Reverse-engineering Faction Power in Elemental: Reforged

 

I have spent a rather unreasonable amount of time trying to work out what the Faction Power number in Elemental: Reforged actually measures.

This started because my own faction often looked surprisingly weak in the Power rankings despite being economically and technologically well developed and quite capable of mobilising a large army quickly. So I was thinking: Why is the game looking at what I'm doing as being so much worse than what the AI is doing, when I cæearly am not having any trouble with the AI in the game.

What followed involved save-file comparisons, XML digging, controlled cheat experiments and eventually direct editing of the save files.

I have not succeeded in reconstructing the exact formula.

But I think we have learned quite a lot about what it is — and, perhaps more importantly, what it is not.

The first obvious theory: military strength

The Kingdom Report actually labels these values as Faction Military Ratings. But it is always the same as the number that is being displayed as Faction Power in the main game screen.

So the natural assumption was that Power was basically aggregate military strength. That certainly seemed plausible. My Power rose as I expanded and improved my military, while factions maintaining large standing armies tended to score highly.

But raw troop count quickly failed as an explanation. Some factions maintained many more formations than others without anything resembling proportional Power.

Then controlled experiments made things stranger. I created one ordinary unit, then made that same unit progressively ridiculous — eventually a level-30 monster with enormous Attack, HP and Accuracy. Its effect on Faction Power was tiny. I added 10 of them. Still tiny effect.

Adding ten ordinary formations also moved displayed Power surprisingly little, although it did materially improve how the other factions assessed my military diplomatically.

That was the first strong indication that:

Displayed Faction Power and the AI's comparative military assessment are not the same thing

The XML supports that distinction. The AI has its own internal combat-evaluation machinery, including AIComparativeCombatRating and explicit Combat Rating values attached to some abilities.

So whatever Faction Power is, it is not simply the number the diplomatic AI uses when asking “could this faction beat me up?”

Then city development looked like the answer 

Across several factions and saves, total city levels correlated extraordinarily well with Power.

For a while, this approximation looked almost embarrassingly good:

Faction Power ≈ 5 × sum of city levels

Regression repeatedly produced something close to that coefficient. Unfortunately, controlled testing broke it. I had two cities level simultaneously. After reload, Power did not change at all.

Other infrastructure tests were similarly disappointing:

• 1 Timber Mill: no change
• 5 Timber Mills: no change
• several additional outposts: no change
• additional Zone of Control: no change
• granting essentially all technologies: only about +4 Power

So city levels were apparently an excellent correlate of something, but not a simple direct five-points-per-level term.

That was an important lesson throughout this investigation: things that predict Power beautifully across factions do not necessarily cause it.

 

Tax, however, definitely matters

 

Tax produced one of the cleanest results of the entire investigation.

Starting repeatedly from the same reloaded save and changing only the tax rate gave:

 

 

Tax	Power
0%	124
10%	126
30%	130
50%	134
70%	138
90%	143

 From 0–70% the relationship is essentially perfect:

 +10 percentage points tax ≈ +2 Power 

 

So Faction Power cannot be purely military.

Whether tax itself enters the formula directly, or whether the game immediately derives some economic value from it, I cannot prove. But the causal effect is real.

Power is also recalculated on load 

 

 

Another major complication is that the game does not always preserve the displayed rating consistently between a live session and a reload.

One especially clear example that I observed was:

• Play session displayed power: 140
• Save and reload: 126

 

Reloading again remained 126. Closing the game entirely and loading again also produced 126.

Later, while reverse-engineering one the the other factions, I found the actual serialized faction-level Power field in the save.

I changed that faction's (Pariden's) stored value from roughly 80 to 5000.

The game loaded and displayed 76 anyway.

Saving immediately afterwards showed that the engine had rewritten the stored field back to 76.

So the save contains a cached Power result, but the engine recalculates it when loading.

That means all controlled comparisons need to use equivalent freshly loaded states. Comparing a live-session value with a reloaded value can be completely misleading.

Pariden then became the best natural experiment

The really interesting faction turned out to be Pariden.

Across 65 turns

 

Season	Power
292	76
323	227
342	991
357	2197

 

That was an extraordinary increase. But also one that I couldn't reconcile with what I saw ingame. Pariden was not stronger than me. Pariden was behind me on basivally all metrics that matter: Tech, Gildar, Mana, etc. etc.

So I reconstructed Pariden's state across all four saves. The obvious empire variables do not remotely explain it.

Pariden keeps the same eight cities throughout. Their combined city level remains 18 throughout.

