Gentleman Jon — It already seems like an irrelevance with the sudden onset of Season 18 but I’m doing the Season 17 Prem stats review anyway, to commemorate the brief existence of a non-Epsilon world if nothing else. We’ll have the usual array of barely supportable hyperbole and the shocking revelation of who is the best demoman of Season 17. Well, there’s a joke in there and the joke of the statistical truth as well (which is as good as a lie of course), which faux philosophical rambling leads me to….
What do you do with a problem like Michael?
We all know he’s a ridiculously good player, and an exceptional performance as soldier, particularly as pocket soldier, is hardly a surprise. But when Mike plays 3 games as Demo and posts absurd stats what’s an analyst to do? It’s fair to say his demo style was atypical but seeing as two of the three were posted against (a pre KnOxXx) TLR and AWS we’re not really talking about the soft option.
Suffice to say that in this season without the enormous dominance of a single team and some genuinely close playoff games some top player’s stats suffered, and some player’s statistical achievements should be taken with a pinch of salt.
Logs against the dropped team aren’t used (in contrast to the TFTV stats) which prevents some players getting the chance to buff their stats that others didn’t. There’s a time limit for inclusion that amounts to roughly a couple of full matches.
Demomen
We might as well get this out of the way straight away
The graph shows damage per minute in blue, K/D ratio in red and the player rating in yellow. The K/D is measured against the top axis while DPM and rating against the bottom axis.
So what can you say? It’s only 3 matches, but Mike got the highest DPM of any demo and only had a lower KD than Sickless. He didn’t have to actually face off against Kaidus so didn’t get the treatment most of the others did of being locked out of any middle they played him on before they arrived, and he got to fill his boots against Lego. Even so he’s only a whisker above Kaidus who actually went the proper route of playing all the opponents (including KnOxXx led TLR twice). Also only Hys got a similar amount of heals.
Hys got very close to Mike’s DPM, and seeing as he had to face AWS and MiG twice it’s fair to say that it’s a harder fought total. In the heat of playoffs though his KD definitely suffered as a prime target on his team. In recent seasons where Epsilon just crushed everyone the differences caused by the playoff rounds in final season stats didn’t seem to make much difference, but here the added burden of the playoff games has supressed Hys’s (and to an extent Kaidus’s) stats as they were better placed at the end of the regular season.
Sickless is also a surprise here and appears to have caused some consternation amongst my stat viewing public, and like Mike his position is based on a small number of appearances, in this case against Dotter, Lego and early season MiG which isn’t the toughest schedule. At a stretch it could be considered representative. It’s unusual for a demo to have such a clear KD over DPM like his, which hints at a sort of cleanup style of play although of course wiping the occasional team stacking a point with a crit sticky helps. This is also seen with Ryb who seemed to hang back noticeably this season.
After a promising TFTV Invitational Dr. Phil seems to have had a bit of a nightmare season.
This illustrates the relationship between damage and kills for the individual players. The yellow bar is the average damage they get for every landed shot and is measured against the bottom axis. The blue bar is the amount of damage they had to cause to register a kill or assist and is also measured against the bottom axis. The red bar is the number of kills achieved per minute, which is measured by the top axis.
A chink in Mike’s armour appears here, getting a slow kill rate, matched by a very high amount of damage required to register a kill or assist. Evidently he piled on the damage without regard for finishing the kills (or he wasn’t blessed with a team cleaning up for him) although his KD was pretty good so it wasn’t suicidal aggression.
In total contrast is Sickless who is very strongly kill focused getting the fastest kill rate while taking the least damage to register a kill although of course he didn’t have the toughest targets. I’ve heard this kind of clean up style described as baiting your team for stats. Whether this unkind moniker applies here is unclear, but somebody has to finish those frags. Listening for calls might explain rumours of him playing in almost total silence.
In contrast to Sickless and Ryb, Exon cannot be considered a clean-up demo posting a very high kill rate by getting in people’s faces.
Lego’s passive style left Dr. Phil with a very slow kill rate but even so he had to do a lot of damage when he was fighting to register a kill or assist.
Hys and Kaidus registered more normal looking stats, with Kaidus looking perhaps a bit more death-matchy.
The ultimate test of Demo skill
As anybody with any understanding of TF2 knows the ultimate test of demo skill is hitting scouts with pipes.
And there you were thinking Dr Phil had a bad season, whereas in actual fact he takes the Season 17 statistical demo crown. Proof that Mike is actually rubbish on Demo regardless of any other numeric frippery.
Soldiers
So, putting Mike’s demo performances behind us move onto soldier where we’re on slightly more solid ground and we can see how Mike got on here as well.
This kid. Mike didn’t play a full season for MIG either, but the stats were racked up again TLR, AWS and AYO, so the 2nd best team, the 3rd best team and a mid-table team although that’s partially based on him playing demo for them… Ultimately we all know he can put up these numbers so there’s not much point denying his soldier performance any legitimacy.
