The Forums Are Now Closed!

The content will remain as a historical reference, thank you.

Under rank 200 as a pugger... what I've learned about stats

By on July 24, 2009 12:33:43 PM from Demigod Forums Demigod Forums

Not long ago when Frogboy announced that accurate stat tracking is at the top of his to do list I balked, and when he said that custom games should no longer be ranked I was delighted. I was wrong on both counts.

Proper stat tracking seems pointless when you look at the top 100 and see that the majority have well over 90% win rates, when you see the rank 8 player joining "no premade" games with a premade, and in a way it really is.  The current system offers no insight into whether the rank 1, 10, or 20 players would win if some life or death emergency actually forced them to play each other.

However the stats are actually excellent at guaging skill level disparity at lower ranks.  In my experience a team full of rank 1,000 players will consistently beat one of rank 10,000 or lower players.  Rank 200 players do genuinely seem to win more consistently than rank 500 players.

The proof of that reliability is, in my mind, my own rank.  Now bear in mind my skill didn't improve very much if at all from when I was rank 400 or 500'ish, what actually changed is that I started looking up everyone's stats before hitting the ready button, to the chagrin of many a lobby.  I found that the majority of the time the stats genuinely reflected players' relative skill level. My improved rank reflects how well I can gauge the outcome of matches from the much-disparaged ranking system. Again that's not to say it's especially precise in that players within 100-200 ranks of each other won't always have distinguishable skill level differences, but nonetheless the stats are actually incredibly helpful for a pugger who's trying to find good games.

Now I can see the argument against custom game stats.  The idea is that players on the top end of the ladder have figured out the most effective way to rank up is to win by sheer volume against crappy teams.  The rank 1 player has 327 wins and 7 losses, which makes the ladder look awful.  But let's not put the blame on the wrong party, the purpose of a ladder system is to rank up and try to reach the top.  The real issue is something else I learned:

Game quality matters, but not enough. I've seen a lot of people with win ratios similar to mine who have much, much worse ranking after a comparable number of games. I believe this is because the majority of my losses are against fairly high ranked players and I lose virtaully no points for them, and a handful of my wins are also against high ranked players, and I receive more points for them than from usual games, in other words the ladder works better than we tend to give it credit for. So that leads me to conclude that the issue with pub stomping premades is not actually that their games are ranked, that the ladder exists, or any of those things, the issue is that they get any points whatsoever for it. 

There must currently be a minimum amount of ranking gained for winning, and that minimum needs to be removed completely. If the average rank of your opposing team isn't within 500 ranks of your own team, you should gain zero points for winning, which leads me back to Frogboy's point about accurate score tracking. I was initially nonplussed about FB's emphasis on tracking, but now I agree it's actually a pretty critical improvement, so long as the ranking requirements are tightened enough to actually take advantage of it.

Let's be honest, when ladder ranks are determined by who had 300 pub stomps vs. 305 accurate tracking isn't especially compelling, because the ranks it correlates to are meaningless.  However if ranking is dependent on a few critical matches between high ranked players then accurate stat tracking is quite important.

The process of ranking up has been very enjoyable for me, and not because the number by my name on the pantheon website is important to me.  It's been fun because the current ranking system works well enough for me to match myself into fun games, and I really think that it could work for other players too if only they could see the stats in game. When I play this game on my brother's computer which is ridiculously better than mine when I tab to check stats I have to worry about crashes and I'm sure most people do too. On my own computer the game is inexplicably rock solid and I can afford to check every player every game even if I'm too dumb to remember their names and have to tab out half a dozen times to get them right.

So, the synopsis if this is TLDR:

Frogboy was right about stat tracking accuracy and he was wrong about removing custom ranks completely.  Instead the rankings need to be tightened up in conjunction with stat tracking improvements, and every player needs a reliable way to check stats.

As a long-term goal removing custom game rankings make sense, but only when team auto-matching is impelmented.

+37 Karma | 27 Replies
July 24, 2009 1:07:28 PM from Demigod Forums Demigod Forums

I agree.  Stats tell enough to be useful and should be shown in game.  I don't see any reason to stop ranking custom games even with matchmaking, just make sure all stats/rankings are separate and independent by type.

