Desert war pairings are not random

Some weeks the desert war draw looks winnable, other weeks it is hopeless, and the game never explains why. Analysing collected matchup records shows that pairing works by matching an internal value best described as strength: across several hundred observed pairs, the two sides' values differed by a median of roughly one percent. That is not coincidence.

This page ranks that value for the alliances we have collected, so you can see which band your alliance sits in and what kind of opponents that band produces.

What drives the value

What the data supports: the value rises with the strength of the roster registered for that round (when the same alliance changes its roster, it moves about 1.4× the change in registered power). But each server applies its own scaling, so identical power and headcount on different servers can produce values dozens of times apart — raw values only mean something within the same pool. Between alliances on the same server, the roster's total hero power explained the differences best. No exact formula is confirmed yet; the full write-up is in Desert War matchmaking, reverse-engineered.

One property matters more than the rest: the value does not respond to results. Winning does not raise it and losing does not lower it; it is recomputed each round from the registered roster.

Which means tanking does not work

The alliance-chat theory — throw this round and next week's draw gets easier — does not survive the data. Comparing value changes after wins against value changes after losses produced no statistically meaningful difference.

What actually moves the number is who registers. Bring more strong accounts and it rises; sit your core out and it falls. If you want to influence the draw, the roster is the lever, not the result.

The opponent pool follows migration groups

Opponents come from a contiguous block of server numbers, broadly aligned with migration groups, and those block boundaries move from week to week rather than staying fixed — one week the pool ran from server 3 to roughly 420, another week it extended to 676. No pairing was observed crossing a boundary.

That is why no permanent table of "our server always faces this range" can be published, and why this page recomputes the grouping from collected data each week.

Using it in practice

Prepare before the draw lands. Knowing your band shows you the power level of the alliances sitting in it. Pairings are generally settled the day before, which leaves room to adjust registration first.

Treat a five percent power gap as the line. Because matchmaking equalises the value, a real power difference beyond about five percent inside the same band is a genuine gap. Inside it, formation quality and turnout decide the match.

Limits of this analysis

None of this is published by the developer; it is regularity observed in collected data. Two earlier conclusions drawn during this work were later withdrawn as the sample grew, and a game update could invalidate what is written here. If it does not match your own alliance's experience, check the actual pairings in desert war records.