Avoid Ray for week 2 (vs NV & TSM). Both NV and TSM gave away a low average number of deaths in week 1. I'm considering this a risky pick, despite the whole APX team performing brilliantly in week 1.
Not convinced by TL & Piglet (vs CLG & FOX). Although CLG gave away the highest average number of deaths per game in NA, it was against two strong teams (TSM and APX). And even though there's a chance you may get decent points from a scrappy game vs CLG, I wouldn't come to expect that from CLG every week. You may also just as likely get shut out given TL's current form, and FOX don't give away that many deaths.
That leaves, Perkz, Reignover & Ryu (from your poll).
Regarding Reignover (vs NRG & TSM): IMT did ok in week 1, and I would probably play him in a First 2 Games format. But while the NA region has the better average deaths per game, EU has the greatest extremes.
For a Best Game format, you should be looking at potential for extreme number of deaths, rather than potential for a higher consistent performance. NRG have an above average deaths per game, but ROC & GIA have way more (and what's even more; they both held the 2nd highest and highest average number of deaths per game in EU last split too, respectively).
That leaves you either Perkz (vs ROC & SPY) or Ryu (vs GIA & FNC).
Perkz didn't really perform that great in week 1, even against GIA, so I'd go with Ryu. (Or Jankos, like you've done already. Jankos had a 94% Kill Participation, whereas Ryu had only a 77% KP in week 1)
Deaths Per Game
- Above Average
EU 18.50 GIA
EU 17.00 ROC
NA 16.40 CLG
NA 16.25 TL
NA 14.80 IMT
NA 14.75 P1
NA 13.60 NRG
EU 12.50 UOL
EU 12.50 OG
EU 12.50 SPY
- NA Average 12.45
- NA+EU Average 11.70
- EU Average 10.95
- Below Average
NA 10.80 APX
NA 10.33 C9
NA 09.40 NV
NA 09.40 FOX
NA 08.75 TSM
EU 08.50 H2K
EU 08.50 VIT
EU 07.00 FNC
EU 06.50 G2
EU 06.00 S04
Keep in mind, these are stats after just 1 week. Although the data is slightly better than we used to get from the best of one format, it may still change greatly in the next week or two.