July 24, 2009 1:18:28 PM from Demigod Forums Demigod Forums

I don't see any reason to stop ranking custom games even with matchmaking, just make sure all stats/rankings are separate and independent by type.
Here's some reasons why, as a long-term goal which coincides with team mathcmaking or comes afterwards, I think custom game rankings should be removed:

  1. The top end of custom game ranking is not about competition and likely never will be
  2. There are too many types of custom games for a coherent overall rank. This will become more pronounced when modding support is added
  3. A better functioning auto-matching system won't be instantly popular. If custom games are ranked many people won't switch. Bottom feeders will stick around to be the new king of the hill if competitive players move on to automatching
  4. Zero map rotation enforcement, aka Big Game Hunters Syndrome. This will become more pronounced if more maps are added to the game
  5. Team stacking
  6. Game stacking, as in sneaking in wacky settings like regulus hosts (it's always regulus) picking high towers, players with heaven's wrath putting in high reinforcements, etc.
  7. The game community just isn't that big. Consolodation is important
July 24, 2009 3:57:55 PM from Demigod Forums Demigod Forums

Good post, and Good research.

It's been mentoned that we need stats in game, but just to clarify the premade issue a little more, perhaps we could get a simple button click on each player that will pop up a small overlay window of thier stat breakdown. Including:

Rank, XP, Heroes used, teamates and win/loss for the last say.. 3 games.  If we get a tab on each players page showing these items, the pop up can link to and center on that info.

Should be a solution that Frogboy can have SD work on.

July 24, 2009 4:01:38 PM from Demigod Forums Demigod Forums

Quoting Obscenitor,
Zero map rotation enforcement, aka Big Game Hunters Syndrome. This will become more pronounced if more maps are added to the game
Mmmm, great point.

And yeah, making the matchmaking more popular for competitive play has got to be the goal, so that makes sense too.

July 24, 2009 4:16:52 PM from Demigod Forums Demigod Forums

I think that there is an asymmetry in stats:

Highly ranked players are rarely terrible players.  They are at least decent and usually pretty good.

Poorly ranked players (meaning bad rank but at least 50/50 wins) can be either good or bad.  I've seen some horrible players around there, but also some very good players.

So a highly ranked player is probably decent to good, but a poorly ranked player is not necessarily a poor player.

Of course, horribly ranked players (bad rank and much less than 50/50 win %) are almost always bad.

So yeah, there is predictive power in stats, but its not as straightforward as you might think.

A key thing about stats is that they arent dominated by how good a player is.  At one point I was joining games without checking teammate stats and usually as one of the last 1 or 2 people to join.  This meant that I was playing with a lot of really bad players against a lot of premades.  I was winning probably around 40%.  But when I am more careful about joining (meaning I'm one of the first people to join a game basically and check to make sure that my teammates arent completely horrible) I win substantially over 60% as a PUG only guy.  That jump in winning percentage wasnt me getting better, but me choosing games more carefully.

There is other stuff like that.  For example, if you host a lot of games you should win a lot more games.  I'm not saying that it should be because you are dropping to prevent stats, but because high ranked opponent hosts cant drop to save stats.  I find that a pretty high percentage of the time when I am playing a team full of highly ranked players, the host on their team drops if I am winning.

July 24, 2009 5:41:17 PM from Demigod Forums Demigod Forums

So yeah, there is predictive power in stats, but its not as straightforward as you might think.

A key thing about stats is that they arent dominated by how good a player is.

I'm not trying to say that I can, as a third party, predict the outcome of a match by players' stats (except in the most obvious cases).  I'm saying that the stats, even as-is, are sufficient to enable players to pit temselves against challenging but not unbeatable opponents. 

In other words any player with moderate experience would do well to use the stats to make the game more fun for themselves, and rather than remove custom games from rankings completely before premade automatching is implemented I think the custom rankings should remain in place as they're a powerful albeit imperfect tool at the moment.

In-game stats would be a massive improvement and would go a long way towards rectifying the premade issues, as premades would have much harder times finding unsuspecting victims and they may have an easier time finding each other because of it.

July 24, 2009 6:10:54 PM from Demigod Forums Demigod Forums

Oh yes, there is not question about that.  Already I use stats all the time and its annoying to have to alt tab all the time.

But there are still many situations in which stats can be very misleading due to either known bugs or the fact that the ranking system used for the ladder still doesnt seem to completely make sense to me.  I mean, take StAcK3D_ActR.  He is winning over 60% of his games, but some guy who is 0-2 with 2 ragequits is ranked higher than him.  I'm a little hesitant to have stats out that consistently rank ragequit losing against highly ranked players as better than beating mediocre players.  So while stats do good things, the fact that the ranking can be EXTREMELY misleading will lead to a new set of problems too.

July 24, 2009 6:14:53 PM from Demigod Forums Demigod Forums

ive given up on the idea that a serious competitive system can be implemented in Demigod without totally overhauling the ranking system. the stat tracking is both inaccurate and far too easily spoofed. its way too possible to game the system so as to avoid being fairly ranked based on games against real competition. 

 

in real competitive gaming scenes a player cannot choose his own matches. he is matched by a tournament organizer with a player of similar ability. this was supposed to be the service provided by Pantheon, but it has totally failed because Pantheon doesn't work very well. 

