Stats such as these are actually MORE significant on average than they appear, not less. For example, due to premade teams being significantly more effective pugs, and the number of prem vs pug games, there is actually a strong mellowing agent there.  That is, no matter what the Prem plays, they will likely win.  All such games add to the noise of a report like this.  If there are still spikey points in the 60% range, then remember it is getting past this noise already.

maybe all premades use sedna + UB + erebus

It's possible, but why do you think they choose those? Thing is, you can't actually factor out player preference from skill.  If only the best players use sedna, it is probably because she is quite awesome.

The odd turnaround of this, is that it means that the 'best' players may also defend sedna, because she is their pet. (This of course applies to all of the high usage DG's.).

No doubt its a strong combination. 

This is what I was saying earlier. How much are UB/LE/Sed's win rates a function of their raw power? How much are they a function of the best players picking them because of their raw power? We really need a large pool of premade vs premade replays to analyze in order to figure out whats what (something we obviously don't have atm). 

I doubt many pre's would be comfortable going into a pre vs pre match without a Sedna player, whereas pug groups frequently play without one. I'd guess teams without a Sedna lose to teams with one in the majority of cases.

The other two you can change up without big issues, UB for Oak and LE for Reg would provide a different playstyle without being any weaker for the majority of the game, though in my opinion there is no better character than UB for lategame.

Inexperienced players also take a lot longer to finish games. They can't knock the citadel down in 15 minutes because they are either too aggressive (and die) or not aggressive enough. This could mean the late game strength of UB is exaggerated in the stats.

Or the other team is competant enough to not die and feed them thus prolonging the game.

The games I have been in as UB and most other Demis, don't end before level 15.  This isn't because of a lack of trying on anyone's part.  Its just that its more difficult to acheive wins v good players. 

i beg to to differ

me and my premade friends often do 3on3 or 2on2 matches within our group (so skill is evenly matched)

and in my experience UB/Reg/Queen can own Ereb/Torch/Sedna(+friggin High Priests)

just one example from last night... my point is that you don't need sedna in a prem. vs. prem. situation

Sedna's effectiveness is inversely proportional to the number of players in a game, as she has no AoE damage, and is only capable of keeping one target up at a time with healing. As such, she can rip faces off in 1v1 situations, where AoE matters much less, but tends to suffer in groups, where a TB can deal 2400 damage with a single RoI, compared to her 1500 heal. This is, in fact, where TB and Slam Rook are at their zenith, so it could be the low player counts in games resulting in a skewed win/loss ratio for many DG's.

Speaking from experience, you can only be as good as the rest of your team. Theres really no easy picks for a Demigod that will turn a poor player into a unstoppable killingmachine.

One thing ive noticed though, if you try all the Demigods a couple games each, it REALLY improves your survival chances in overall. Not to mention it makes it easier for you to work together with the rest of your team, knowing everyones strong points and weak sides.

Is queen really that differant form sedna. Queen just also gets a lot of crappy players and doens't do the damage. But you made up with that for UB and Regulus who is 4th demigod if i'm not mistaken mostly because of mines.

This is what i've done. I can play all of the demigods evenly (except sedna) and i notice a trend. My UB and erebus games almost always are a win while my QoT games are almost always a lose. Now sometimes i'll lose an erebus game or win a QoT game but this is in public games so it pratially is my teammates. My best demigod is UB and i've played less than 10 games with him and over 30 with torchbearer. Is this coincidence?

UB is faster than TB
UB does more damage than TB
UB's abilities do more damage than TB's abilities (lvl 10 ability does more dmg than lvl 15 ability)
UB has more HP than TB

It's a joke really.

I'm not sure on the relative power of the DG's, but some DG's are less complicated to play than others.

There are really no gotchas with UB, it's fairly straightforward to play. Sedna, focus on not dying, don't get Yetis, and you'll be decent enough. Lord Erebus, not as simple as the other two, but can do pretty much everything, and in the hands of a good plyer it will be very hard to kill.

The rest of the DG's, with maybe the exception on Regulus (also fairly straightforward to play, as long as people remember to invest on Mines), are more tricky. Fire TB requires careful play, Ice TB is good in the hands of a good player. QoT, needs a good player to be worth it, same for Rook, a bad Rook player just dies a lot. And Oak, can also be good, but again, requires experience.

TBs abilities are by and large all ranged to one degree or another.  UBs are not.

TBs abilities are by and large AOE damage abilities and therefor to MORE damage on multiple targets. (Single target attacks for TB are 2 for UB its everything by ooze and plague) 

TB has TWO perfectly wonderous interupts one of which does more damage than the UBs single interupt. AND its ranged.

The second is an AOE that puts all the creeps on hold.

It's a joke really.

I guess I'd argue even in a 2v2 or a 3v3 confrontation focus firing on a single opponent is better than damaging all of them slightly in a very large number of cases, even when theres a Sedna that can only heal a single target every cooldown. Getting one member of a team of three to the point where they have to retreat represents a victory for a lot of situations.

Not even close to damaging only slightly for most of the AOE abilities in this game...Focusing a single target UB can spit and hit with ooze for 650 off the bat?, 150ish is ooze, QoT does 650ish to all three, TB does 850ish to all three,  you've done ~2400 hps to a single target who is also taking AA and likely another closely followed volley of AoE goodness in spike wave for 600 to all three and more ooze ticks at 150 a piece, plus a shatter for another 1300 in under 3 seconds ..

AoEs numbers may not be OMG huge but they sure do add up...

Think about this for a second: if you took a Fire TB, a thorns QoT, a slam rook, and a... I dunno, pick two other AoE'ers, and then  put them up against spit UB's, you could expect the UB's to kill one, maybe two of them, before being utterly destroyed.

Uh... no. 

Sure, they will all bunch up waiting for you to rain fiery death on them. Or they use their superior power/speed, to control the map.

These same characters also happen to have effective enough lane control and tower destruction that I doubt this would be a problem. Anyways, the point I'm trying to make isn't UB's versus Rook/whatever. The point is the (often lost) potential of AoE builds.

A much better team against five spit UBs would be Sedna, Erebus, Oak, another UB, and an Ice TB.