Deep Dives

Root Faction Scoring Models

Recently, I started keeping track of the round-by-round points for each factions after Rk9 showed interest. It was a stat that I hadn’t really tracked before and I’m not sure why I didn’t because it’s so fascinating. Looking back of all the tournaments, I found that there were two sets of data: the scoring paths from AdSet and the scoring paths from the Plus One Pool Draft. These get delved into by faction in the sections below. All data on the AdSet faction includes only the first two rounds of the 2021 Winter Tournament.

For all the data, I found the average scoring model and created a best-fit linear line with a high R2 number. The R2 represents how closely the movement of the dependent variable is correlated to the independent variable. Since the R2 for these numbers were relatively high, a linear model is appropriate to predict the expected scoring. While numbers are great, there are always exceptions especially in a game like Root that has so much player input.

Check out the Tournament Stats page to see the stats from other tournaments! Or check out The Woodland for most posts on Root!

I was completed fascinated by the fact that the factions, on average, stick so close together when counting all plays in AdSet. It’s no surprise to me that when considering all plays the woodland alliance is the one to take the lead. This resonates with the win rate we typically see in tournaments. Also leaning on the quicker to victory side is the Eyrie, Duchy, Riverfolk, Warlord, and Keepers. The slower side consists of the Lizards, Corvids, Cats, and Vagabond. The Vagabond being the lagging faction really surprised me. I thought that it would be on a faster clock then it was.

However, the faction comparison look drastically different from what they were in the Plus One Pool Draft. I was shocked that the Woodland Alliance had such a major shift. Lizards and cats are falling behind from what it use to be as well! Just all around interesting.

The graphs that only look at game where those factions win show something completely different. After playtesting the new marauder factions, I heard the saying “The [Keepers in Iron] are over-tuned”. I always assumed this meant that you had to play them in just the right way to win. While they may be over tuned, they are not the faction that is the most overturned. That honor goes to the Riverfolk Company!  I also find it interesting that the Vagabond and Cats are generally on the quicker side when it comes to finding a victory.

And poor corvids are hanging way out in left field. Which is a drastic difference from what we saw in the Plus One Pool Draft. Although that is going off a lot less data… The comparison still feels very stark. The AdSet draft style seems to have brought the factions more into balance.

The base factions

marquise de cat

In both AdSet and the Plus One Pool draft, we have seen numerous plays with the cats. It leads to these fun and sprawling graphs. But one scoring model is not data but the average. And for some reason, I made it very purple. The model between the two draft methods produced two slopes with a difference of 0.1183 VP per Round. That’s barely anything!

We haven’t seen many wins for the cats in the first two round of the 2021 tournament. Give that I only had 3 sets of data, I’m not completely comfortable with say it’s a good representation.  I do find it interesting thought that all the Plus One Pool Draft wins come to a concise point at Round 7 and 8. But even here, there is no predetermined winning path for the cats.

What I found to be the most interesting about the Cats is how close the data follows each other. The two models of data that came from all plays is extremely close. As is the two models of data that came from the plays where the cats one. I did run an t-test to compare the sets of data and found no statistical significance between the sets of data. I guess I could have smashed these ones together after all!

The eyrie dynasties

The Eyrie graphs are some of my favorites because they look so wonky. This is mainly because the Eyrie are the only faction that can lose points. Being the only faction to lose points makes them have a unique graph type although their model is still linear. We haven’t seen as many Eyrie plays using the AdSet draft as we have of the Plus One Pool Draft. However, the AdSet models are significantly less sporadic with an increased 0.7 VP/round in the average model slope.

There is more consistency in the AdSet winning plays model then there is in the Plus One Pool winning play models. It reminds me of a bouquet as they all start out on roughly the same trajectory and expand from there. The Plus One Pool draft has produced a wide array of models. The consistency though is really reflected in the difference of slopes between the two average models. Once again, there is an increase of 0.7 per round. 

Looking at this graph, I pretty much assumed that there would be a statistical significance between the two sets of data. And I was correct! There is enough variation between the two sets of models that they cannot be compared. I do think Eyrie has gotten a significant boost moving after transition to the AdSet Draft style. The freedom of placement has become a major benefit to the Eyrie.

the woodland alliance

I think the Woodland Alliance graphs are probably the strangest looking graphs out of all the factions. The AdSet data have a very odd curve to it which makes me think that an exponential model would fit the data better. However, the linear model had enough accuracy that I felt comfortable keeping all the data linear.  The Plus One Pool draft models were a little more scattered. There was a significant difference between the two slops of these models. I was surprised to see how many rounds it took the Woodland alliance to win using this draft method.