 

what would have to be done is to not track stats for custom games at all. they would be purely for fun/practice. only Pantheon/Skirmish matches would be ranked. this could only be acceptable if persistent teams and team ranks were included as part of the service. it would also require complete elimination of AI substitutes for dropped players. and it would also, and perhaps most importantly, require weighted point ranking system so that you gain more points for defeating a higher ranked team and actually LOSE points if you take a loss to a lower ranked team. 

July 24, 2009 7:32:32 PM from Demigod Forums Demigod Forums

As a long-term goal removing custom game rankings make sense, but only when team auto-matching is impelmented.

Weren't you arguing against this before? I only read the TL:DR part.

Not trash talking, I'm actually surprised you changed your perspective on it. Glad we can agree on it.

July 24, 2009 7:56:40 PM from Demigod Forums Demigod Forums

Quoting AstarothLBC,

As a long-term goal removing custom game rankings make sense, but only when team auto-matching is impelmented.


Weren't you arguing against this before? I only read the TL:DR part.

Not trash talking, I'm actually surprised you changed your perspective on it. Glad we can agree on it.
I would say that I've changed my mind in that I accept rankings as a more effective matching tool than I did before. My big conditional statement though is that I think the ladder rankings need to be visible in-game though so that they're a useful tool for everyone, not just the small portion of the playerbase who tabs out to check stats every game.

July 24, 2009 8:20:33 PM from Demigod Forums Demigod Forums

3 Solid Ideas for Stardock:

1. Get rid of host quitting exploit which erases stats (priority)

2. Change the interface in lobby so we can see the stats (alt tabbing causes frequent crashing)

3. Give hosts a way to filter players based on stats. ie Win pct, Minutes Played, Number of Disconnects

July 24, 2009 9:36:49 PM from Demigod Forums Demigod Forums

Quoting Obscenitor,


I think the ladder rankings need to be visible in-game though so that they're a useful tool for everyone, not just the small portion of the playerbase who tabs out to check stats every game.

I don't think anyway can argue with the need for this.

July 24, 2009 10:47:23 PM from Demigod Forums Demigod Forums

Cherry picking ranked games based on other players stats = FAIL.

July 24, 2009 11:30:09 PM from Demigod Forums Demigod Forums

Quoting Fearzone,
Cherry picking ranked games based on other players stats = FAIL.

July 24, 2009 11:51:01 PM from Demigod Forums Demigod Forums

Quoting Fearzone,
Cherry picking ranked games based on other players stats = FAIL.

Commenting on player habits in online games with 0 online experience = EPIC FAIL?

July 25, 2009 1:32:38 AM from Demigod Forums Demigod Forums

Frogboy was right about stat tracking accuracy and he was wrong about removing custom ranks completely.  Instead the rankings need to be tightened up in conjunction with stat tracking improvements, and every player needs a reliable way to check stats.

As a long-term goal removing custom game rankings make sense, but only when team auto-matching is impelmented.

 

Tottaly agree here. Custom games should be the place to go and "fool around" Not climbing to the top of the ladder. Skirmish and Pantheon needs improved so they become the place to go for evenly matched competative play and rankings. Keep the ranks in custom for now, but once Pantheon and Skirmish matching is improved, dump it. It would be nice to see people in the top 10 that actaully play the tournaments.

 

My big conditional statement though is that I think the ladder rankings need to be visible in-game though so that they're a useful tool for everyone, not just the small portion of the playerbase who tabs out to check stats every game.

This would also help against "unbalnced" pre-made situations. You join a game with 3 guys in the top 100 on one team, and the other has guys in the top...10,000, you could atleast atempt to balance the teams based on the ranks shown. People will then know what they are getting themselves into instantly instead of having to do the alt+tab search over and over.

July 25, 2009 2:26:52 AM from Demigod Forums Demigod Forums

Quoting Colonel_Jessep,

Quoting Fearzone,
reply 13
Cherry picking ranked games based on other players stats = FAIL.

Commenting on player habits in online games with 0 online experience = EPIC FAIL?

ha!
Seriously though, great post Obscenitor.  Thanks for taking the time to write it up.   for you!  

July 25, 2009 4:10:46 AM from Demigod Forums Demigod Forums

This would also help against "unbalnced" pre-made situations. You join a game with 3 guys in the top 100 on one team, and the other has guys in the top...10,000, you could atleast atempt to balance the teams based on the ranks shown. People will then know what they are getting themselves into instantly instead of having to do the alt+tab search over and over.
I'd thought about this quite a bit. The premade vs. pug thing right now is kind of a layered problem because not only does the premade have the advantage of coordination, the pug team's chances are further diminished by any relatively experienced player seeing identifying the premade jumping ship, virtually guaranteeing the most lopsided fight possible.

