Hello, everyone!  It’s Micah and Flash, back with another installment of our series investigating and analyzing the data collected as part of the Eternal Draft Project.  Today we’ll be going over various strange things in the data and something that many readers have been calling for:  Comparing how certain cards fair in different factions!

FLASH

So now that we’ve taken a nice long look at the data and went through the best and worst 15 cards, the next thing to do is try and make some sense of the data! Once we have some concepts or theories, we can then run further analysis to test out these theories! For today, I’m going to explore 3 things:

-Comparing the card’s performance when it’s on-faction, splashed or in a 4F/5F greed pile
-Thinking about synergistic effects
-Exploring why lifesteal units are so good!

On-faction vs Splash vs 4F/5F Performance

One nice point that was brought up multiple times on the previous reddit threads was looking at the performance of a card when it’s on-faction, on the splash and when it’s in 4F/5F decks. Coincidentally, this was also something that I was interested in because there were multiple cards that I felt performed better when they were on the splash (e.g. Highwind Glider). In contrast, there were also cards that we rationalized were poor performers because drafters were all too willing to splash for it despite a shaky power base (e.g. Smuggler’s Stash).

For this analysis, I simply took my model, split the card of interest into 3 separate regressors based on the factions of the deck and re-ran the model. I did not split all the cards simultaneously because that would likely introduce a huge amount of overfitting. All the results that are being discussed, together with their errors, can be found on the Google sheet (under Card Breakdowns).

Awakened Student
(On-faction WRM: 1.48, Splashed WRM: -0.23, 4F/5F WRM: 0.67)
As any good scientist would tell you, before running a new model on new data, you should always test on known data to make sure the model works. For this purpose, I’ve chosen the card Awakened Student. This card is probably the best Combrei common, but it is much weaker on the splash because you aren’t able to get it down consistently and quickly enough. The stats do back up this claim, with a great WRM when it is on-faction, but bad WRMs in the other 2 cases. Some 4F/5F piles could still be base Combrei and have enough fixing to support an Awakened Student, which might explain the slight improvement for 4F/5F WRM.

Highwind Glider
(On-faction WRM: -8.56, Splashed WRM: -1.82, 4F/5F WRM: -2.86)
Upon ensuring this model works, Highwind Glider was the first card I really wanted to test this model on. As I said in Part 2, Highwind Glider would have made a great card if it was in a more durdly faction pairing. However, Hooru does not have the units to back up such a durdly unit and thus, I would expect it to shine much more when it’s on the splash or in a 4F/5F greed pile. This is exactly what we see with our stats, a horrendous -8.56 WRM when played in base Hooru decks, and a much more reasonable -1.82/-2.86 in other decks. The consistent negative ratings do suggest that despite being better as a splash, Highwind Glider is still slightly overrated.

Hunter’s Harpoon
(On-faction WRM: -5.24, Splashed WRM: -0.33, 4F/5F WRM(Tiny sample): -5.58)
Hunter’s Harpoon was another interesting card. I felt that this card was pretty bad whereas Finkel argued that it was simply overrated and people were splashing it when they shouldn’t. Well, the stats showed that both of us were wrong. Hunter’s Harpoon actually performs best on the splash and in retrospect, this does make sense. I panned Hunter’s Harpoon because I felt that it didn’t synergize well with Skycrag’s aggro plan and was easily blown out by removal. However, if Harpoon was splashed in a more value-orientated deck, in decks with big units (to abuse the quickdraw) or units with aegis (to prevent removal blowout), it can easily take over the game. These stats actually makes complete sense to me now, and definitely made me rethink Hunter’s Harpoon in general.

Smuggler’s Stash
(On-faction WRM: 3.29, Splashed WRM: -6.03, 4F/5F WRM (Tiny sample) -0.49)
Upon looking at the previous 2 cards, I was actually beginning to doubt my theory about drafters being too greedy with their splash. Smuggler’s Stash came in the nick of time to rescue the theory with a shocking 3.29 WRM when its on-faction and a terrible -6.03 WRM when it’s splashed. This data does back up the claim that Smuggler’s Stash is still a great card, but people are just too greedy with splashing it.

When I first coded this bit, I didn’t have high expectations and I expected noise to dominate most of the analysis. To my pleasure and surprise, the results were actually convincing and clean enough to showcase the differences in WRM. It backed up most of my theories, while also providing some unexpected results and generating new theories. If you have any cards that you would like me to do this analysis on, do let me know in the reddit thread. (Note: I can’t do this for most rares since the sample size is way too small.)

Thinking about Synergistic Effects

Another thought-provoking card in the top lists was Icebow. As I alluded to in the article (and pointed out by many on the reddit thread), a likely significant contributory factor of Icebow’s great rating was the power of its synergy with specific cards and effects (namely, infiltrate and deadly). Unfortunately, we don’t have enough data to model interaction effects, but I came up with an ingenious (if I may say so myself) solution. Given that most of the synergistic units were only found in Feln and Elysian, a good way to detect the effects would be to compare the relative WRM of Icebow in each faction pairing.
(Feln WRM: 3.11, Elysian WRM: 2.90, Hooru WRM: 0.88, Skycrag WRM: 0.01)
The results completely backed up our theory in this case! With a clear distinction between Feln and Elysian vs Hooru and Skycrag, we can reasonably conclude that Icebow performs significantly better with synergistic cards while being a pretty mediocre weapon in other decks.