Once again, the AdSet draft models have that curve. This really reflect a similar playstyle among the Woodland Alliance wins as the scores are low in the beginning few rounds and rapidly escalate over time. The large jump in slope is also seen when comparing the winning models of the two draft types. I don’t love that the average model of the Plus One Pool draft dips down in round 10. While this heavily influences the slope of the scored, there are multiple data points which shouldn’t be ignored.

Not going to lie, I did not expect the t-test to result in a statistically significant for the Woodland Alliance when I first started this project. But here we are. What’s even more surprising is how closely the lines from each draft stick together. The line models for AdSet plays and AdSet wins are nearly identical. It baffles me. This is why the Woodland Alliance has always done so well. If nothing else, they are extremely consistent.

the vagabond

The Vagabond has been making a lot of appearance is 2021 Winter Tournament so there is a lot more data models for AdSet over the Plus One Pool draft. But both sets of data look extremely similar in distribution. The AdSet slope did increase from the Plus One Pool draft. There are a lot of changes that could affect the Vagabonds scoring. The ability to curate your starting hand is highly beneficial but despot infamy is a bit of a setback. I also wonder if the new meta of not hitting the Vagabond every turn is also allowing the scoring to stay on a similar trajectory.

The winning play models are, again, very similar. The main difference being that Vagabond in AdSet typically starts off at a high points value and increases from there. There is also a higher linear relationship on the average model of the AdSet data.

Looking between the similar sets of data, the t-test found no statistical significance between the two types of draft. This blows my mind. I thought we would see a lot more variance with the change from normal infamy to despot infamy. There must have been a shift in table attitude to accommodate this difference. That’s the only thing that I can think of. 

the riverfolk expansion

the riverfolk company

The Riverfolk Company has an interesting parallel going between these two sets of data.  The Plus One Pool draft shows a tighter grouping of its model data while the AdSet draft has a wider variance. The difference between these slopes is almost negligibly small. It’s interesting how these things balance out, right? That paths look nearly identical!

We have not seen as many wins using the AdSet Drafts as we have seen in the Plus One Pool Drafts. The slope difference between the two lines do show a great amount of variation. The winning games for the Riverfolk in AdSet have shown a higher scoring per round that previously in the Plus One Pool draft. I wonder if a shift in mentality about the rate in which the Riverfolk Company score would help boost the amount of wins we’re seeing.

It’s very interesting to me that the wining strategy in the plus one pool draft was to start almost behind the average score of all plays and rapidly ramp up the scoring temp. In AdSet, you score a lot to begin with and keep the pace up. Although, this line feels a bit more inline with the two average all play models. Comparing the sets of data for each draft type was found to not be statistically significant by the t-test.

the lizard cult

Oh, my dear, dear lizard cult. Just like the chaos you cause in the world; you cause to your own scoring models. These graphs are all over the place regardless of the draft style. I do find it interesting that the Plus One Pool Draft has more consistency over their model sets. It’s got that bouquet vibe going where the AdSet draft feels more spread out and chaotic.

So… I only had one win from AdSet to go off for this comparison so I’m not sure it there is much to compare. That being said, only having one win also stands out in it’s own way. What’s even more interesting is that the Lizards cults average scoring model slope decreases from the Plus One Pool Draft. This is the opposite of many of the other trends that we have been seeing. But, basing this off of only one set of data doesn’t mean that much.

While there are some difference, the slopes of the average models are relatively similar. After putting the data through a t-test, I found that there was no statistical significance between the two sets of data. On of the more interesting models on this graph is the Plus One Pool draft Winning plays model. It allows for no scoring earlier in the game and ramps up rapidly at a higher intensity than any of the other models.

The underworld expansion

the underground duchy

The Underground Duchy has been one of the most played factions in the tournaments. I believe this is because they are in quite a few of the draft and they are a great faction with a fairly decent win rate. The Duchy is tied for the second highest slope in AdSet but they only had the fifth highest slope in the Plus One Pool Draft. Looks like AdSet gave the Duchy a little boost.

Looking at the winning games, the duchy has seen more consistency in their games. I feel like the AdSet graph has shifted forward to  have more winning models reach 30 in earlier rounds. The duchy has landed right in the center of the win models against other factions. I feel like they have the most flexibility in the models to be able to pull out the win.

This is the set of data that has me losing my confidence in the t-test. According to the T-test, there is not a statistically significance between the sets of data. But, just by looking at the graph, that makes absolutely no sense. There is a 0.7307 difference in slope between the average all play models and a 0.6496 difference in slope between the average winning play models. This is very similar to the Eyrie (which was deemed statistically significant) so I’m very confused.  It must be due to the fact that all models were considered and the noise was much less significant when putting all of the data together.

the corvid conspiracy

I feel like the Corvids scoring trajectory mimics the nature of the faction: random and utterly chaotic. There are scoring models on ends of the spectrum and veer off hard left and right. I find it fascinating that the graphs for the Plus One Pool Draft and the AdSet Draft give off a similar feeling while having a vastly different slope. The AdSet draft has an increased 0.4 points per round increase which can prove to be a major difference.

