1350 + Any clan Mono, And SOA (Roots) Half decks. Of course it changes every week, but right now Roots are dominant.

5/3 splits

I do not understand your point behind them. Seriously, take it from a player who has made consecutive tops in ELO. They do not work as well as 4/4 splits.

I havent been able to play too much this week but its obvious that root half decks are tier1 this week and banger half decks aswell(any time Beeboy is usable...Bangers are t1 lol).
5/3 decks are overrated...its all about 3/3/2 hahaha lol

1000-1100: "Yeah, I know Timber has 6 power, but he's really pretty good!" At this point in your UR "career," you probably have 1 or 2 "unplayable" cards in your ELO deck.
1200+: Player understands that Fury does NOT help win a round, begins to grasp basic synergies. As the player is starting to recognize OP cards, decks begin to be built around them. In fact, there's a good chance that this player's deck relies too much on an OP card or two. RIP, Askai!
1300+: Player understands more advanced synergies. Decks often feature some kind of theme and may sometimes be built to counter popular setups that week.
1400+: Your deck doesn't fold over and die when facing SoA, SoB, an attack manip clan, or heavy DR. If it features two clans, you can still win with many of your 3/1 splits. In addition, it is often built to shut down an opponent's strategy (typically with SoA or DR) or to threaten a 3-round win.
1500+: Same as above, but your cards are covered in little flecks of virtual sweat. You own one or more voodoo doll with another high-rated player's name written on it. It is considered completely appropriate to time out if a lower-rated player beats you at this point. You insist that "ELO is pure luck" despite the fact that your own records clearly demonstrate otherwise.
1600+: You have done it. You have surpassed the gods. In retrospect, you have no idea what the heck you were trying to prove. You refer to your favorite deck as "my precioussssssss!"
1700+: ???

1000-1200 you cruise on easy victories by players new to ELO who don't know what they are doing and make all sorts of strategic mistakes or win by counting on drawing big bombs or win by sticking to an easy system (such as Rescue). By the time you approach 1300, you have to work for every victory so really what you want from your preset are "nuts" hands that transcend playskill, or characters that are tricky, that even experienced players make mistakes playing against. Above 1300 you want hands that can consistently draw at least 2 hard hitters and/ or some topend DRs. In other words, there is also a significant gap between presets which maximize winning by waiting for good players to make small mistakes and presets that maximize actively forcing mistakes out of even the most experienced players.

Ofcourse, the hands you run into are consistently very intimidating past 1300. Sometimes a player may get a lousy draw, but all the characters are above curve to topend.

In the current meta, you seldomly ever run into Piranas or Nightmare as you approach and pass 1300. This is what causes Roots and mono to be rampant in topend ELO right now.

As for Timber, I actually have 3 presets with Timber (Nightmare, Junkz and Frozn) that I use to go from 1000-1200. Skeelz would also be good with Timber except that Sasha/ Sandro/ Falkenstein are so topend (if it were not for the support synergy, I'd rate Timber equal or better than Michael). It's far more entertaining than playing with a killer preset and I love inducing ragequits. Of course he doesn't quite cut it as you approach 1300, although he's decent against mono Roots. A 7/6 or 8/5 Timber would be amazing.

Timber is definitely not "unplayable." He's actually really good (just not 1300+ good) in the right clan. When I use my Timber presets, I tend to win > 90% of the hands that I draw him. That's much better than any Nightmare, Frozn (post Kalindra banning) or Skeelz 5*.

1550+: This is a bot programmed by a MIT math/ CS major (future job- Microsoft)
1600+: This is a bot programmed by a Caltech math/ CS major (future job- US federal government)
1650+: This is a bot programmed by a Stanford math/ CS major (future job- Google)
1700+: This is a bot programmed by a Cal math/ CS major (future job- hacker)

As far as deck-building goes, pretty much what ghelas said. Additionally, at higher ratings, deck composition is less important than reading opponent's hands, knowing when to take risks, and the ability to do quick mental math. Case in point: I ran into tonton elo, a very solid and smart player, getting to the 1300s with a Roots/Rescue deck. Read that last bit again.

Anyway, I think that's why so many top players play mono-SOA-- against most hands, a "wait-and-see" strategy (some people call it "olding", apparently?) is perfectly viable and minimizes risk, especially against inexperienced players. But there's still plenty of room in the game for all sorts of deck designs, even when certain cards are unbanned.