the strugglers, the mathematical munging hasn’t produced a fairy tale ending.

Snipers

41

Did that marvellous i52 video released by Publiclir show Dennia with his face pressed up against his monitor? Well if it did then stats confirm: it works. As long as you don’t want headshots that is. That’s a lot of bodyshots but Dennia got a faster sniper kill rate than the almighty king of sniping Sheep but in terms of head popping accuracy there’s no comparison.

In the absence of Jukebox, Sheep is uncontested as the kind of skull cracking headshots getting twice the headshot rate of his nearest competitor. Tek is best of the rest although Stark is notable for very high headshot accuracy but with a much slower rate. Clock should stay on scout, although if the price of having him on your team is the odd sniper excursion then just let him do it.

Uber dropping stats are a bit boring and not worth being illustrated in a fancy chart, but Tek got another Medic drop to add to his collection, Stark got 3 leading the pack and Sheep amongst a few others got one, except Dennia who found two although probably not with headshots.

Medics

Medic is the final core class and I won’t be turning my attention to any novel ratings, rating a medic is even more impossible than rating players so they will merely be subject to my ill informed judgements and taste for hyperbole. So let’s get on with it.

42

Heal rates are the blue line on the bottom axis, kills+assists/deaths is in red measured on the top, and assists per minute is the rather squashed up line in yellow also measured along the top.

Healing rates don’t vary a great deal, Ace playing for Ayo Gurl of course suffers in all stats as it’s hard to get anything done when you’re fleeing for your life, but Harbleu leads the way and it is noticeable that the three non-european medics lead the heal rates. Something for Europe to ponder?

Danger loving Mirelin predictably leads the kills per minute with Zebbosai or Kaidus to latch on to and considering his risky style his KAD doesn’t take an excessive hit. Most of the medics who had a fragging machine to heal did OK, Vani with Smirre for example, but Shade didn’t seem to excessively benefit from Lansky showing some considerably discretion while KnOxXx also played carefully, and both those medics had significantly higher KADs than the rest.

Of course this is also largely down to the team’s ability to keep them alive and Shade had a considerable advantage in that area, whereas KnOxXx was a slippery French bastard © Admirable.

43

Ubers per minute are in the blue measured against the bottom axis and drops per minute are in yellow and measured against the top axis.

Epsilon’s KnOxXx manages to build ubers the fastest but is also the most guilty of dropping them, certainly amongst the top medics. Is old age starting to catch up to the right click reflexes? Harbleu comes closest but that’s at least partly down to some inspired work by Tek. Vani actually takes the award for the lowest rate of uber drops demonstrating the fastest trigger finger but at the sharp end of the competition Shade and bonobo did the best job of popping.

Bonobo and IM built uber the least effectively but I don’t have a single kritz on record for them so that may have something to do with it.

44

Predictably Mirelin went for the Kritz most often with Vani and Bulle playing hard for the tide turning Kritz too – perhaps Smirre’s considerable damage and frag numbers have their root here? Where Huhy’s went is a mystery.

KnOxXx also wasn’t afraid to Kritz but of course War’s once immense kritzing ability has long disappeared and this counts more as a handicap to Epsilon than a stat booster. Mixup rarely employed the Kritz which makes their uber building more impressive and lends some credence to the idea that their scout ratings were hit by both building uber and generating excessive healing numbers. This may also be where Harbleu’s lead in healing comes from.

The grand final effect

So just how much were Epsilon’s overall stats hurt by the way the final went and how much were Froyotech’s boosted? Certainly it seems that some Epsilon players were ranked far worse than the results their team managed to achieve. Well, the results aren’t earth shattering. None of the Froyotech players have a significant negative and in fact B4nny’s numbers improve if the final is taken out which might give you an idea of how important meeting enemy aggression was for his game.

Epsilon gain more than Froyotech lose, the main winners in terms of raw stats are Kiler, Bash and Mike but the main beneficiaries of discounting the final thrashing in terms of their positions in the rankings are War, Kiler and Mike. War would squeeze ahead of Duwatna and Mike would gain a small edge over Yuki in the soldier table. Kiler would gain a couple of spots in the scout rating table, but Bash would only get closer to the mid table.

For a single match it’s quite a big effect but grand final performance can’t be blamed for all of Epsilon’s statistical woes and realistically the Immunity log might have had more of an effect in terms of improving their overall lot by suppressing a near competitor.

Team of the tournament

So, the team of the tournament is a tricky one, for no other reason than Froyotech are either top or near top in all areas. Termo is the only player who clearly looks better than his Froyotech counterpart Duwatna (remember kids, we think we’ve adjusted for their relative roles in their teams and healing support), I certainly think the scouts Clockwork and B4nny as a pair are worth taking and Lansky was the man of steel, apparently indestructible posting an enormous KD along with leading all the aggression stats.

On roamer Blaze got the best basic stats but there was a relative lack of big play ability there in comparison to say Tek or A Seagull, but Tek had a hard time with his general stats whereas A Seagull got the best of both worlds being more of a constant presence, so I think I’d take him. Tek perhaps if I wanted the sick sniper and spy plays, but why would I when I can Caw Caw!

Medic is less clear cut, Shade was solid but unspectacular, made sure he got and looked after his Uber with care and mixed in a dash of Kritz for added spice. KnOxXx dropped too often by comparison, but with killing machine Lansky racking up huge kill rates I have questions about Shade’s assists per minute figure – support your man of steel. Obviously Harbleu brings the big plays, and to be honest if it meant I kept the double kill on Snakewater last I’d go with him. Mirelin was more exciting than most but wasn’t able to overcome the obvious risks of his play style to keep up KAD and uber build rates particularly with so much Kritz chosen.

So who shall we go with? The “who won” tiebreaker says Shade.

Leaving us with the horrific result of not a single European player on the team of the tournament. Rather than wail in self pity in our dark emo Eurotrash bedrooms, let’s have a look at the continental top performers.

War leads the Demo selection (sorry Perm) leading Kaidus by a small margin. Zebbosai fairly cleanly takes the pocket soldier role and the i49 finalist takes on Lansky head to head in style. Rising takes it on roamer, he’s played on and off at the top level and hopefully this i52 showing encourages him to strive to improve, and this was with significant physical injury.

On the scout front this may prove to be controversial (yeah, like anybody gives a shit) but I’d follow the stats and go with Smzi and Stark. There could be problems in that they both tend to pick and choose their moment rather than play an aggressive space creating style, but with War and Zebbo doing their best to overextend we need some people with brains to hold the team together.

And for Medic I think Mirelin is a fairly clear choice, as KnOxXx has developed a case of the drops just when he needed to keep it rock solid. There aren’t many medic stats that are just down to the medic themselves and some drops are unavoidable, but you have to look for a low drop rate.

Capable of competing with the NA/AUS mix? Maybe with the right map picks but generally speaking probably not, EU you have about a year to get a decent team together capable of getting up in the morning and calling a taxi in English, and doing things like knowing who your opponents best maps are. Go to it! Any Swedish super teams are also welcome.

So that’s it

It’s over already? Ah well, all good things come to an end and unfortunately both i52 (good thing) and this nonsense (open to question) are no exceptions. Hopefully this will stand as some kind of final testament to the men and women who made i52 a great tournament, and I may have recovered from the effort of writing this to consider doing another article when the next one comes around although by then surely the body won’t still be kicking?