While the winning games produce few models for the corvids in AdSet, there are even fewer from the Plus One Pool Draft. Corvids have a large spread of rounds in which they achieved the win. In AdSet, we saw wins on round 6, 8, and 10. And it caused out average line to become very spikey. In the Plus One Pool Draft, our single win was at Round 7. I think this speaks to the flexibility that corvids has as they can burst early or do a slow build if the game remains on a slower pace.

The model comparison show a wide variation between the average models for all plays and for all wins. After running the t-test, these variations were found to be statistically significant. Which means that I cannot combine this data as they are not compatible with each other. One thing I find absolutely fascinating about the above graph is that the model of all plays in AdSet and the model for won games in AdSet are extremely close. I believe this is why we are seeing more Corvid wins in the 2021 Root Winter Tournament because AdSet really has helped bring Corvids into the game as a competitive faction.

the mauraders expansion

the lord of the hundreds

Since the Lord of the Hundreds is a new faction, I don’t have any data from the Plus One Pool draft to do a comparison on. But the Lord of the hundreds has been an extremely popular faction from the first two round of the 2021 Root Winter Tournament. Looking at all the plays, we can see the bouquet effect where the faction starts in a similar range before spreading out in various directions. 

Not only have we seen a lot of plays of the Lord of the Hundreds; we’ve seen a lot of wins as well! The Lord of the Hundreds have proven themselves to be a very steady faction with a lot of flexibility. This is reflected by the rounds in which we have been able to see wins. It has ranged from as early as round 6 to as late as round 10.

It’s intriguing that the average scoring models for all plays and winning plays run in parallel with each other. It would suggest that those who get more points in round one are more likely to have a wining play. There is a minute difference between the slopes so the only thing separating these plays is the points scored during the first round.

the keepers in iron

Since Keepers in Iron is a new faction, I don’t have any data from the Plus One Pool Draft to do a comparison on. The plays that we have seen so far have proven to be chaotic and varying. I think this is mainly attributed to players learning the faction. The Keepers have proven to be quite complex. If you’re curious to learn more about them, check out the Keepers in Iron guide under Deep Delves of the Game section!

Because we have not seen many badgers wins, I do not have much data. And I’m not sure that three sets of data can really make a statistically accurate average. But when has that ever stopped me? It’s absolutely fascinating that the badgers have the highest slope for their VP to round ratio when looking at the average winning game models. This would be even higher is we didn’t consider the one model that took 8 rounds to win.  

The Keepers in Iron have a vastly increased slope between the two different average models. But it’s not the largest difference that we’ve seen. I’m still amazed the 4.22 victory points per round slope. It’s insane for a faction to consistently scoring that quickly!!


The table below is the average slope for all plays and the winning plays of the models in the diagrams above.

Faction All Plays
All Plays
Plus One Pool
Winning Plays AdSet Winning Plays Plus One Pool
Marquise de Cat2.94592.82763.80564.0595
Eyrie Dynasties3.52662.82663.91923.2192
Woodland Alliance3.89672.76484.0122.8298
Riverfolk Company3.30463.2144.19443.7808
Lizard Cult3.30613.25383.38333.7556
Underground Duchy3.5282.79733.92673.2771
Corvid Conspiracy3.20492.82023.18384.6429
Lord of the Hundreds3.46053.5437
Keepers in Iron3.42674.2222

The table below is the calculating the difference in slope in each category by taking the slope AdSet minus the slope of the Plus One Pool draft.

FactionsAll PlaysWinning Plays
Marquise de Cat0.1183-0.2539
Eyrie Dynasties0.70.7
Woodland Alliance1.13191.1822
Riverfolk Company0.09060.4136
Lizard Cult0.0523-0.3723
Underground Duchy0.73070.6496
Corvid Conspiracy0.3847-1.4591

The table low is calculated the difference in slope between the winning plays model and the all plays model. This is kept separate by the type of draft.

FactionAdSetPlus One Pool
Marquise de Cat0.85971.2319
Eyrie Dynasties0.39260.3926
Woodland Alliance0.11530.065
Riverfolk Company0.88980.5668
Lizard Cult0.07720.5018
Underground Duchy0.39870.4798
Corvid Conspiracy-0.02111.8227
Lord of the Hundreds0.0832
Keepers in Iron0.7955

Leave a Reply