Population rises only:

702 → 869 → 894 → 939

Constructed city improvements:

67 → 69 → 71 → 71

Military unit objects:

85 → 93 → 100 → 100

Represented soldiers:

245 → 274 → 298 → 296

And the last interval is the killer:

Power: 991 → 2197

While Pariden has the same number of unit objects, slightly fewer represented soldiers, exactly the same total city level, no net increase in city improvements, and only modest growth in ordinary resources and infrastructure.

Among 98 exact surviving units (clearly identifiable from Unit IDs in the save games) from Season 342 to 357:

• 70 did not gain a level
• 27 gained one
• 1 gained two
• only 5 changed equipment

Yet Power increased by 1,206.

That effectively killed any simple theory based on current visible military strength.

I then tested population directly

Pariden's real population at Season 292 (the season I used as baseline for all my Pariden tests because it was the earliest save I had where I hadn't noticed Pariden's massive growth yet) was 702.

I edited the actual population of one city so that the national population became:

702 → 779

The game correctly displayed 779.

Faction Power remained:

76

So population is not simply being converted directly into Power at any large coefficient.

Population may still contribute indirectly or at a very small weight, but it clearly cannot explain Pariden's runaway score.

Formation size also disappointed

I changed ten existing Pariden Parties from three soldiers to four.

The edit was accepted: Pariden's monthly deficit became noticeably worse because it was paying wages for the additional soldiers.

Power nevertheless remained:

76

So the economy recognised the extra soldiers, but displayed Power did not visibly care at that scale.

Equipment definitely matters — but not nearly enough

Equipment finally gave us a clean military effect.

Giving ten suitable Pariden formations the endgame Veinseeker axe changed Power:

76 → 78

Using weaker but still powerful weapons gave:

Pendulum: 77
Sunderer: 77

Compared to:
Veinseeker: 78

An armour-only upgrade on the same ten formations remained:

76

So offensive equipment absolutely can affect Faction Power.

But the magnitude is revealing.

Those ten Veinseeker swaps represented a very substantial improvement in the offensive strength of the modified troops — roughly a 14% increase in our whole-army equipment-derived Pierce Attack value for Pariden — yet Power only rose about 2.6%.

Meanwhile Pariden's historical Power rose hundreds or thousands of points.

So weapons are unquestionably one input, but they are not the missing explanation.

 

I then widened the search to the whole empire

At this point I was more or less ready to throw the whole project into the flames of Mordor, but I thought I'd attack the problem from just one more angle. So I then compared city buildings, mines, clay pits, crystal quarries, iron, horses, magic shards, Mana, Gildar, Metal, Crystal, Fame, research and quest-state progression across the same Pariden saves and against other factions.

Nothing obvious tracks Pariden's explosion.

For example, from Season 342 to 357 Pariden gains only modest strategic infrastructure. Other factions gain similar numbers of mines or shards without their Power exploding.

Pariden's Fame rises sharply — but Resoln gains roughly as much Fame while its Power actually falls.

Pariden's economy becomes worse, not better: deeper Gildar deficit and lower Mana while Power more than doubles.

There is simply no visible empire statistic we have found that behaves remotely like the Power number.

So what is Faction Power?

The safest conclusion I can offer is:

Faction Power is a composite rating with at least economic and military inputs, but the dominant late-game behaviour remains unexplained.

I have experimentally demonstrated that:

• tax rate affects it;
• offensive military equipment affects it;
• it is recalculated on load;
• the diplomatic AI's military assessment is a separate system.

I have strongly weakened simple explanations based on:

• population;
• city count;
• city levels;
• ordinary buildings;
• outposts;
• Zone of Control;
• technology;
• raw troop count;
• formation size;
• treasury/resources;
• visible equipment quantity;
• straightforward current military combat strength.

And Pariden - at least in me test game - remains the unsolved problem.

Some important state may exist that I have not decoded: generated AI unit designs, historical/cumulative military counters, victory data, cached combat evaluations, stack-level objects, or some other runtime state.

It is also entirely possible that there is simply a bug somewhere in the rating system — for example something being accumulated or counted repeatedly when it should not be.

With the saves available, I do not think there is enough information left to distinguish those possibilities confidently.

So after far too many hours, I am stopping one step short of The Big Formula™.

But at least I am reasonably confident of one thing:

Faction Power should not be interpreted as a straightforward measure of how strong a faction actually is.

It measures something real. Several real things, apparently.

It just occasionally decides that Pariden is worth 2,197 while owning essentially the same empire and army that recently rated 991.

And at that point I think the burden of explanation transfers back to Stardock.  

Thanks