Apart from the illustrated differences above KnOxXx went after high value targets more than Mike getting a higher proportion of his kills against both medic and demo. Although some players dedicated more of their targeting to one or the other no one else combined attacking both so frequently.
The top 3 pocket soldiers come from the top 3 teams which is no surprise really, although Tek is part of a little group around 3rd with Klar and Zebbosai very close. Like everyone on Fnx, Klar didn’t play a full season and the strongest team he faced was a pre-KnOxXx TLR so perhaps his position is a little overstated, and his rating does benefit from getting a lot of medic kills, an even larger share of his kills than KnOxXx.
Zebbosai did well in a team that didn’t meet expectations although once again he didn’t have to face the top teams with their strongest lineups. He has the most aggressive profile, pushing for engagements at the expense of his KD.
Stark suffers a bit here having played against MIG twice while taking up Merc duties but even so it was a bit of a lacklustre season for him.
One interesting point is the illustration of the level of statistical upgrade Mike and KnOxXx represent over Rising and ipz, although particularly in Ipz’s case there’s a loss of calling and vocal team leadership, whereas TLR gained a large slice of that.
Gear forces his way in amongst the pocket soldiers again, the only roamer who does so, with only Mike and KnOxXx putting up clearly more damage. Next best in the roamer charts is Kalho whose perennial use of the black box led to him playing a more defensive risk averse roamer style. From his high KD to DPM ratio we can see a rare thing indeed, the clean-up roamer.
There’s not a great deal to say about the other roamers, T-Mac is the best of the rest with his high risk/reward style where he backs himself to pull off any play he can imagine.
Somewhat surprising here is Tek’s relatively slow kill rate amongst his peers at the top of the soldier charts, although with pretty efficient use of his damage perhaps it’s indicative of unusually tight support play clearing up his damage efficiently – Exon had a high kill rate for example.
Klar is actually the most efficient damager although once again this could be a result of slightly weaker opposition
Once again we can see that Rising’s and Ipz’s kill rates are well behind their replacements although both had similar efficiency of damage producing kills/assists which may indicate being a bit more careful about which engagements to take on.
As to be expected the roaming soldiers suffered the most with damage wastage with Crizzl and Sideshow the most egregious perpetrators. Gear of course posted stats that would look more at home on a pocket soldier.
Scouts
Unlike classic Epsilon season of the recent past there was some actual competition amongst the scouts this season, although it may prove to be somewhat illusory on further examination.
There’s quite a few cleanup scouts at the top of the rankings this season, and Smzii is foremost amongst them sacrificing a bit of DPM in favour of an enormous KD score. Some might say he was baiting his team for stats, others might say that he was listening to his team’s calls – team comms certainly showed he was active in trying to improve them.
Gf18’s lofty position is a bit false, he played against Lego, Ipz-edition MiG and Dotter so didn’t face the toughest ride, but a lesser scout still wouldn’t have posted such good stats. However Bash played 8 games including TLR twice so even though he’s slightly behind statistically I don’t think there’s any comparison in terms of their achievements this season and Bash is realistically the top aggro scout.
Kiler4fun on the other hand played against the whole gamut despite infamously missing the final (now you’re in Epsilon you’ll need to clear your schedule young man). Essentially his stats make him a mini Smzii.
Sheepy is fnx’s other aggro scout and also benefits from fnx’s revolving door policy on lineups. His 4 games played are another difficult to interpret track record, facing both TLR and MIG although they were the early season versions but also Lego and Dotter who both struggled. We know from last season he can put up good numbers so I certainly wouldn’t dismiss them but a pinch of salt is required.
Flippy rounds out the prevalence of the top clean up scouts. Dotter’s scout pair seem to be pretty aggro based in both cases with Beavern leaning very slightly towards a better KD, and the bottom of the listings are dominated by aggro scouts that couldn’t maintain a competitive KD after initiating combat.
Unsurprisingly Smzii has the lowest damage/kill assist efficiency, and his kill rate must be partially credited to the rest of his team racking up the initial damage. The Dotter scouts built up very high kill rates and Beavern had Smzii level damage efficiency too. Unfortunately they found themselves being cleaned up almost as often as when they were doing the cleaning.
Gf18 matches the elite scouts kill rates against the weakened schedule he faced, and along with Beavern set the highest mark for damage per landed shot, dealing out the meatshots. Sheepy by contrast had to do a lot more damage for his kills.
At the other end of the scale helberg and zen labour for kills doing a lot of damage for no reward, and helberg has a very slow kill rate.
