Betfair Zero house edge Blackjack & single deck BJ

In Betfair casino, there's this Zero Blackjack game which Betfair claims to have zero house edge and sources online differ on the house advantage. The rules are as follow:
1. Dealer always peek for BJ
2. 8 decks
3. double after splitting allowed
4. no re-splitting
5. player CANNOT split Aces
6. double down on hard totals of 8/9/10/11
7. suited BJ pays 2-1, unsuited BJ pays 3-2 (all BJ are considered to be equal rank, ie. if you have suited BJ & dealer has unsited BJ or vice versa, it's a push)
8. 5-cards-21 is an automatic winner & pays 2-1 (applies to 5-cards-21 only, 6-cards-21 / 7-cards-21 is considered to be a 21 point hand)

I wonder what's the house edge is, various sources online differ slightly (ranging from a very small player's advantage to house advantage of 0.02%)

Betfair also offers a single deck BJ game with the following rules:
1. Dealer does not peek for BJ
2. no re-splitting, no double-down after splitting
3. BJ pays 3-2
4. double down on 9-11 (incl. A-8 & A-9)
The house edge on this game is about 0.22%.

As with online casino, cards are shuffled after each round, but Betfair allows player to play up to 3 spots in both of these games. I wonder how much does playing multi-hand cuts down the house edge assume you have a computer while playing (not unreasonable given it's an online casino) & can play perfectly and utilize all information (ie. cards in other hands)?

The house edge for the 8 deck game you are looking at is 0.003%
Minimal but still a "House Edge".

Ok, thanks for the input. Another different number, wizardsofodds had the house edge at 0.02%.
More fine points on the rules:
dealer stand on all 17's
when splitting, the additional card is dealt to the 2 new hands immediately (normally, you would play through the first hand before the 2nd hand gets its 2nd card dealt).

I computed the EV (see URL) for each combination of initial 2 cards (these numbers are after the dealer peeked for BJ), first column is the dealer upcard, 2nd being the player's initial 2 cards (1=Ace). 3rd column is the EV for standing, 4th column is the EV for hitting, 5th column is EV for double down, 6th column is EV for splitting. NA means in 5th or 6th column means this action is not allowed.

I'm not sure if I implemented 5-card-21 pays 2-1 correctly.

The exact top-of-the-deck EV using a Total Dependent Strategy for the player is:

TDEV: 0.017439820602692%

Using a Composition Dependent strategy the player advantage increases to:

CDEV: 0.02633831413312%

I'm not sure where GMan got his numbers but while they're close they're incorrect. Mine are from my Combinatorial Analyzer which handles all of the rules you've mentioned.

8 deck shoe with S17: -0.57%
Double on 8-9-10-11 only: -0.089%
DAS WHEN NO RESPLITS 8 decks: +0.086%
Only one pair split allowed: 0%
No Ace split: -0.18%
5 cards 21 pays 2:1 : +0.2%
Suited BJ pays 2:1 : +0.55%

Total : -0.003%

Can you show your numbers please ?

"8 deck shoe with S17: -0.57%
Double on 8-9-10-11 only: -0.089%
DAS WHEN NO RESPLITS 8 decks: +0.086%
Only one pair split allowed: 0%
No Ace split: -0.18%
5 cards 21 pays 2:1 : +0.2%
Suited BJ pays 2:1 : +0.55%

Total : -0.003%

Can you show your numbers please ?"

Nope - I do it all at once as it should be :) There are rule-rule interactions that you can't account for by going one at a time. Your number is close but it's an estimate. Not meaning to be rude or insulting, but just clarifying.

After verification, it seems that the source where my numbers came from wasn't 100% accurate. The main difference being for the 8 deck suite BJ giving 0.55% when in fact it should be 0.56575% (say 0.566), an additional player advantage of 0.016.This alone will bring the advantage from -0.003% to 0.013 making it slighly favorable for the player.

Thanks for pointing this out !

Would be appreciated if MGP can comment on the correctness or incorrectness of EVs for each initial 2 card combination I post earlier since you have performed combinatorics analysis. My calculated house edge is very close to yours and also have a players' positive EV listed. Unfortunately, anytime a casino game is known to have a players' edge in its 'default' state, one cannot help but want to scrutinize it (how can the casino give its $$ away). Especially since the strategy for game is not tough to learn and even total dependent strategy can turn a positive EV along with the fact sources on the internet does not seem to have a consensus as to the house edge of the game (from house edge of 0.02% to players' edge of 0.06%).

Thanks.

Last but not least, simulation of 500 million hands comes up with house edge of 0.04%. Of course, it could be the fact of the PRNG. I accumulated various dealer totals (17 thru 21 & bust) and the frequency match the theoretical combinatorial analysis. But then, PRNG maybe faulty in other ways. Random numbers come from Windows library function, CryptGenRandom rather than C++ default random number generator.