So what would happen if the average player could see premades for what they are? Would premade wait times scale up dramatically and drive off the 'competitive' crowd completely or would the decline in newbies to stomp actually make it easier for premades to find each other since they wouldn't be preoccupied with other opponents? I have no idea.

July 25, 2009 8:06:37 AM from Demigod Forums Demigod Forums

I enjoy playing against premades, provides more of a challenge ^^

btw Obscenitor, your rook build is full of win, I tried it out today and it was pretty good fun, orb of defiance and all
i think i went towers -> arrows -> transfer -> towers -> stun -> light tower -> towers -> siege hat -> save -> stun+towers -> hammer hammer hammer hammer hammer
it seems kinda unstoppable

July 25, 2009 2:55:43 PM from Demigod Forums Demigod Forums

That particular permutation might struggle against minion swarmers depending on how aggressive/good they are. Still though it's a super fun build and the orb is amazing. What do you think of the early siege hat? People knock it, but I like it quite a bit. It's 500 a pop on buildings and 200+ on low armored DGs, I think it's really underrated.

Also beware of trebuchets.  A pure tower build really, really struggles with them on fortress mode. I had a crucible game where the guy got them early and the things were literally killing me, I was taking 200'ish damage every 3-4 seconds from off the map, it easily outpaced my regen and when I finally out of their absurd range they devestated my towers.

July 25, 2009 5:30:48 PM from Demigod Forums Demigod Forums

LOL Obscenitor, I've played enough Pantheon games with you to know that shouting "Here is a friendly hint to avoid dying, stop standing on the FUCKING fire!" is mostly pointless to my teammates who end up in that situation in the first place.

That's not to say you aren't a really good Fire TB (best I've played so far in Pantheon)... but playing TB on forces of light in Pantheon by exploiting a UI bug is cheatXoring.... a little.

So here is a set of questions I have to you, to see if you aren't skewing your rank a little bit in another area as well... When you are playing Pantheon, and you look at connection info... if you see someone on you team is terribly bad (from memory), do you quit? If you see someone is terribly bad on the other side (from memory), do you quit?

Just asking....

July 25, 2009 7:23:53 PM from Demigod Forums Demigod Forums

Nah, pantheon/skirmish are a different beast entirely. I usually get a drink, make some food, or browse the internet while I wait on connections and my memory isn't as reliable on bad players as it is on good ones.

Also it takes so absurdly long to get games going at times that it wouldn't be worth the effort.

 

As for being a light TB, I initially found out how it works when I got dark rook on accident...  that was funny at first but when I ended up with rook and rook vs. rook and some other guy the humor faded a bit.

*edit* also as you can see from my pantheon rank, I don't play that bracket competitively. It just takes too long to get reliable games.

July 27, 2009 2:31:13 AM from Stardock Forums Stardock Forums

It's good to see another decently ranked Pantheon player. I was wondering if I was the only one, since all the high rankers are merely pre-mades.

July 27, 2009 2:43:02 AM from Demigod Forums Demigod Forums

Quoting xclavex,
It's good to see another decently ranked Pantheon player. I was wondering if I was the only one, since all the high rankers are merely pre-mades.

Pantheon is a joke these days. I'm 19-2(and that's because the same teammate quit on me twice). Forces of Light are getting crushed; they always seem to get AI teammates. I've seen entire teams of AI in a row.

July 27, 2009 3:02:44 AM from Stardock Forums Stardock Forums

Quoting Shade,

Quoting xclavex, reply 23It's good to see another decently ranked Pantheon player. I was wondering if I was the only one, since all the high rankers are merely pre-mades.
Pantheon is a joke these days. I'm 19-2(and that's because the same teammate quit on me twice). Forces of Light are getting crushed; they always seem to get AI teammates. I've seen entire teams of AI in a row.

 

Doesn't matter. I still prefer the randomness. It's like gambling, you never know what you'll get.

Stardock Forums v1.0.0.0    #108435  walnut2   Server Load Time: 00:00:00.0000687   Page Render Time:

Stardock Magazine | Register | Online Privacy Policy | Terms of Use

Copyright ?? 2012 Stardock Entertainment and Gas Powered Games. Demigod is a trademark of Gas Powered Games. All rights reserved. All other trademarks and copyrights are the properties of their respective owners. Windows, the Windows Vista Start button and Xbox 360 are trademarks of the Microsoft group of companies, and 'Games for Windows' and the Windows Vista Start button logo are used under license from Microsoft. ?? 2012 Advanced Micro Devices, Inc. All rights reserved. AMD, the AMD Arrow logo and combinations thereof are trademarks of Advanced Micro Devices, Inc.