Sniping
Unsurprisingly Flippy is the top sniper again, with Rake making the most creditable challenge. Beavern and Smzii are notably big threats to the uber regardless of their kill totals with Smzii famously hitting the Granary headshot in the finals. It’s an interesting difference between using the sniper as a general class and using it as a game changing class seeking to swing a key mechanic in one teams favour.
Medics
In this graph the blue bar shows the kills and assists per death and is measured against the bottom axis, the red shows healing per minute and is measured against the top axis and the yellow shows assists per minute also measured against the bottom axis.
Seeing as MIG’s players had a slight kill rate superiority over their TLR counterparts it’s interesting to see that Mirelin has a small edge in assists per minute and KAD over Byte. Byte is no shrinking violet but it appears Mirelin was the stronger battle medic.
There are other medics posting stronger assist per minute with similar KAD stats, but they seem to have paid with lower overall healing output.
As you’d expect from Lego’s overall passive style F2 is low on the assists rate but it didn’t help with the KAD ratio. Bulle was also relatively conservative in terms of facing danger.
The payoff for Byte’s relative safety compared to Mirelin is illustrated here with lower drop rates and slightly more ubers.
Bottom dropper is Skeej which considering he appeared to be healing in combat a lot is creditable. F2 and Bulle also reap the rewards of some conservative play, F2 probably held his ubers a lot in a conservative style so Bulle as the faster uber rate.
Thun is the top dropper by some considerable distance, and his successful uber rate is largely down to a liking for the fast charging Kritz.
Overall KOS has a nice combination of an acceptable drop rate, uber rate, assists per minute, KAD and healing. He’s the middle of the road prem medic for Season 17.
Medic Killers
This is the part of the review that sends chills down Sideshow’s spine as I reveal which players relentlessly brutalised medics on the field of battle.
KnOxXx leads the way in ignoring all conventions of civilised warfare, but with a stat that may surprise many Exon comes in 2nd, not only getting a very high percentage of medic kills, but apparently using a combination of devious traps and bombing from the stratosphere to drop their ubers as well.
Only his team mate Flippy was more prolific in dropping uber, Awsomniac considering it unsporting to prevent the medic getting uber when you can just take it away from him instead.
Hys is also a bit of a surprising Medic killer as he doesn’t really have the reputation as a dynamic Demoman, but presumably on the occasions KnOxXx found himself in the spawn queue when a medic needed chasing down Hys was dispatched to do the job.
3 of the next 4 places on the list are taken by MiG players with ipz and Gear combining and Kaidus following up to set a strong team representation without the same scale of heavily invested individuals of other teams. Interestingly Mike doesn’t appear in this cream of the crop list, and is in fact well off the bottom. He’s got a healthy respect for doctors, or he didn’t get many chances playing Demo for Ayo.
All of the top 8 places are taken by playoff team members, and the vast majority of the list is dominated by them, reflecting that the medic usually dies when the rest of his team is killed first, and successful teams are far more likely to be in that position. Klar and Sickless represent Fnx and Stark is there for Ayo.
So, statistical team of the season then
This used to be easy, bring back Epsilon!
For demo the stats tell me I should be going with Mike. Fortunately seeing as he’ll also be soldier, I actually have to go with Kaidus who was the real top demo. Statistics work.
For soldier it’s the perennial Mike on pocket and perennial Gear on roamer. With Gear looking like he’s bowing out for a season we might actually get a change soon for any roamers who want to step it up… oh hold on, Mike’s playing roamer.
Looking at the scouts, Smzii is an easy pick for the defensive clean up guy raising this to an art form. Based on just the raw stats gf18 should be my aggressive instigator, but really you have to take into account his soft schedule and go with Bash although of course this means we’re starting to converge on an MiG lineup. If you want to be bold go with the upstart.
On Medic the drop rate is a bit of a concern with Mirelin, but with Byte you do have to accept him running the team comms. If you want dependability Bulle is probably a good choice, but I think Mirelin is probably the most interesting pick as we already have a bunch of MIG players all over the team anyway.
So, it’s like Epsilon anyway, but with 2 different players. You can’t keep a good team down I guess.
These stats seem a bit messed up
It’s true to say that the stats this season were a moving target with the usual Fnx chaotic lineups and the top teams shifting major personnel as well, resulting in some unusual results that required explanation. To clear this up a set of stats that takes into account the quality of the opposing rosters would be needed that gave a moving rating based on the performances in relation to the strength of opposition.
The problem is that most players are barely playing 6 games or more so it’s hard to come up with something that adapts fast enough to have meaning. It may well be a subject for future study, although I’m reluctant to undertake an entire complex rating system to satisfy the mysteries of Skeej’s roster management style.
Originally published on VanillaTV (vanillatf2.org) on 2014-04-29 by diemasher.
Recovered from the Internet Archive (archived copy) as part of tf2esports.com's preservation of VanillaTV's TF2 coverage. The original site is no longer online. All rights remain with the original author; reposted for historical reference.