Why is lifesteal so great?

One persistent trend that I noticed in the data was that lifesteal units generally outshone their “equivalents” significantly (Cabal Cutthroat vs Lethrai Ranger, Xenan Destroyer vs Ravenous Thornbeast). So I decided to set about exploring this.
Initially, my suspicions were that these lifesteal units were actually NOT truly better than their non-lifesteal counterparts, but their WRM was been buffed due to a transference of WRM from lifeforce decks. I have always felt that lifeforce was a do-able, slightly below par, drafting archetype, but most players hate it because they force it too hard when it’s not open. As such, when people crash and burn with lifeforce, I think that they often end up with some payoff (given that this was probably what enticed them initially) but not enough enablers. Since I’m not modeling interaction effects, what this means is that the win rate effects from good lifeforce decks (with both enablers and payoffs) will mostly get assigned to the enablers, since the model sees bad lifeforce decks with just payoffs. This theory sounds about right to me, so I decided to test it similar to Icebow. Given lifeforce is a Xenan archetype, I looked at the WRM for these cards when they are in Xenan decks specifically.

 
Card Played in Times Played WR Modifier Error
Cabal Cutthroat Xenan 9 3.90 1.34
Argenport 12 3.10 1.16
Xenan Destroyer Xenan 26 0.19 2.79
Argenport 36 6.02 2.37
4F/5F 13 2.01 3.95
Skeeter Xenan 21 -1.86 1.76
Argenport 30 2.00 1.47
4F/5F 15 -0.08 2.08

Cabal Cutthroat did show some promise in supporting this theory but 3.9 is not significantly larger than 3.1. However, Xenan destroyer was quick to shut this lifeforce theory down with only 0.19 WRM and a staggering 6.02 WRM in Argenport.
This seems to indicate that my initial theory of Xenan Destroyer having a good win rate due to synergy with weapons is closer to the truth (since most of the good weapons are in Argenport). Running the same analysis on Skeeter shows a similar trend. This seems to suggest that the high win rate of lifesteal units arises because of the weapons synergy. By suiting up a giant lifesteal unit, this completely shuts the window on your opponent being able to race you and forces them to find a way to answer your threat. Without lifesteal, weapons would still be scary, but it would be possible to play around, such as chump blocking the giant threat and racing or going wide around it.

Micah

Splashing for Shadow Removal

Many of the content creators that I watch reccomend splashing Shadow for its hard removal spells like Execute and Annihilate.  The implication being that it will make your deck better.  Let’s take a look at the data!

 
Card Played in Times Played WR Modifier Error
Suffocate Primary 20 0.04 1.44
Splash 13 -2.03 1.78
4F/5F 0 0.00 0.00
Annihilate Primary 39 3.44 2.38
Splash 22 0.48 3.17
4F/5F 6 0.55 6.07
Execute Primary 106 3.99 0.55
Splash 45 -0.62 0.84
4F/5F 19 2.57 1.29

And the answer is that it depends!  All three cards are much stronger when Shadow is one of the decks primary factions.   Now, this doesn’t tell us that you shouldn’t be splashing for hard removal, but you need to be extremely sure that it isn’t compromising your powerbase.  Another possible interpretation is that if you need to splash for removal you are not drafting the open faction/archetype for your draft position.

So a Priest and a Camel walk into a bar…

One of the stranger things we noticed while sifting through the data was the extremely high positive WR Modifier for Sanctuary Priest and Amaran Camel.  As we discussed our various theories we came to similar conclusions.  They are both reasonable blockers and incidental lifegain is a large advantage in race and board stall situations.

I can already hear the screams of, “But Micah, a 1/2 can’t block anything!”  But it is pretty good at double blocking.  The majority of the time a player will choose to kill the larger of the two units you are double blocking with, leaving this unassuming priest to sit around and generate value.

Similarly, the camel seems to be great in decks with many fliers.  It will block a stranger every turn and gain 1 life per turn cycle (or more with additional card draw).  That is effectively 3 life gained per turn, which can be hard to overcome.

 
Card Played in Times Played WR Modifier Error
Sancutary Priest Xenan 18 1.54 0.29
Others 8 1.41 0.44
4F/5F 6 4.73 0.50
Amaran Camel Elysian 14 4.10 2.11
Xenan 9 0.46 2.63

So you want to play a 5F greed pile?

Besides 3 Yeti Spies and 3 Trailblaze, you should probably also grab every Amber Acolyte and Amaran Archeologist that you see.  And while this may be an obvious point, its always nice to see that your assumptions are backed up by the data.

 
Card Played in Times Played WR Modifier Error
Amber Acolyte Primary 185 -0.98 0.18
Splash 5 -0.90 1.10
4F/5F 20 1.09 0.55
Amaran Archeologist Primary 146 0.96 0.92
Splash 3 0.11 6.42
4F/5F 12 2.69 3.21

That’s it for this week everyone!  If you haven’t been submitting your draft decks to the Eternal Draft Project, please start!  Follow the link below to start contributing.

Submit your decks today!

LEAVE A REPLY

Please enter your comment!
Please enter your name here